1pub const DNNL_RUNTIME_NONE: u32 = 0;
4pub const DNNL_RUNTIME_SEQ: u32 = 1;
5pub const DNNL_RUNTIME_OMP: u32 = 2;
6pub const DNNL_RUNTIME_TBB: u32 = 4;
7pub const DNNL_RUNTIME_THREADPOOL: u32 = 8;
8pub const DNNL_RUNTIME_OCL: u32 = 256;
9pub const DNNL_RUNTIME_SYCL: u32 = 512;
10pub const DNNL_RUNTIME_DPCPP: u32 = 512;
11pub const DNNL_VENDOR_NONE: u32 = 0;
12pub const DNNL_VENDOR_INTEL: u32 = 1;
13pub const DNNL_VENDOR_NVIDIA: u32 = 2;
14pub const DNNL_VENDOR_AMD: u32 = 4;
15pub const DNNL_VENDOR_GENERIC: u32 = 8;
16pub const DNNL_CPU_THREADING_RUNTIME: u32 = 4;
17pub const DNNL_CPU_RUNTIME: u32 = 512;
18pub const DNNL_GPU_RUNTIME: u32 = 512;
19pub const DNNL_GPU_VENDOR: u32 = 1;
20pub const BUILD_TRAINING: u32 = 1;
21pub const BUILD_INFERENCE: u32 = 0;
22pub const BUILD_PRIMITIVE_ALL: u32 = 1;
23pub const BUILD_BATCH_NORMALIZATION: u32 = 0;
24pub const BUILD_BINARY: u32 = 0;
25pub const BUILD_CONCAT: u32 = 0;
26pub const BUILD_CONVOLUTION: u32 = 0;
27pub const BUILD_DECONVOLUTION: u32 = 0;
28pub const BUILD_ELTWISE: u32 = 0;
29pub const BUILD_GROUP_NORMALIZATION: u32 = 0;
30pub const BUILD_INNER_PRODUCT: u32 = 0;
31pub const BUILD_LAYER_NORMALIZATION: u32 = 0;
32pub const BUILD_LRN: u32 = 0;
33pub const BUILD_MATMUL: u32 = 0;
34pub const BUILD_POOLING: u32 = 0;
35pub const BUILD_PRELU: u32 = 0;
36pub const BUILD_REDUCTION: u32 = 0;
37pub const BUILD_REORDER: u32 = 0;
38pub const BUILD_RESAMPLING: u32 = 0;
39pub const BUILD_RNN: u32 = 0;
40pub const BUILD_SDPA: u32 = 0;
41pub const BUILD_SHUFFLE: u32 = 0;
42pub const BUILD_SOFTMAX: u32 = 0;
43pub const BUILD_SUM: u32 = 0;
44pub const BUILD_PRIMITIVE_CPU_ISA_ALL: u32 = 1;
45pub const BUILD_SSE41: u32 = 0;
46pub const BUILD_AVX2: u32 = 0;
47pub const BUILD_AVX512: u32 = 0;
48pub const BUILD_AMX: u32 = 0;
49pub const BUILD_PRIMITIVE_GPU_ISA_ALL: u32 = 1;
50pub const BUILD_GEN9: u32 = 0;
51pub const BUILD_GEN11: u32 = 0;
52pub const BUILD_XELP: u32 = 0;
53pub const BUILD_XEHP: u32 = 0;
54pub const BUILD_XEHPG: u32 = 0;
55pub const BUILD_XEHPC: u32 = 0;
56pub const BUILD_XE2: u32 = 0;
57pub const BUILD_XE3: u32 = 0;
58pub const BUILD_GEMM_KERNELS_ALL: u32 = 1;
59pub const BUILD_GEMM_KERNELS_NONE: u32 = 0;
60pub const BUILD_GEMM_SSE41: u32 = 0;
61pub const BUILD_GEMM_AVX2: u32 = 0;
62pub const BUILD_GEMM_AVX512: u32 = 0;
63pub const DNNL_MAX_NDIMS: u32 = 12;
64pub const DNNL_VERSION_MAJOR: u32 = 3;
65pub const DNNL_VERSION_MINOR: u32 = 8;
66pub const DNNL_VERSION_PATCH: u32 = 1;
67pub const DNNL_ARG_UNDEF: u32 = 0;
68pub const DNNL_ARG_SRC_0: u32 = 1;
69pub const DNNL_ARG_SRC: u32 = 1;
70pub const DNNL_ARG_SRC_LAYER: u32 = 1;
71pub const DNNL_ARG_FROM: u32 = 1;
72pub const DNNL_ARG_SRC_1: u32 = 2;
73pub const DNNL_ARG_SRC_ITER: u32 = 2;
74pub const DNNL_ARG_SRC_2: u32 = 3;
75pub const DNNL_ARG_SRC_ITER_C: u32 = 3;
76pub const DNNL_ARG_SRC_3: u32 = 4;
77pub const DNNL_ARG_AUGRU_ATTENTION: u32 = 4;
78pub const DNNL_ARG_DST_0: u32 = 17;
79pub const DNNL_ARG_DST: u32 = 17;
80pub const DNNL_ARG_TO: u32 = 17;
81pub const DNNL_ARG_DST_LAYER: u32 = 17;
82pub const DNNL_ARG_DST_1: u32 = 18;
83pub const DNNL_ARG_DST_ITER: u32 = 18;
84pub const DNNL_ARG_DST_2: u32 = 19;
85pub const DNNL_ARG_DST_ITER_C: u32 = 19;
86pub const DNNL_ARG_WEIGHTS_0: u32 = 33;
87pub const DNNL_ARG_WEIGHTS: u32 = 33;
88pub const DNNL_ARG_WEIGHTS_LAYER: u32 = 33;
89pub const DNNL_ARG_WEIGHTS_1: u32 = 34;
90pub const DNNL_ARG_WEIGHTS_ITER: u32 = 34;
91pub const DNNL_ARG_WEIGHTS_2: u32 = 35;
92pub const DNNL_ARG_WEIGHTS_PEEPHOLE: u32 = 35;
93pub const DNNL_ARG_WEIGHTS_3: u32 = 36;
94pub const DNNL_ARG_WEIGHTS_PROJECTION: u32 = 36;
95pub const DNNL_ARG_BIAS: u32 = 41;
96pub const DNNL_ARG_REDUCE: u32 = 42;
97pub const DNNL_ARG_MEAN: u32 = 49;
98pub const DNNL_ARG_VARIANCE: u32 = 50;
99pub const DNNL_ARG_SCALE: u32 = 51;
100pub const DNNL_ARG_SHIFT: u32 = 52;
101pub const DNNL_ARG_WORKSPACE: u32 = 64;
102pub const DNNL_ARG_SCRATCHPAD: u32 = 80;
103pub const DNNL_ARG_DIFF_SRC_0: u32 = 129;
104pub const DNNL_ARG_DIFF_SRC: u32 = 129;
105pub const DNNL_ARG_DIFF_SRC_LAYER: u32 = 129;
106pub const DNNL_ARG_DIFF_SRC_1: u32 = 130;
107pub const DNNL_ARG_DIFF_SRC_ITER: u32 = 130;
108pub const DNNL_ARG_DIFF_SRC_2: u32 = 131;
109pub const DNNL_ARG_DIFF_SRC_ITER_C: u32 = 131;
110pub const DNNL_ARG_DIFF_SRC_3: u32 = 132;
111pub const DNNL_ARG_DIFF_AUGRU_ATTENTION: u32 = 132;
112pub const DNNL_ARG_DIFF_DST_0: u32 = 145;
113pub const DNNL_ARG_DIFF_DST: u32 = 145;
114pub const DNNL_ARG_DIFF_DST_LAYER: u32 = 145;
115pub const DNNL_ARG_DIFF_DST_1: u32 = 146;
116pub const DNNL_ARG_DIFF_DST_ITER: u32 = 146;
117pub const DNNL_ARG_DIFF_DST_2: u32 = 147;
118pub const DNNL_ARG_DIFF_DST_ITER_C: u32 = 147;
119pub const DNNL_ARG_DIFF_WEIGHTS_0: u32 = 161;
120pub const DNNL_ARG_DIFF_WEIGHTS: u32 = 161;
121pub const DNNL_ARG_DIFF_WEIGHTS_LAYER: u32 = 161;
122pub const DNNL_ARG_DIFF_WEIGHTS_1: u32 = 162;
123pub const DNNL_ARG_DIFF_WEIGHTS_ITER: u32 = 162;
124pub const DNNL_ARG_DIFF_WEIGHTS_2: u32 = 163;
125pub const DNNL_ARG_DIFF_WEIGHTS_PEEPHOLE: u32 = 163;
126pub const DNNL_ARG_DIFF_WEIGHTS_3: u32 = 164;
127pub const DNNL_ARG_DIFF_WEIGHTS_PROJECTION: u32 = 164;
128pub const DNNL_ARG_DIFF_BIAS: u32 = 169;
129pub const DNNL_ARG_DIFF_SCALE: u32 = 255;
130pub const DNNL_ARG_DIFF_SHIFT: u32 = 256;
131pub const DNNL_ARG_ATTR_ROUNDING_SEED: u32 = 508;
132pub const DNNL_ARG_ATTR_DROPOUT_MASK: u32 = 509;
133pub const DNNL_ARG_ATTR_DROPOUT_PROBABILITY: u32 = 510;
134pub const DNNL_ARG_ATTR_DROPOUT_SEED: u32 = 511;
135pub const DNNL_ARG_ATTR_OUTPUT_SCALES: u32 = 513;
136pub const DNNL_ARG_MULTIPLE_SRC: u32 = 1024;
137pub const DNNL_ARG_MULTIPLE_DST: u32 = 2048;
138pub const DNNL_ARG_ATTR_SCALES: u32 = 4096;
139pub const DNNL_ARG_ATTR_ZERO_POINTS: u32 = 8192;
140pub const DNNL_ARG_ATTR_POST_OP_DW: u32 = 16384;
141pub const DNNL_ARG_ATTR_MULTIPLE_POST_OP_BASE: u32 = 32768;
142pub const DNNL_JIT_PROFILE_NONE: u32 = 0;
143pub const DNNL_JIT_PROFILE_VTUNE: u32 = 1;
144pub const DNNL_JIT_PROFILE_LINUX_PERFMAP: u32 = 2;
145pub const DNNL_JIT_PROFILE_LINUX_JITDUMP: u32 = 4;
146pub const DNNL_JIT_PROFILE_LINUX_JITDUMP_USE_TSC: u32 = 8;
147pub const DNNL_JIT_PROFILE_LINUX_PERF: u32 = 6;
148pub const DNNL_GRAPH_UNKNOWN_NDIMS: i32 = -1;
149pub mod dnnl_status_t {
150 #[doc = " Status values returned by the library functions."]
151 pub type Type = ::std::os::raw::c_uint;
152 #[doc = " The operation was successful"]
153 pub const dnnl_success: Type = 0;
154 #[doc = " The operation failed due to an out-of-memory condition"]
155 pub const dnnl_out_of_memory: Type = 1;
156 #[doc = " The operation failed because of incorrect function arguments"]
157 pub const dnnl_invalid_arguments: Type = 2;
158 #[doc = " The operation failed because requested functionality is not implemented"]
159 pub const dnnl_unimplemented: Type = 3;
160 #[doc = " The last available implementation is reached"]
161 pub const dnnl_last_impl_reached: Type = 4;
162 #[doc = " Primitive or engine failed on execution"]
163 pub const dnnl_runtime_error: Type = 5;
164 #[doc = " Queried element is not required for given primitive"]
165 pub const dnnl_not_required: Type = 6;
166 #[doc = " The graph is not legitimate"]
167 pub const dnnl_invalid_graph: Type = 7;
168 #[doc = " The operation is not legitimate according to op schema"]
169 pub const dnnl_invalid_graph_op: Type = 8;
170 #[doc = " The shape cannot be inferred or compiled"]
171 pub const dnnl_invalid_shape: Type = 9;
172 #[doc = " The data type cannot be inferred or compiled"]
173 pub const dnnl_invalid_data_type: Type = 10;
174}
175pub mod dnnl_data_type_t {
176 #[doc = " Data type specification"]
177 pub type Type = ::std::os::raw::c_uint;
178 #[doc = " Undefined data type, used for empty memory descriptors."]
179 pub const dnnl_data_type_undef: Type = 0;
180 #[doc = " 16-bit/half-precision floating point."]
181 pub const dnnl_f16: Type = 1;
182 #[doc = " non-standard 16-bit (bfloat16 w/ 7 bit mantissa) floating point."]
183 pub const dnnl_bf16: Type = 2;
184 #[doc = " 32-bit/single-precision floating point."]
185 pub const dnnl_f32: Type = 3;
186 #[doc = " 32-bit signed integer."]
187 pub const dnnl_s32: Type = 4;
188 #[doc = " 8-bit signed integer."]
189 pub const dnnl_s8: Type = 5;
190 #[doc = " 8-bit unsigned integer."]
191 pub const dnnl_u8: Type = 6;
192 #[doc = " 64-bit/double-precision floating point."]
193 pub const dnnl_f64: Type = 7;
194 #[doc = " Boolean data type. Size is C++ implementation defined."]
195 pub const dnnl_boolean: Type = 8;
196 #[doc = " [OFP8 standard 8-bit floating-point](https://www.opencompute.org/documents/ocp-8-bit-floating-point-specification-ofp8-revision-1-0-2023-06-20-pdf)\n with a 5-bit exponent and a 2-bit mantissa."]
197 pub const dnnl_f8_e5m2: Type = 9;
198 #[doc = " [OFP8 standard 8-bit floating-point](https://www.opencompute.org/documents/ocp-8-bit-floating-point-specification-ofp8-revision-1-0-2023-06-20-pdf)\n with a 4-bit exponent and a 3-bit mantissa."]
199 pub const dnnl_f8_e4m3: Type = 10;
200 #[doc = " 4-bit signed integer."]
201 pub const dnnl_s4: Type = 11;
202 #[doc = " 4-bit unsigned integer."]
203 pub const dnnl_u4: Type = 12;
204 #[doc = " [MX-compliant 8-bit compliant scale data type](https://www.opencompute.org/documents/ocp-microscaling-formats-mx-v1-0-spec-final-pdf) with 8-bit exponent."]
205 pub const dnnl_e8m0: Type = 13;
206 #[doc = " [MX-compliant 4-bit float data type](https://www.opencompute.org/documents/ocp-microscaling-formats-mx-v1-0-spec-final-pdf) with 2-bit exponent and 1 bit mantissa."]
207 pub const dnnl_f4_e2m1: Type = 14;
208 #[doc = " 4-bit float data type with 3-bit exponent and 0 bit mantissa."]
209 pub const dnnl_f4_e3m0: Type = 15;
210 #[doc = " Parameter to allow internal only data_types without undefined behavior.\n This parameter is chosen to be valid for so long as sizeof(int) >= 2."]
211 pub const dnnl_data_type_max: Type = 32767;
212}
213#[doc = " A type to describe tensor dimension."]
214pub type dnnl_dim_t = i64;
215#[doc = " A type to describe tensor dimensions."]
216pub type dnnl_dims_t = [dnnl_dim_t; 12usize];
217pub mod dnnl_fpmath_mode_t {
218 #[doc = " Floating-point math mode"]
219 pub type Type = ::std::os::raw::c_uint;
220 #[doc = " Default behavior, no downconversions allowed"]
221 pub const dnnl_fpmath_mode_strict: Type = 0;
222 #[doc = " Implicit f32->bf16 conversions allowed"]
223 pub const dnnl_fpmath_mode_bf16: Type = 1;
224 #[doc = " Implicit f32->f16 conversions allowed"]
225 pub const dnnl_fpmath_mode_f16: Type = 2;
226 #[doc = " Implicit f32->f16, f32->tf32 or f32->bf16 conversions allowed"]
227 pub const dnnl_fpmath_mode_any: Type = 3;
228 #[doc = " Implicit f32->tf32 conversions allowed"]
229 pub const dnnl_fpmath_mode_tf32: Type = 4;
230}
231pub mod dnnl_accumulation_mode_t {
232 #[doc = " Accumulation mode"]
233 pub type Type = ::std::os::raw::c_uint;
234 #[doc = " Default behavior, f32/f64 for floating point computation, s32\n for integer"]
235 pub const dnnl_accumulation_mode_strict: Type = 0;
236 #[doc = " Same as strict but allows some partial accumulators to be\n rounded to src/dst datatype in memory."]
237 pub const dnnl_accumulation_mode_relaxed: Type = 1;
238 #[doc = " uses fastest implementation, could use src/dst datatype or\n wider datatype for accumulators"]
239 pub const dnnl_accumulation_mode_any: Type = 2;
240 #[doc = " use s32 accumulators during computation"]
241 pub const dnnl_accumulation_mode_s32: Type = 3;
242 #[doc = " use f32 accumulators during computation"]
243 pub const dnnl_accumulation_mode_f32: Type = 4;
244 #[doc = " use f16 accumulators during computation"]
245 pub const dnnl_accumulation_mode_f16: Type = 5;
246}
247pub mod dnnl_engine_kind_t {
248 #[doc = " @brief Kinds of engines."]
249 pub type Type = ::std::os::raw::c_uint;
250 #[doc = " An unspecified engine."]
251 pub const dnnl_any_engine: Type = 0;
252 #[doc = " CPU engine."]
253 pub const dnnl_cpu: Type = 1;
254 #[doc = " GPU engine."]
255 pub const dnnl_gpu: Type = 2;
256}
257#[doc = " @struct dnnl_engine\n @brief An opaque structure to describe an engine."]
258#[repr(C)]
259#[derive(Debug, Copy, Clone)]
260pub struct dnnl_engine {
261 _unused: [u8; 0],
262}
263#[doc = " @brief An engine handle."]
264pub type dnnl_engine_t = *mut dnnl_engine;
265pub mod dnnl_stream_flags_t {
266 #[doc = " @brief Stream flags."]
267 pub type Type = ::std::os::raw::c_uint;
268 pub const dnnl_stream_in_order: Type = 1;
269 #[doc = " Out-of-order execution."]
270 pub const dnnl_stream_out_of_order: Type = 2;
271 #[doc = " Default stream configuration."]
272 pub const dnnl_stream_default_flags: Type = 1;
273}
274#[doc = " @struct dnnl_stream\n An opaque structure to describe an execution stream."]
275#[repr(C)]
276#[derive(Debug, Copy, Clone)]
277pub struct dnnl_stream {
278 _unused: [u8; 0],
279}
280#[doc = " An execution stream handle."]
281pub type dnnl_stream_t = *mut dnnl_stream;
282#[doc = " A constant execution stream handle."]
283pub type const_dnnl_stream_t = *const dnnl_stream;
284#[doc = " Structure containing version information as per [Semantic\n Versioning](https://semver.org)"]
285#[repr(C)]
286#[derive(Debug, Copy, Clone)]
287pub struct dnnl_version_t {
288 #[doc = "< Major version"]
289 pub major: ::std::os::raw::c_int,
290 #[doc = "< Minor version"]
291 pub minor: ::std::os::raw::c_int,
292 #[doc = "< Patch version"]
293 pub patch: ::std::os::raw::c_int,
294 #[doc = "< Git hash of the sources (may be absent)"]
295 pub hash: *const ::std::os::raw::c_char,
296 #[doc = "< CPU runtime"]
297 pub cpu_runtime: ::std::os::raw::c_uint,
298 #[doc = "< GPU runtime"]
299 pub gpu_runtime: ::std::os::raw::c_uint,
300}
301#[allow(clippy::unnecessary_operation, clippy::identity_op)]
302const _: () = {
303 ["Size of dnnl_version_t"][::std::mem::size_of::<dnnl_version_t>() - 32usize];
304 ["Alignment of dnnl_version_t"][::std::mem::align_of::<dnnl_version_t>() - 8usize];
305 ["Offset of field: dnnl_version_t::major"]
306 [::std::mem::offset_of!(dnnl_version_t, major) - 0usize];
307 ["Offset of field: dnnl_version_t::minor"]
308 [::std::mem::offset_of!(dnnl_version_t, minor) - 4usize];
309 ["Offset of field: dnnl_version_t::patch"]
310 [::std::mem::offset_of!(dnnl_version_t, patch) - 8usize];
311 ["Offset of field: dnnl_version_t::hash"]
312 [::std::mem::offset_of!(dnnl_version_t, hash) - 16usize];
313 ["Offset of field: dnnl_version_t::cpu_runtime"]
314 [::std::mem::offset_of!(dnnl_version_t, cpu_runtime) - 24usize];
315 ["Offset of field: dnnl_version_t::gpu_runtime"]
316 [::std::mem::offset_of!(dnnl_version_t, gpu_runtime) - 28usize];
317};
318unsafe extern "C" {
319 #[doc = " Returns the number of engines of a particular kind.\n\n @param kind Kind of engines to count.\n @returns Count of the engines."]
320 pub fn dnnl_engine_get_count(kind: dnnl_engine_kind_t::Type) -> usize;
321}
322unsafe extern "C" {
323 #[doc = " Creates an engine.\n\n @param engine Output engine.\n @param kind Engine kind.\n @param index Engine index that should be between 0 and the count of\n engines of the requested kind.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
324 pub fn dnnl_engine_create(
325 engine: *mut dnnl_engine_t,
326 kind: dnnl_engine_kind_t::Type,
327 index: usize,
328 ) -> dnnl_status_t::Type;
329}
330unsafe extern "C" {
331 #[doc = " Returns the kind of an engine.\n\n @param engine Engine to query.\n @param kind Output engine kind.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
332 pub fn dnnl_engine_get_kind(
333 engine: dnnl_engine_t,
334 kind: *mut dnnl_engine_kind_t::Type,
335 ) -> dnnl_status_t::Type;
336}
337unsafe extern "C" {
338 #[doc = " Destroys an engine.\n\n @param engine Engine to destroy.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
339 pub fn dnnl_engine_destroy(engine: dnnl_engine_t) -> dnnl_status_t::Type;
340}
341unsafe extern "C" {
342 #[doc = " Creates an execution stream.\n\n @param stream Output execution stream.\n @param engine Engine to create the execution stream on.\n @param flags Stream behavior flags (@sa dnnl_stream_flags_t).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
343 pub fn dnnl_stream_create(
344 stream: *mut dnnl_stream_t,
345 engine: dnnl_engine_t,
346 flags: ::std::os::raw::c_uint,
347 ) -> dnnl_status_t::Type;
348}
349unsafe extern "C" {
350 #[doc = " Returns the engine of a stream object.\n\n @param stream Stream object.\n @param engine Output engine on which the stream is created.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
351 pub fn dnnl_stream_get_engine(
352 stream: const_dnnl_stream_t,
353 engine: *mut dnnl_engine_t,
354 ) -> dnnl_status_t::Type;
355}
356unsafe extern "C" {
357 #[doc = " Waits for all primitives in the execution stream to finish computations.\n\n @param stream Execution stream.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
358 pub fn dnnl_stream_wait(stream: dnnl_stream_t) -> dnnl_status_t::Type;
359}
360unsafe extern "C" {
361 #[doc = " Destroys an execution stream.\n\n @param stream Execution stream to destroy.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
362 pub fn dnnl_stream_destroy(stream: dnnl_stream_t) -> dnnl_status_t::Type;
363}
364unsafe extern "C" {
365 #[doc = " Returns the floating-point math mode that will be used by default\n for all subsequently created primitives.\n\n @param mode Output FP math mode.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
366 pub fn dnnl_get_default_fpmath_mode(mode: *mut dnnl_fpmath_mode_t::Type)
367 -> dnnl_status_t::Type;
368}
369unsafe extern "C" {
370 #[doc = " Sets the floating-point math mode that will be used by default\n for all subsequently created primitives.\n\n @param mode FP math mode. The possible values are:\n #dnnl_fpmath_mode_strict,\n #dnnl_fpmath_mode_bf16,\n #dnnl_fpmath_mode_f16,\n #dnnl_fpmath_mode_tf32,\n #dnnl_fpmath_mode_any.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
371 pub fn dnnl_set_default_fpmath_mode(mode: dnnl_fpmath_mode_t::Type) -> dnnl_status_t::Type;
372}
373unsafe extern "C" {
374 #[doc = " Configures verbose output to stdout.\n\n @note\n Enabling verbose output affects performance.\n This setting overrides the ONEDNN_VERBOSE environment variable.\n\n @param level Verbosity level:\n - 0: no verbose output (default),\n - 1: primitive and graph information at execution,\n - 2: primitive and graph information at creation/compilation and execution.\n @returns #dnnl_invalid_arguments/#dnnl::status::invalid_arguments if the\n @p level value is invalid, and #dnnl_success/#dnnl::status::success on\n success."]
375 pub fn dnnl_set_verbose(level: ::std::os::raw::c_int) -> dnnl_status_t::Type;
376}
377unsafe extern "C" {
378 #[doc = " Returns library version information.\n @returns Pointer to a constant structure containing\n - major: major version number,\n - minor: minor version number,\n - patch: patch release number,\n - hash: git commit hash."]
379 pub fn dnnl_version() -> *const dnnl_version_t;
380}
381pub mod dnnl_format_kind_t {
382 #[doc = " Memory format kind"]
383 pub type Type = ::std::os::raw::c_uint;
384 #[doc = " Undefined memory format kind, used for empty memory descriptors."]
385 pub const dnnl_format_kind_undef: Type = 0;
386 #[doc = " A special format kind that indicates that the actual format will be\n selected by a primitive automatically."]
387 pub const dnnl_format_kind_any: Type = 1;
388 #[doc = " A tensor in a generic format described by the stride and blocking\n values in each dimension."]
389 pub const dnnl_blocked: Type = 2;
390 #[doc = " A special format kind that indicates that tensor format is opaque."]
391 pub const dnnl_format_kind_opaque: Type = 3;
392 #[doc = " Parameter to allow internal only format kinds without undefined\n behavior. This parameter is chosen to be valid for so long as\n sizeof(int) >= 2."]
393 pub const dnnl_format_kind_max: Type = 32767;
394}
395pub mod dnnl_format_tag_t {
396 #[doc = " Memory format tag specification.\n\n oneDNN formats describe physical data layout. The physical layout\n is described as a sequence of the dimensions as they are laid out in the\n memory (from the outer-most to the inner-most). Note that this order\n doesn't affect the logical order of the dimensions that is kept in the\n `dims` field of the dnnl_memory_desc_t structure. The logical order of the\n dimensions is specified by the primitive that uses the tensor.\n\n For example, CNN 5D tensor always has its logical dimensions in the order\n `(batch, channels, depth, height, width)`, while the physical layout might be\n `NCDHW` (corresponds to #dnnl_ncdhw format tag) or\n `NDHWC` (corresponds to #dnnl_ndhwc format tag).\n\n ~~~cpp\n int batch = 2, channels = 16, depth = 13, height = 13, width = 13;\n\n int ndims = 5; // 5D tensor\n dnnl_dims_t dims = {batch, channels, depth, height, width};\n dnnl_memory_desc_t data_in_ncdhw;\n dnnl_memory_desc_create_with_tag(\n &data_in_ncdhw, 5, dims, dnnl_f32, dnnl_ncdhw);\n\n // note that in both cases dims passed are the same\n dnnl_memory_desc_t data_in_ndhwc;\n dnnl_memory_desc_create_with_tag(\n &data_in_ndhwc, 5, dims, dnnl_f32, dnnl_ndhwc);\n\n dnnl_memory_desc_destroy(data_in_ncdhw);\n dnnl_memory_desc_destroy(data_in_ndhwc);\n ~~~\n\n Memory format tags can be further divided into two categories:\n - Domain-agnostic names, i.e. names the do not depend on the tensor usage\n in the specific primitive. These names use letters from `a` to `l` to\n denote logical dimension from 1 to 12, and form the order in which the\n dimensions are laid in memory. For instance, #dnnl_ab is used to denote\n 2D tensor where the second logical dimension (aka `b`) is the innermost,\n i.e. has stride = 1, and the first logical dimension (`a`) laid out in\n memory with stride equal to the size of second dimension. On the other\n hand, #dnnl_ba is just transposed version of the same tensor: the\n first dimension (`a`) becomes the innermost one.\n - Domain-specific names, i.e. names that make sense only in the context of\n a certain domain, such as CNN. This names are just aliases to the\n corresponding domain-agnostic tags and used mostly for the convenience.\n For example, #dnnl_nc is used to denote 2D CNN activations tensor\n memory format, where channels are the innermost dimension and batch is an\n outermost one. Moreover, #dnnl_nc is just an alias to #dnnl_ab,\n since for oneDNN CNN primitives the logical dimensions of\n activations tensors come in order: batch, channels, spatial.\n In other words, batch corresponds to the first logical dimension (`a`),\n channels correspond to the second one (`b`).\n\n The following domain-specific notation applies to memory format tags:\n - @c 'n' denotes the mini-batch dimension\n - @c 'c' denotes a channels dimension\n - When there are multiple channel dimensions (for example, in convolution\n weights tensor), @c 'i' and @c 'o' denote dimensions of input and output\n channels\n - @c 'd', @c 'h', and @c 'w' denote spatial depth, height, and width\n respectively\n\n Upper-case letters indicate that the data is laid out in blocks for a\n particular dimension. In such cases, the format name contains both upper-\n and lower-case letters for that dimension with a lower-case letter preceded\n by the block size. For example: #dnnl_nChw8c describes a format where the\n outermost dimension is mini-batch, followed by the channel block number,\n followed by the spatial height and width, and finally followed by 8-element\n channel blocks.\n\n @sa @ref dev_guide_understanding_memory_formats"]
397 pub type Type = ::std::os::raw::c_uint;
398 #[doc = " Undefined memory format tag"]
399 pub const dnnl_format_tag_undef: Type = 0;
400 #[doc = " Undefined memory format tag.\n The primitive selects a format automatically."]
401 pub const dnnl_format_tag_any: Type = 1;
402 #[doc = "< plain 1D tensor"]
403 pub const dnnl_a: Type = 2;
404 #[doc = "< plain 2D tensor"]
405 pub const dnnl_ab: Type = 3;
406 #[doc = "< plain 3D tensor"]
407 pub const dnnl_abc: Type = 4;
408 #[doc = "< plain 4D tensor"]
409 pub const dnnl_abcd: Type = 5;
410 #[doc = "< plain 5D tensor"]
411 pub const dnnl_abcde: Type = 6;
412 #[doc = "< plain 6D tensor"]
413 pub const dnnl_abcdef: Type = 7;
414 #[doc = "< plain 7D tensor"]
415 pub const dnnl_abcdefg: Type = 8;
416 #[doc = "< plain 8D tensor"]
417 pub const dnnl_abcdefgh: Type = 9;
418 #[doc = "< plain 9D tensor"]
419 pub const dnnl_abcdefghi: Type = 10;
420 #[doc = "< plain 10D tensor"]
421 pub const dnnl_abcdefghij: Type = 11;
422 #[doc = "< plain 11D tensor"]
423 pub const dnnl_abcdefghijk: Type = 12;
424 #[doc = "< plain 12D tensor"]
425 pub const dnnl_abcdefghijkl: Type = 13;
426 #[doc = "< permuted 2D tensor"]
427 pub const dnnl_ba: Type = 14;
428 #[doc = "< permuted 3D tensor"]
429 pub const dnnl_acb: Type = 15;
430 #[doc = "< permuted 3D tensor"]
431 pub const dnnl_bac: Type = 16;
432 #[doc = "< permuted 3D tensor"]
433 pub const dnnl_bca: Type = 17;
434 #[doc = "< permuted 3D tensor"]
435 pub const dnnl_cab: Type = 18;
436 #[doc = "< permuted 3D tensor"]
437 pub const dnnl_cba: Type = 19;
438 #[doc = "< permuted 4D tensor"]
439 pub const dnnl_abdc: Type = 20;
440 #[doc = "< permuted 4D tensor"]
441 pub const dnnl_acbd: Type = 21;
442 #[doc = "< permuted 4D tensor"]
443 pub const dnnl_acdb: Type = 22;
444 #[doc = "< permuted 4D tensor"]
445 pub const dnnl_adbc: Type = 23;
446 #[doc = "< permuted 4D tensor"]
447 pub const dnnl_adcb: Type = 24;
448 #[doc = "< permuted 4D tensor"]
449 pub const dnnl_bacd: Type = 25;
450 #[doc = "< permuted 4D tensor"]
451 pub const dnnl_bcda: Type = 26;
452 #[doc = "< permuted 4D tensor"]
453 pub const dnnl_cdab: Type = 27;
454 #[doc = "< permuted 4D tensor"]
455 pub const dnnl_cdba: Type = 28;
456 #[doc = "< permuted 4D tensor"]
457 pub const dnnl_dcab: Type = 29;
458 #[doc = "< permuted 5D tensor"]
459 pub const dnnl_abced: Type = 30;
460 #[doc = "< permuted 5D tensor"]
461 pub const dnnl_abdec: Type = 31;
462 #[doc = "< permuted 5D tensor"]
463 pub const dnnl_acbde: Type = 32;
464 #[doc = "< permuted 5D tensor"]
465 pub const dnnl_acdeb: Type = 33;
466 #[doc = "< permuted 5D tensor"]
467 pub const dnnl_adecb: Type = 34;
468 #[doc = "< permuted 5D tensor"]
469 pub const dnnl_bacde: Type = 35;
470 #[doc = "< permuted 5D tensor"]
471 pub const dnnl_bcdea: Type = 36;
472 #[doc = "< permuted 5D tensor"]
473 pub const dnnl_cdeab: Type = 37;
474 #[doc = "< permuted 5D tensor"]
475 pub const dnnl_cdeba: Type = 38;
476 #[doc = "< permuted 5D tensor"]
477 pub const dnnl_decab: Type = 39;
478 #[doc = "< permuted 6D tensor"]
479 pub const dnnl_abcdfe: Type = 40;
480 #[doc = "< permuted 6D tensor"]
481 pub const dnnl_abdefc: Type = 41;
482 #[doc = "< permuted 6D tensor"]
483 pub const dnnl_abdfce: Type = 42;
484 #[doc = "< permuted 6D tensor"]
485 pub const dnnl_acbdef: Type = 43;
486 #[doc = "< permuted 6D tensor"]
487 pub const dnnl_adefcb: Type = 44;
488 #[doc = "< permuted 6D tensor"]
489 pub const dnnl_defcab: Type = 45;
490 #[doc = "< permuted 7D tensor"]
491 pub const dnnl_abcdegf: Type = 46;
492 #[doc = "< permuted 8D tensor"]
493 pub const dnnl_abcdefhg: Type = 47;
494 #[doc = "< permuted 9D tensor"]
495 pub const dnnl_abcdefgih: Type = 48;
496 #[doc = "< permuted 10D tensor"]
497 pub const dnnl_abcdefghji: Type = 49;
498 #[doc = "< permuted 11D tensor"]
499 pub const dnnl_abcdefghikj: Type = 50;
500 #[doc = "< permuted 12D tensor"]
501 pub const dnnl_abcdefghijlk: Type = 51;
502 pub const dnnl_Abc16a: Type = 52;
503 pub const dnnl_ABc16a16b: Type = 53;
504 pub const dnnl_ABc32a32b: Type = 54;
505 pub const dnnl_ABc4a4b: Type = 55;
506 #[doc = " 3D tensor blocked by 2nd dimension with block size 16"]
507 pub const dnnl_aBc16b: Type = 56;
508 #[doc = " 3D tensor blocked by 2nd dimension with block size 16"]
509 pub const dnnl_ABc16b16a: Type = 57;
510 #[doc = " 3D tensor blocked by 2nd dimension with block size 16"]
511 pub const dnnl_Abc4a: Type = 58;
512 #[doc = " 3D tensor blocked by 2nd dimension with block size 32"]
513 pub const dnnl_aBc32b: Type = 59;
514 #[doc = " 3D tensor blocked by 2nd dimension with block size 4"]
515 pub const dnnl_aBc4b: Type = 60;
516 #[doc = " 3D tensor blocked by 2nd dimension with block size 4"]
517 pub const dnnl_ABc4b16a4b: Type = 61;
518 #[doc = " 3D tensor blocked by 2nd dimension with block size 4"]
519 pub const dnnl_ABc2b8a4b: Type = 62;
520 #[doc = " 3D tensor blocked by 2nd dimension with block size 4"]
521 pub const dnnl_ABc16b16a4b: Type = 63;
522 #[doc = " 3D tensor blocked by 2nd dimension with block size 4"]
523 pub const dnnl_ABc16b16a2b: Type = 64;
524 #[doc = " 3D tensor blocked by 2nd dimension with block size 4"]
525 pub const dnnl_ABc4b4a: Type = 65;
526 #[doc = " 3D tensor blocked by 2nd dimension with block size 4"]
527 pub const dnnl_ABc8a16b2a: Type = 66;
528 #[doc = " 3D tensor blocked by 2nd dimension with block size 4"]
529 pub const dnnl_ABc8a8b: Type = 67;
530 #[doc = " 3D tensor blocked by 2nd dimension with block size 4"]
531 pub const dnnl_ABc8a4b: Type = 68;
532 #[doc = " 3D tensor blocked by 2nd dimension with block size 8"]
533 pub const dnnl_aBc8b: Type = 69;
534 #[doc = " 3D tensor blocked by 2nd dimension with block size 8"]
535 pub const dnnl_ABc8b16a2b: Type = 70;
536 #[doc = " 3D tensor blocked by 2nd dimension with block size 8"]
537 pub const dnnl_BAc8a16b2a: Type = 71;
538 #[doc = " 3D tensor blocked by 2nd dimension with block size 8"]
539 pub const dnnl_ABc8b8a: Type = 72;
540 #[doc = " 3D tensor blocked by 2nd dimension with block size 8"]
541 pub const dnnl_Abcd16a: Type = 73;
542 #[doc = " 3D tensor blocked by 2nd dimension with block size 8"]
543 pub const dnnl_Abcd8a: Type = 74;
544 #[doc = " 3D tensor blocked by 2nd dimension with block size 8"]
545 pub const dnnl_ABcd16a16b: Type = 75;
546 #[doc = " 3D tensor blocked by 2nd dimension with block size 8"]
547 pub const dnnl_Abcd32a: Type = 76;
548 #[doc = " 3D tensor blocked by 2nd dimension with block size 8"]
549 pub const dnnl_ABcd32a32b: Type = 77;
550 #[doc = " 4D tensor blocked by 2nd dimension with block size 16"]
551 pub const dnnl_aBcd16b: Type = 78;
552 #[doc = " 4D tensor blocked by 2nd dimension with block size 16"]
553 pub const dnnl_ABcd16b16a: Type = 79;
554 #[doc = " 4D tensor blocked by 2nd dimension with block size 16"]
555 pub const dnnl_aBCd16b16c: Type = 80;
556 #[doc = " 4D tensor blocked by 2nd dimension with block size 16"]
557 pub const dnnl_aBCd16c16b: Type = 81;
558 #[doc = " 4D tensor blocked by 2nd dimension with block size 16"]
559 pub const dnnl_Abcd4a: Type = 82;
560 #[doc = " 4D tensor blocked by 2nd dimension with block size 32"]
561 pub const dnnl_aBcd32b: Type = 83;
562 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
563 pub const dnnl_aBcd4b: Type = 84;
564 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
565 pub const dnnl_ABcd4b16a4b: Type = 85;
566 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
567 pub const dnnl_ABcd16b16a4b: Type = 86;
568 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
569 pub const dnnl_ABcd16b16a2b: Type = 87;
570 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
571 pub const dnnl_ABcd4b4a: Type = 88;
572 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
573 pub const dnnl_ABcd4a4b: Type = 89;
574 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
575 pub const dnnl_aBCd2c4b2c: Type = 90;
576 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
577 pub const dnnl_aBCd4b8c2b: Type = 91;
578 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
579 pub const dnnl_aBCd4c16b4c: Type = 92;
580 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
581 pub const dnnl_aBCd2c8b4c: Type = 93;
582 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
583 pub const dnnl_aBCd16c16b4c: Type = 94;
584 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
585 pub const dnnl_aBCd16c16b2c: Type = 95;
586 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
587 pub const dnnl_aBCd4c4b: Type = 96;
588 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
589 pub const dnnl_aBCd4b4c: Type = 97;
590 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
591 pub const dnnl_ABcd8a16b2a: Type = 98;
592 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
593 pub const dnnl_ABcd2b8a4b: Type = 99;
594 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
595 pub const dnnl_ABcd8a8b: Type = 100;
596 #[doc = " 4D tensor blocked by 2nd dimension with block size 4"]
597 pub const dnnl_ABcd8a4b: Type = 101;
598 #[doc = " 4D tensor blocked by 2nd dimension with block size 8"]
599 pub const dnnl_aBcd8b: Type = 102;
600 #[doc = " 4D tensor blocked by 2nd dimension with block size 8"]
601 pub const dnnl_aBCd4c8b2c: Type = 103;
602 #[doc = " 4D tensor blocked by 2nd dimension with block size 8"]
603 pub const dnnl_ABcd8b16a2b: Type = 104;
604 #[doc = " 4D tensor blocked by 2nd dimension with block size 8"]
605 pub const dnnl_aBCd8b16c2b: Type = 105;
606 #[doc = " 4D tensor blocked by 2nd dimension with block size 8"]
607 pub const dnnl_BAcd8a16b2a: Type = 106;
608 #[doc = " 4D tensor blocked by 1st and 2nd dimension with block size 8"]
609 pub const dnnl_ABcd8b8a: Type = 107;
610 #[doc = " 4D tensor blocked by 1st and 2nd dimension with block size 8"]
611 pub const dnnl_aBCd8b8c: Type = 108;
612 #[doc = " 4D tensor blocked by 1st and 2nd dimension with block size 8"]
613 pub const dnnl_aBCd8b4c: Type = 109;
614 #[doc = " 4D tensor blocked by 1st and 2nd dimension with block size 8"]
615 pub const dnnl_aBCd8c16b2c: Type = 110;
616 #[doc = " 4D tensor blocked by 1st and 2nd dimension with block size 8"]
617 pub const dnnl_ABcde8a16b2a: Type = 111;
618 #[doc = " 4D tensor blocked by 1st and 2nd dimension with block size 8"]
619 pub const dnnl_aCBd8b16c2b: Type = 112;
620 #[doc = " 4D tensor blocked by 1st and 2nd dimension with block size 8"]
621 pub const dnnl_aBCd8c8b: Type = 113;
622 #[doc = " 4D tensor blocked by 1st and 2nd dimension with block size 8"]
623 pub const dnnl_Abcde16a: Type = 114;
624 #[doc = " 4D tensor blocked by 1st and 2nd dimension with block size 8"]
625 pub const dnnl_Abcde32a: Type = 115;
626 #[doc = " 4D tensor blocked by 1st and 2nd dimension with block size 8"]
627 pub const dnnl_ABcde16a16b: Type = 116;
628 #[doc = " 4D tensor blocked by 1st and 2nd dimension with block size 8"]
629 pub const dnnl_BAcde8a16b2a: Type = 117;
630 #[doc = " 4D tensor blocked by 3rd dimension with block size 4"]
631 pub const dnnl_aBCd2b4c2b: Type = 118;
632 #[doc = " 5D tensor blocked by 1st dimension with block size 16"]
633 pub const dnnl_ABcde4b16a4b: Type = 119;
634 #[doc = " 5D tensor blocked by 1st dimension with block size 8"]
635 pub const dnnl_ABcde2b8a4b: Type = 120;
636 #[doc = " 5D tensor blocked by 2nd dimension with block size 16"]
637 pub const dnnl_aBcde16b: Type = 121;
638 #[doc = " 5D tensor blocked by 2nd dimension with block size 16"]
639 pub const dnnl_ABcde16b16a: Type = 122;
640 #[doc = " 5D tensor blocked by 2nd dimension with block size 16"]
641 pub const dnnl_aBCde16b16c: Type = 123;
642 #[doc = " 5D tensor blocked by 2nd dimension with block size 16"]
643 pub const dnnl_aBCde16c16b: Type = 124;
644 #[doc = " 5D tensor blocked by 2nd dimension with block size 16"]
645 pub const dnnl_aBCde2c8b4c: Type = 125;
646 #[doc = " 5D tensor blocked by 2nd dimension with block size 16"]
647 pub const dnnl_Abcde4a: Type = 126;
648 #[doc = " 5D tensor blocked by 2nd dimension with block size 32"]
649 pub const dnnl_aBcde32b: Type = 127;
650 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
651 pub const dnnl_aBcde4b: Type = 128;
652 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
653 pub const dnnl_ABcde4b4a: Type = 129;
654 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
655 pub const dnnl_ABcde4a4b: Type = 130;
656 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
657 pub const dnnl_aBCde4b4c: Type = 131;
658 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
659 pub const dnnl_aBCde2c4b2c: Type = 132;
660 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
661 pub const dnnl_aBCde4b8c2b: Type = 133;
662 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
663 pub const dnnl_aBCde4c16b4c: Type = 134;
664 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
665 pub const dnnl_aBCde16c16b4c: Type = 135;
666 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
667 pub const dnnl_aBCde16c16b2c: Type = 136;
668 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
669 pub const dnnl_aBCde4c4b: Type = 137;
670 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
671 pub const dnnl_Abcde8a: Type = 138;
672 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
673 pub const dnnl_ABcde8a8b: Type = 139;
674 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
675 pub const dnnl_ABcde8a4b: Type = 140;
676 #[doc = " 5D tensor blocked by 2nd dimension with block size 4"]
677 pub const dnnl_BAcde16b16a: Type = 141;
678 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
679 pub const dnnl_aBcde8b: Type = 142;
680 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
681 pub const dnnl_ABcde8b16a2b: Type = 143;
682 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
683 pub const dnnl_aBCde8b16c2b: Type = 144;
684 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
685 pub const dnnl_aBCde4c8b2c: Type = 145;
686 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
687 pub const dnnl_aCBde8b16c2b: Type = 146;
688 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
689 pub const dnnl_ABcde8b8a: Type = 147;
690 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
691 pub const dnnl_ABcde32a32b: Type = 148;
692 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
693 pub const dnnl_aBCde8b8c: Type = 149;
694 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
695 pub const dnnl_aBCde8b4c: Type = 150;
696 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
697 pub const dnnl_ABc4a8b8a4b: Type = 151;
698 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
699 pub const dnnl_ABcd4a8b8a4b: Type = 152;
700 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
701 pub const dnnl_ABcde4a8b8a4b: Type = 153;
702 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
703 pub const dnnl_BAc4b8a8b4a: Type = 154;
704 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
705 pub const dnnl_BAcd4b8a8b4a: Type = 155;
706 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
707 pub const dnnl_BAcde4b8a8b4a: Type = 156;
708 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
709 pub const dnnl_ABcd2a8b8a2b: Type = 157;
710 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
711 pub const dnnl_aBCd4b8c8b4c: Type = 158;
712 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
713 pub const dnnl_aBCde4b8c8b4c: Type = 159;
714 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
715 pub const dnnl_aBCde2b8c8b2c: Type = 160;
716 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
717 pub const dnnl_aBCde8c16b2c: Type = 161;
718 #[doc = " 5D tensor blocked by 2nd dimension with block size 8"]
719 pub const dnnl_aBCde8c8b: Type = 162;
720 #[doc = " 5D tensor blocked by 3rd dimension with block size 4"]
721 pub const dnnl_aBCde2b4c2b: Type = 163;
722 #[doc = " 6D tensor blocked by 2nd dimension with block size 16"]
723 pub const dnnl_aBcdef16b: Type = 164;
724 #[doc = " 6D tensor blocked by 2nd dimension with block size 16"]
725 pub const dnnl_aBCdef16b16c: Type = 165;
726 #[doc = " 6D tensor blocked by 2nd dimension with block size 16"]
727 pub const dnnl_aBCdef16c16b: Type = 166;
728 #[doc = " 6D tensor blocked by 2nd dimension with block size 16"]
729 pub const dnnl_aBCdef4c16b4c: Type = 167;
730 #[doc = " 6D tensor blocked by 2nd dimension with block size 8"]
731 pub const dnnl_aBCdef2c8b4c: Type = 168;
732 #[doc = " 6D tensor blocked by 2nd dimension with block size 8"]
733 pub const dnnl_aBCdef4c8b2c: Type = 169;
734 #[doc = " 6D tensor blocked by 3rd dimension with block size 4"]
735 pub const dnnl_aBCdef2b4c2b: Type = 170;
736 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
737 pub const dnnl_aBcdef4b: Type = 171;
738 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
739 pub const dnnl_aBCdef4c4b: Type = 172;
740 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
741 pub const dnnl_aBCdef4b4c: Type = 173;
742 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
743 pub const dnnl_aBCdef2c4b2c: Type = 174;
744 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
745 pub const dnnl_aBCdef4b8c2b: Type = 175;
746 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
747 pub const dnnl_aBCdef8b8c: Type = 176;
748 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
749 pub const dnnl_aBCdef8b4c: Type = 177;
750 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
751 pub const dnnl_aBCdef8c16b2c: Type = 178;
752 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
753 pub const dnnl_aBCdef4b8c8b4c: Type = 179;
754 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
755 pub const dnnl_aBCdef8b16c2b: Type = 180;
756 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
757 pub const dnnl_aCBdef8b16c2b: Type = 181;
758 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
759 pub const dnnl_aBCdef8c8b: Type = 182;
760 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
761 pub const dnnl_aBdc16b: Type = 183;
762 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
763 pub const dnnl_aBdC16b2c: Type = 184;
764 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
765 pub const dnnl_aBdC16b4c: Type = 185;
766 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
767 pub const dnnl_aBdc4b: Type = 186;
768 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
769 pub const dnnl_aBdc8b: Type = 187;
770 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
771 pub const dnnl_aBdec16b: Type = 188;
772 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
773 pub const dnnl_aBdeC16b2c: Type = 189;
774 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
775 pub const dnnl_aBdeC16b4c: Type = 190;
776 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
777 pub const dnnl_aBdec32b: Type = 191;
778 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
779 pub const dnnl_aBdec4b: Type = 192;
780 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
781 pub const dnnl_aBdec8b: Type = 193;
782 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
783 pub const dnnl_aBdefc16b: Type = 194;
784 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
785 pub const dnnl_aBdefC16b2c: Type = 195;
786 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
787 pub const dnnl_aCBdef16c16b: Type = 196;
788 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
789 pub const dnnl_aBdefc4b: Type = 197;
790 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
791 pub const dnnl_aBdefc8b: Type = 198;
792 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
793 pub const dnnl_Abcdef16a: Type = 199;
794 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
795 pub const dnnl_Abcdef32a: Type = 200;
796 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
797 pub const dnnl_aBedc16b: Type = 201;
798 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
799 pub const dnnl_Acb16a: Type = 202;
800 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
801 pub const dnnl_AcB16a2b: Type = 203;
802 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
803 pub const dnnl_AcB16a4b: Type = 204;
804 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
805 pub const dnnl_Acb4a: Type = 205;
806 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
807 pub const dnnl_Acb8a: Type = 206;
808 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
809 pub const dnnl_aCBd16b16c: Type = 207;
810 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
811 pub const dnnl_aCBd16c16b: Type = 208;
812 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
813 pub const dnnl_aCBde16b16c: Type = 209;
814 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
815 pub const dnnl_aCBde16c16b: Type = 210;
816 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
817 pub const dnnl_Acdb16a: Type = 211;
818 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
819 pub const dnnl_AcdB16a2b: Type = 212;
820 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
821 pub const dnnl_AcdB16a4b: Type = 213;
822 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
823 pub const dnnl_Acdb32a: Type = 214;
824 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
825 pub const dnnl_Acdb4a: Type = 215;
826 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
827 pub const dnnl_Acdb8a: Type = 216;
828 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
829 pub const dnnl_Acdeb16a: Type = 217;
830 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
831 pub const dnnl_AcdeB16a2b: Type = 218;
832 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
833 pub const dnnl_Acdeb4a: Type = 219;
834 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
835 pub const dnnl_Acdeb8a: Type = 220;
836 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
837 pub const dnnl_Adcb16a: Type = 221;
838 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
839 pub const dnnl_BAc16a16b: Type = 222;
840 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
841 pub const dnnl_BAc16b16a: Type = 223;
842 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
843 pub const dnnl_BAcd16a16b: Type = 224;
844 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
845 pub const dnnl_BAcd16b16a: Type = 225;
846 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
847 pub const dnnl_aCBd4c8b8c4b: Type = 226;
848 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
849 pub const dnnl_aCBde4c8b8c4b: Type = 227;
850 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
851 pub const dnnl_aCBdef4c8b8c4b: Type = 228;
852 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
853 pub const dnnl_BAcde16a16b: Type = 229;
854 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
855 pub const dnnl_aCBdef16b16c: Type = 230;
856 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
857 pub const dnnl_ABc16b32a: Type = 231;
858 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
859 pub const dnnl_ABc16b64a: Type = 232;
860 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
861 pub const dnnl_ABc4b32a4b: Type = 233;
862 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
863 pub const dnnl_ABc4b64a4b: Type = 234;
864 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
865 pub const dnnl_ABc8b32a2b: Type = 235;
866 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
867 pub const dnnl_ABc8b64a2b: Type = 236;
868 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
869 pub const dnnl_AB16b16a: Type = 237;
870 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
871 pub const dnnl_AB16b32a: Type = 238;
872 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
873 pub const dnnl_AB16b64a: Type = 239;
874 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
875 pub const dnnl_AB8b16a2b: Type = 240;
876 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
877 pub const dnnl_AB8b32a2b: Type = 241;
878 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
879 pub const dnnl_AB8b64a2b: Type = 242;
880 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
881 pub const dnnl_AB4b16a4b: Type = 243;
882 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
883 pub const dnnl_AB4b32a4b: Type = 244;
884 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
885 pub const dnnl_AB4b64a4b: Type = 245;
886 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
887 pub const dnnl_AB16b16a4b: Type = 246;
888 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
889 pub const dnnl_ABcd16b32a: Type = 247;
890 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
891 pub const dnnl_ABcd16b64a: Type = 248;
892 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
893 pub const dnnl_ABcd4b32a4b: Type = 249;
894 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
895 pub const dnnl_ABcd4b64a4b: Type = 250;
896 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
897 pub const dnnl_ABcd8b32a2b: Type = 251;
898 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
899 pub const dnnl_ABcd8b64a2b: Type = 252;
900 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
901 pub const dnnl_ABcde4b32a4b: Type = 253;
902 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
903 pub const dnnl_ABcde4b64a4b: Type = 254;
904 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
905 pub const dnnl_ABcde16b16a4b: Type = 255;
906 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
907 pub const dnnl_ABcde16b16a2b: Type = 256;
908 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
909 pub const dnnl_ABcde16b32a: Type = 257;
910 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
911 pub const dnnl_ABcde16b64a: Type = 258;
912 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
913 pub const dnnl_ABcde8b32a2b: Type = 259;
914 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
915 pub const dnnl_ABcde8b64a2b: Type = 260;
916 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
917 pub const dnnl_aBCdef16c16b4c: Type = 261;
918 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
919 pub const dnnl_aBCdef16c16b2c: Type = 262;
920 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
921 pub const dnnl_AB32a32b8a4b: Type = 263;
922 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
923 pub const dnnl_AB8a4b: Type = 264;
924 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
925 pub const dnnl_AB32a32b8a2b: Type = 265;
926 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
927 pub const dnnl_AB8a2b: Type = 266;
928 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
929 pub const dnnl_abDc32d: Type = 267;
930 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
931 pub const dnnl_abDC32d4c: Type = 268;
932 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
933 pub const dnnl_abdEc32e: Type = 269;
934 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
935 pub const dnnl_abdEC32e2c: Type = 270;
936 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
937 pub const dnnl_abdEC32e4c: Type = 271;
938 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
939 pub const dnnl_aBdefC16b4c: Type = 272;
940 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
941 pub const dnnl_AcdeB16a4b: Type = 273;
942 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
943 pub const dnnl_ABcd16a16b2a: Type = 274;
944 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
945 pub const dnnl_ABc16a16b2a: Type = 275;
946 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
947 pub const dnnl_aBCd16b16c2b: Type = 276;
948 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
949 pub const dnnl_aBCde16b16c2b: Type = 277;
950 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
951 pub const dnnl_Acb32a: Type = 278;
952 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
953 pub const dnnl_AcB32a2b: Type = 279;
954 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
955 pub const dnnl_AcB32a4b: Type = 280;
956 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
957 pub const dnnl_Acb48a: Type = 281;
958 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
959 pub const dnnl_AcB48a2b: Type = 282;
960 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
961 pub const dnnl_AcB48a4b: Type = 283;
962 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
963 pub const dnnl_Acb64a: Type = 284;
964 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
965 pub const dnnl_AcB64a2b: Type = 285;
966 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
967 pub const dnnl_AcB64a4b: Type = 286;
968 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
969 pub const dnnl_cBa2b: Type = 287;
970 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
971 pub const dnnl_cBa4b: Type = 288;
972 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
973 pub const dnnl_aBdc32b: Type = 289;
974 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
975 pub const dnnl_aBdC32b2c: Type = 290;
976 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
977 pub const dnnl_aBdC32b4c: Type = 291;
978 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
979 pub const dnnl_aBdc48b: Type = 292;
980 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
981 pub const dnnl_aBdC48b2c: Type = 293;
982 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
983 pub const dnnl_aBdC48b4c: Type = 294;
984 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
985 pub const dnnl_aBdc64b: Type = 295;
986 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
987 pub const dnnl_aBdC64b2c: Type = 296;
988 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
989 pub const dnnl_aBdC64b4c: Type = 297;
990 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
991 pub const dnnl_adCb2c: Type = 298;
992 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
993 pub const dnnl_adCb4c: Type = 299;
994 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
995 pub const dnnl_AcdB32a2b: Type = 300;
996 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
997 pub const dnnl_AcdB32a4b: Type = 301;
998 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
999 pub const dnnl_Acdb48a: Type = 302;
1000 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1001 pub const dnnl_AcdB48a2b: Type = 303;
1002 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1003 pub const dnnl_AcdB48a4b: Type = 304;
1004 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1005 pub const dnnl_Acdb64a: Type = 305;
1006 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1007 pub const dnnl_AcdB64a2b: Type = 306;
1008 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1009 pub const dnnl_AcdB64a4b: Type = 307;
1010 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1011 pub const dnnl_cdBa2b: Type = 308;
1012 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1013 pub const dnnl_cdBa4b: Type = 309;
1014 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1015 pub const dnnl_aBdeC32b2c: Type = 310;
1016 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1017 pub const dnnl_aBdeC32b4c: Type = 311;
1018 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1019 pub const dnnl_aBdec48b: Type = 312;
1020 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1021 pub const dnnl_aBdeC48b2c: Type = 313;
1022 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1023 pub const dnnl_aBdeC48b4c: Type = 314;
1024 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1025 pub const dnnl_aBdec64b: Type = 315;
1026 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1027 pub const dnnl_aBdeC64b2c: Type = 316;
1028 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1029 pub const dnnl_aBdeC64b4c: Type = 317;
1030 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1031 pub const dnnl_adeCb2c: Type = 318;
1032 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1033 pub const dnnl_adeCb4c: Type = 319;
1034 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1035 pub const dnnl_Acdeb32a: Type = 320;
1036 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1037 pub const dnnl_AcdeB32a2b: Type = 321;
1038 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1039 pub const dnnl_AcdeB32a4b: Type = 322;
1040 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1041 pub const dnnl_Acdeb48a: Type = 323;
1042 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1043 pub const dnnl_AcdeB48a2b: Type = 324;
1044 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1045 pub const dnnl_AcdeB48a4b: Type = 325;
1046 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1047 pub const dnnl_Acdeb64a: Type = 326;
1048 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1049 pub const dnnl_AcdeB64a2b: Type = 327;
1050 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1051 pub const dnnl_AcdeB64a4b: Type = 328;
1052 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1053 pub const dnnl_cdeBa2b: Type = 329;
1054 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1055 pub const dnnl_cdeBa4b: Type = 330;
1056 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1057 pub const dnnl_aBdefc32b: Type = 331;
1058 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1059 pub const dnnl_aBdefC32b2c: Type = 332;
1060 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1061 pub const dnnl_aBdefC32b4c: Type = 333;
1062 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1063 pub const dnnl_aBdefc48b: Type = 334;
1064 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1065 pub const dnnl_aBdefC48b2c: Type = 335;
1066 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1067 pub const dnnl_aBdefC48b4c: Type = 336;
1068 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1069 pub const dnnl_aBdefc64b: Type = 337;
1070 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1071 pub const dnnl_aBdefC64b2c: Type = 338;
1072 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1073 pub const dnnl_aBdefC64b4c: Type = 339;
1074 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1075 pub const dnnl_adefCb2c: Type = 340;
1076 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1077 pub const dnnl_adefCb4c: Type = 341;
1078 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1079 pub const dnnl_AB16b32a4b: Type = 342;
1080 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1081 pub const dnnl_AB16b48a4b: Type = 343;
1082 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1083 pub const dnnl_AB16b64a4b: Type = 344;
1084 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1085 pub const dnnl_AB16b16a2b: Type = 345;
1086 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1087 pub const dnnl_AB16b32a2b: Type = 346;
1088 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1089 pub const dnnl_AB16b48a2b: Type = 347;
1090 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1091 pub const dnnl_AB16b64a2b: Type = 348;
1092 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1093 pub const dnnl_ABc16b32a4b: Type = 349;
1094 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1095 pub const dnnl_ABc16b48a4b: Type = 350;
1096 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1097 pub const dnnl_ABc16b64a4b: Type = 351;
1098 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1099 pub const dnnl_ABc16b32a2b: Type = 352;
1100 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1101 pub const dnnl_ABc16b48a2b: Type = 353;
1102 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1103 pub const dnnl_ABc16b64a2b: Type = 354;
1104 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1105 pub const dnnl_ABcd16b32a4b: Type = 355;
1106 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1107 pub const dnnl_ABcd16b48a4b: Type = 356;
1108 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1109 pub const dnnl_ABcd16b64a4b: Type = 357;
1110 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1111 pub const dnnl_ABcd16b32a2b: Type = 358;
1112 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1113 pub const dnnl_ABcd16b48a2b: Type = 359;
1114 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1115 pub const dnnl_ABcd16b64a2b: Type = 360;
1116 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1117 pub const dnnl_ABcde16b32a4b: Type = 361;
1118 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1119 pub const dnnl_ABcde16b48a4b: Type = 362;
1120 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1121 pub const dnnl_ABcde16b64a4b: Type = 363;
1122 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1123 pub const dnnl_ABcde16b32a2b: Type = 364;
1124 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1125 pub const dnnl_ABcde16b48a2b: Type = 365;
1126 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1127 pub const dnnl_ABcde16b64a2b: Type = 366;
1128 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1129 pub const dnnl_ABc32a16b: Type = 367;
1130 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1131 pub const dnnl_ABcd32a16b: Type = 368;
1132 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1133 pub const dnnl_ABcde32a16b: Type = 369;
1134 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1135 pub const dnnl_AB48a16b: Type = 370;
1136 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1137 pub const dnnl_AB48a32b: Type = 371;
1138 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1139 pub const dnnl_ABc40a16b: Type = 372;
1140 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1141 pub const dnnl_ABc40a32b: Type = 373;
1142 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1143 pub const dnnl_aBC48b16c: Type = 374;
1144 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1145 pub const dnnl_aBC48b32c: Type = 375;
1146 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1147 pub const dnnl_ABcd40a16b: Type = 376;
1148 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1149 pub const dnnl_ABcd40a32b: Type = 377;
1150 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1151 pub const dnnl_abCd32c: Type = 378;
1152 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1153 pub const dnnl_abdCe32c: Type = 379;
1154 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1155 pub const dnnl_abdCE32c2e: Type = 380;
1156 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1157 pub const dnnl_BA16a16b2a: Type = 381;
1158 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1159 pub const dnnl_BA16a32b2a: Type = 382;
1160 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1161 pub const dnnl_BA16a48b2a: Type = 383;
1162 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1163 pub const dnnl_BA16a64b2a: Type = 384;
1164 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1165 pub const dnnl_BA16a16b4a: Type = 385;
1166 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1167 pub const dnnl_BA16a32b4a: Type = 386;
1168 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1169 pub const dnnl_BA16a48b4a: Type = 387;
1170 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1171 pub const dnnl_BA16a64b4a: Type = 388;
1172 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1173 pub const dnnl_ABcd8a2b: Type = 389;
1174 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1175 pub const dnnl_aBdeC16c16b2c: Type = 390;
1176 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1177 pub const dnnl_aBdeC16c16b4c: Type = 391;
1178 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1179 pub const dnnl_aBdefC16c16b2c: Type = 392;
1180 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1181 pub const dnnl_AcB16b16a2b: Type = 393;
1182 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1183 pub const dnnl_AcB16b16a4b: Type = 394;
1184 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1185 pub const dnnl_AcdB16b16a2b: Type = 395;
1186 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1187 pub const dnnl_AcdB16b16a4b: Type = 396;
1188 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1189 pub const dnnl_AcdeB16b16a2b: Type = 397;
1190 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1191 pub const dnnl_aBdefC16c16b4c: Type = 398;
1192 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1193 pub const dnnl_AcdeB16b16a4b: Type = 399;
1194 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1195 pub const dnnl_AcB16b32a2b: Type = 400;
1196 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1197 pub const dnnl_AcB16b32a4b: Type = 401;
1198 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1199 pub const dnnl_AcB16b48a2b: Type = 402;
1200 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1201 pub const dnnl_AcB16b48a4b: Type = 403;
1202 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1203 pub const dnnl_AcB16b64a2b: Type = 404;
1204 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1205 pub const dnnl_AcB16b64a4b: Type = 405;
1206 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1207 pub const dnnl_aBdC16c16b2c: Type = 406;
1208 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1209 pub const dnnl_aBdC16c16b4c: Type = 407;
1210 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1211 pub const dnnl_aBdC16c32b2c: Type = 408;
1212 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1213 pub const dnnl_aBdC16c32b4c: Type = 409;
1214 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1215 pub const dnnl_aBdC16c48b2c: Type = 410;
1216 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1217 pub const dnnl_aBdC16c48b4c: Type = 411;
1218 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1219 pub const dnnl_aBdC16c64b2c: Type = 412;
1220 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1221 pub const dnnl_aBdC16c64b4c: Type = 413;
1222 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1223 pub const dnnl_AcdB16b32a2b: Type = 414;
1224 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1225 pub const dnnl_AcdB16b32a4b: Type = 415;
1226 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1227 pub const dnnl_AcdB16b48a2b: Type = 416;
1228 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1229 pub const dnnl_AcdB16b48a4b: Type = 417;
1230 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1231 pub const dnnl_AcdB16b64a2b: Type = 418;
1232 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1233 pub const dnnl_AcdB16b64a4b: Type = 419;
1234 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1235 pub const dnnl_aBdeC16c32b2c: Type = 420;
1236 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1237 pub const dnnl_aBdeC16c32b4c: Type = 421;
1238 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1239 pub const dnnl_aBdeC16c48b2c: Type = 422;
1240 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1241 pub const dnnl_aBdeC16c48b4c: Type = 423;
1242 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1243 pub const dnnl_aBdeC16c64b2c: Type = 424;
1244 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1245 pub const dnnl_aBdeC16c64b4c: Type = 425;
1246 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1247 pub const dnnl_AcdeB16b32a2b: Type = 426;
1248 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1249 pub const dnnl_AcdeB16b32a4b: Type = 427;
1250 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1251 pub const dnnl_AcdeB16b48a2b: Type = 428;
1252 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1253 pub const dnnl_AcdeB16b48a4b: Type = 429;
1254 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1255 pub const dnnl_AcdeB16b64a2b: Type = 430;
1256 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1257 pub const dnnl_AcdeB16b64a4b: Type = 431;
1258 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1259 pub const dnnl_aBdefC16c32b2c: Type = 432;
1260 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1261 pub const dnnl_aBdefC16c32b4c: Type = 433;
1262 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1263 pub const dnnl_aBdefC16c48b2c: Type = 434;
1264 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1265 pub const dnnl_aBdefC16c48b4c: Type = 435;
1266 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1267 pub const dnnl_aBdefC16c64b2c: Type = 436;
1268 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1269 pub const dnnl_aBdefC16c64b4c: Type = 437;
1270 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1271 pub const dnnl_decbA16a: Type = 438;
1272 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1273 pub const dnnl_ABc4a2b: Type = 439;
1274 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1275 pub const dnnl_ABc8a2b: Type = 440;
1276 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1277 pub const dnnl_aBCd8b2c: Type = 441;
1278 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1279 pub const dnnl_ABcde4a2b: Type = 442;
1280 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1281 pub const dnnl_ABcde8a2b: Type = 443;
1282 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1283 pub const dnnl_ABcde40a16b: Type = 444;
1284 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1285 pub const dnnl_ABcde40a32b: Type = 445;
1286 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1287 pub const dnnl_aBCde8b2c: Type = 446;
1288 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1289 pub const dnnl_ABcde4a8b8a2b: Type = 447;
1290 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1291 pub const dnnl_ABcd4a8b8a2b: Type = 448;
1292 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1293 pub const dnnl_ABc4a8b8a2b: Type = 449;
1294 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1295 pub const dnnl_aBCdef4b8c8b2c: Type = 450;
1296 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1297 pub const dnnl_aBCde4b8c8b2c: Type = 451;
1298 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1299 pub const dnnl_aBCd4b8c8b2c: Type = 452;
1300 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1301 pub const dnnl_BAcde4b8a8b2a: Type = 453;
1302 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1303 pub const dnnl_BAcd4b8a8b2a: Type = 454;
1304 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1305 pub const dnnl_BAc4b8a8b2a: Type = 455;
1306 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1307 pub const dnnl_aCBdef4c8b8c2b: Type = 456;
1308 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1309 pub const dnnl_aCBde4c8b8c2b: Type = 457;
1310 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1311 pub const dnnl_aCBd4c8b8c2b: Type = 458;
1312 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1313 pub const dnnl_aBCdef8b2c: Type = 459;
1314 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1315 pub const dnnl_AB32a16b: Type = 460;
1316 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1317 pub const dnnl_AB32a32b: Type = 461;
1318 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1319 pub const dnnl_BA4b8a8b2a: Type = 462;
1320 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1321 pub const dnnl_BA4b8a8b4a: Type = 463;
1322 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1323 pub const dnnl_aBC32b16c: Type = 464;
1324 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1325 pub const dnnl_aBC32b32c: Type = 465;
1326 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1327 pub const dnnl_aCB4c8b8c2b: Type = 466;
1328 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1329 pub const dnnl_aCB4c8b8c4b: Type = 467;
1330 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1331 pub const dnnl_ABcd4a2b: Type = 468;
1332 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1333 pub const dnnl_ABc2b8a16b4a: Type = 469;
1334 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1335 pub const dnnl_ABcd2b8a16b4a: Type = 470;
1336 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1337 pub const dnnl_ABcde2b8a16b4a: Type = 471;
1338 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1339 pub const dnnl_ABc2a8b16a4b: Type = 472;
1340 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1341 pub const dnnl_ABc2a8b16a2b: Type = 473;
1342 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1343 pub const dnnl_ABc2b32a8b: Type = 474;
1344 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1345 pub const dnnl_ABcd2a8b16a4b: Type = 475;
1346 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1347 pub const dnnl_ABcd2a8b16a2b: Type = 476;
1348 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1349 pub const dnnl_aCBd2c8b16c2b: Type = 477;
1350 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1351 pub const dnnl_ABcd2b32a8b: Type = 478;
1352 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1353 pub const dnnl_aBCd2c8b16c2b: Type = 479;
1354 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1355 pub const dnnl_ABcde2a8b16a4b: Type = 480;
1356 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1357 pub const dnnl_ABcde2a8b16a2b: Type = 481;
1358 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1359 pub const dnnl_aCBde2c8b16c2b: Type = 482;
1360 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1361 pub const dnnl_ABcde2b32a8b: Type = 483;
1362 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1363 pub const dnnl_aBC2b8c16b2c: Type = 484;
1364 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1365 pub const dnnl_aBCd2b8c16b2c: Type = 485;
1366 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1367 pub const dnnl_aBCde2b8c16b2c: Type = 486;
1368 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1369 pub const dnnl_aBCdef2b8c16b2c: Type = 487;
1370 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1371 pub const dnnl_BAcde2b8a16b4a: Type = 488;
1372 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1373 pub const dnnl_BAcd2b8a16b4a: Type = 489;
1374 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1375 pub const dnnl_BAc2b8a16b4a: Type = 490;
1376 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1377 pub const dnnl_BAcde2b8a16b2a: Type = 491;
1378 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1379 pub const dnnl_BAcd2b8a16b2a: Type = 492;
1380 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1381 pub const dnnl_BAc2b8a16b2a: Type = 493;
1382 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1383 pub const dnnl_aBCde2c8b16c2b: Type = 494;
1384 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1385 pub const dnnl_aBCdef2c8b16c2b: Type = 495;
1386 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1387 pub const dnnl_aCBdef2c8b16c2b: Type = 496;
1388 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1389 pub const dnnl_aBCd2b8c16b4c: Type = 497;
1390 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1391 pub const dnnl_aBCde2b8c16b4c: Type = 498;
1392 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1393 pub const dnnl_BA4b8a16b2a: Type = 499;
1394 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1395 pub const dnnl_BA4b8a16b4a: Type = 500;
1396 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1397 pub const dnnl_aCB4c8b16c2b: Type = 501;
1398 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1399 pub const dnnl_aCB4c8b16c4b: Type = 502;
1400 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1401 pub const dnnl_BA16a16b: Type = 503;
1402 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1403 pub const dnnl_BA16a32b: Type = 504;
1404 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1405 pub const dnnl_BA16a48b: Type = 505;
1406 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1407 pub const dnnl_BA16a64b: Type = 506;
1408 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1409 pub const dnnl_aCB16c2b: Type = 507;
1410 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1411 pub const dnnl_aCB16c4b: Type = 508;
1412 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1413 pub const dnnl_BA16b2a: Type = 509;
1414 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1415 pub const dnnl_BA16b4a: Type = 510;
1416 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1417 pub const dnnl_aBC16b16c: Type = 511;
1418 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1419 pub const dnnl_aBC16b32c: Type = 512;
1420 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1421 pub const dnnl_AB16a16b: Type = 513;
1422 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1423 pub const dnnl_AB16a32b: Type = 514;
1424 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1425 pub const dnnl_ABcde16a16b2a: Type = 515;
1426 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1427 pub const dnnl_aBCdef16b16c2b: Type = 516;
1428 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1429 pub const dnnl_Acedb16a: Type = 517;
1430 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1431 pub const dnnl_aBdfec16b: Type = 518;
1432 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1433 pub const dnnl_abdEC64e2c: Type = 519;
1434 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1435 pub const dnnl_abdEC64e4c: Type = 520;
1436 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1437 pub const dnnl_aCB16b16c: Type = 521;
1438 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1439 pub const dnnl_aCB16b32c: Type = 522;
1440 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1441 pub const dnnl_aCB16b48c: Type = 523;
1442 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1443 pub const dnnl_aCB16b64c: Type = 524;
1444 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1445 pub const dnnl_aCB16b16c2b: Type = 525;
1446 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1447 pub const dnnl_aCB16b32c2b: Type = 526;
1448 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1449 pub const dnnl_aCB16b48c2b: Type = 527;
1450 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1451 pub const dnnl_aCB16b64c2b: Type = 528;
1452 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1453 pub const dnnl_aCB16b16c4b: Type = 529;
1454 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1455 pub const dnnl_aCB16b32c4b: Type = 530;
1456 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1457 pub const dnnl_aCB16b48c4b: Type = 531;
1458 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1459 pub const dnnl_aCB16b64c4b: Type = 532;
1460 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1461 pub const dnnl_abCd4c: Type = 533;
1462 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1463 pub const dnnl_abCde4c: Type = 534;
1464 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1465 pub const dnnl_abCdef4c: Type = 535;
1466 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1467 pub const dnnl_abCde32c: Type = 536;
1468 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1469 pub const dnnl_abCdef32c: Type = 537;
1470 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1471 pub const dnnl_ABcd16a32b: Type = 538;
1472 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1473 pub const dnnl_decbA8a: Type = 539;
1474 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1475 pub const dnnl_aCdefB16b32c2b: Type = 540;
1476 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1477 pub const dnnl_aCdefB16b32c4b: Type = 541;
1478 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1479 pub const dnnl_aCdefB16b48c2b: Type = 542;
1480 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1481 pub const dnnl_aCdefB16b48c4b: Type = 543;
1482 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1483 pub const dnnl_aCdefB16b64c2b: Type = 544;
1484 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1485 pub const dnnl_aCdefB16b64c4b: Type = 545;
1486 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1487 pub const dnnl_BcdeA16a32b2a: Type = 546;
1488 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1489 pub const dnnl_BcdeA16a32b4a: Type = 547;
1490 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1491 pub const dnnl_BcdeA16a48b2a: Type = 548;
1492 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1493 pub const dnnl_BcdeA16a48b4a: Type = 549;
1494 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1495 pub const dnnl_BcdeA16a64b2a: Type = 550;
1496 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1497 pub const dnnl_BcdeA16a64b4a: Type = 551;
1498 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1499 pub const dnnl_aCdefb32c: Type = 552;
1500 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1501 pub const dnnl_aCdefB32c2b: Type = 553;
1502 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1503 pub const dnnl_aCdefB32c4b: Type = 554;
1504 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1505 pub const dnnl_aCdefb48c: Type = 555;
1506 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1507 pub const dnnl_aCdefB48c2b: Type = 556;
1508 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1509 pub const dnnl_aCdefB48c4b: Type = 557;
1510 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1511 pub const dnnl_aCdefb64c: Type = 558;
1512 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1513 pub const dnnl_aCdefB64c2b: Type = 559;
1514 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1515 pub const dnnl_aCdefB64c4b: Type = 560;
1516 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1517 pub const dnnl_Bcdea32b: Type = 561;
1518 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1519 pub const dnnl_BcdeA32b2a: Type = 562;
1520 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1521 pub const dnnl_BcdeA32b4a: Type = 563;
1522 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1523 pub const dnnl_Bcdea48b: Type = 564;
1524 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1525 pub const dnnl_BcdeA48b2a: Type = 565;
1526 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1527 pub const dnnl_BcdeA48b4a: Type = 566;
1528 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1529 pub const dnnl_Bcdea64b: Type = 567;
1530 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1531 pub const dnnl_BcdeA64b2a: Type = 568;
1532 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1533 pub const dnnl_BcdeA64b4a: Type = 569;
1534 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1535 pub const dnnl_Bca32b: Type = 570;
1536 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1537 pub const dnnl_BcA32b2a: Type = 571;
1538 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1539 pub const dnnl_BcA32b4a: Type = 572;
1540 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1541 pub const dnnl_Bca48b: Type = 573;
1542 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1543 pub const dnnl_BcA48b2a: Type = 574;
1544 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1545 pub const dnnl_BcA48b4a: Type = 575;
1546 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1547 pub const dnnl_Bca64b: Type = 576;
1548 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1549 pub const dnnl_BcA64b2a: Type = 577;
1550 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1551 pub const dnnl_BcA64b4a: Type = 578;
1552 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1553 pub const dnnl_aCdb32c: Type = 579;
1554 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1555 pub const dnnl_aCdB32c2b: Type = 580;
1556 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1557 pub const dnnl_aCdB32c4b: Type = 581;
1558 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1559 pub const dnnl_aCdb48c: Type = 582;
1560 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1561 pub const dnnl_aCdB48c2b: Type = 583;
1562 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1563 pub const dnnl_aCdB48c4b: Type = 584;
1564 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1565 pub const dnnl_aCdb64c: Type = 585;
1566 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1567 pub const dnnl_aCdB64c2b: Type = 586;
1568 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1569 pub const dnnl_aCdB64c4b: Type = 587;
1570 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1571 pub const dnnl_BcA16a16b2a: Type = 588;
1572 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1573 pub const dnnl_BcA16a16b4a: Type = 589;
1574 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1575 pub const dnnl_BcdA16a16b2a: Type = 590;
1576 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1577 pub const dnnl_BcdA16a16b4a: Type = 591;
1578 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1579 pub const dnnl_BcdeA16a16b2a: Type = 592;
1580 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1581 pub const dnnl_BcdeA16a16b4a: Type = 593;
1582 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1583 pub const dnnl_aCdB16b16c2b: Type = 594;
1584 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1585 pub const dnnl_aCdB16b16c4b: Type = 595;
1586 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1587 pub const dnnl_aCdeB16b16c2b: Type = 596;
1588 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1589 pub const dnnl_aCdeB16b16c4b: Type = 597;
1590 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1591 pub const dnnl_aCdefB16b16c2b: Type = 598;
1592 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1593 pub const dnnl_aCdefB16b16c4b: Type = 599;
1594 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1595 pub const dnnl_BcA16a32b2a: Type = 600;
1596 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1597 pub const dnnl_BcA16a32b4a: Type = 601;
1598 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1599 pub const dnnl_BcA16a48b2a: Type = 602;
1600 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1601 pub const dnnl_BcA16a48b4a: Type = 603;
1602 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1603 pub const dnnl_BcA16a64b2a: Type = 604;
1604 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1605 pub const dnnl_BcA16a64b4a: Type = 605;
1606 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1607 pub const dnnl_aCdB16b32c2b: Type = 606;
1608 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1609 pub const dnnl_aCdB16b32c4b: Type = 607;
1610 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1611 pub const dnnl_aCdB16b48c2b: Type = 608;
1612 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1613 pub const dnnl_aCdB16b48c4b: Type = 609;
1614 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1615 pub const dnnl_aCdB16b64c2b: Type = 610;
1616 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1617 pub const dnnl_aCdB16b64c4b: Type = 611;
1618 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1619 pub const dnnl_BcdA16a32b2a: Type = 612;
1620 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1621 pub const dnnl_BcdA16a32b4a: Type = 613;
1622 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1623 pub const dnnl_BcdA16a48b2a: Type = 614;
1624 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1625 pub const dnnl_BcdA16a48b4a: Type = 615;
1626 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1627 pub const dnnl_BcdA16a64b2a: Type = 616;
1628 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1629 pub const dnnl_BcdA16a64b4a: Type = 617;
1630 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1631 pub const dnnl_aCdeB16b32c2b: Type = 618;
1632 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1633 pub const dnnl_aCdeB16b32c4b: Type = 619;
1634 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1635 pub const dnnl_aCdeB16b48c2b: Type = 620;
1636 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1637 pub const dnnl_aCdeB16b48c4b: Type = 621;
1638 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1639 pub const dnnl_aCdeB16b64c2b: Type = 622;
1640 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1641 pub const dnnl_aCdeB16b64c4b: Type = 623;
1642 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1643 pub const dnnl_Bca16b: Type = 624;
1644 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1645 pub const dnnl_BcA16b2a: Type = 625;
1646 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1647 pub const dnnl_BcA16b4a: Type = 626;
1648 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1649 pub const dnnl_Bcda16b: Type = 627;
1650 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1651 pub const dnnl_BcdA16b2a: Type = 628;
1652 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1653 pub const dnnl_BcdA16b4a: Type = 629;
1654 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1655 pub const dnnl_Bcdea16b: Type = 630;
1656 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1657 pub const dnnl_BcdeA16b2a: Type = 631;
1658 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1659 pub const dnnl_BcdeA16b4a: Type = 632;
1660 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1661 pub const dnnl_aCdb16c: Type = 633;
1662 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1663 pub const dnnl_aCdB16c2b: Type = 634;
1664 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1665 pub const dnnl_aCdB16c4b: Type = 635;
1666 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1667 pub const dnnl_aCdeb16c: Type = 636;
1668 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1669 pub const dnnl_aCdeB16c2b: Type = 637;
1670 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1671 pub const dnnl_aCdeB16c4b: Type = 638;
1672 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1673 pub const dnnl_aCdefb16c: Type = 639;
1674 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1675 pub const dnnl_aCdefB16c2b: Type = 640;
1676 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1677 pub const dnnl_aCdefB16c4b: Type = 641;
1678 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1679 pub const dnnl_Bcda32b: Type = 642;
1680 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1681 pub const dnnl_BcdA32b2a: Type = 643;
1682 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1683 pub const dnnl_BcdA32b4a: Type = 644;
1684 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1685 pub const dnnl_Bcda48b: Type = 645;
1686 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1687 pub const dnnl_BcdA48b2a: Type = 646;
1688 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1689 pub const dnnl_BcdA48b4a: Type = 647;
1690 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1691 pub const dnnl_Bcda64b: Type = 648;
1692 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1693 pub const dnnl_BcdA64b2a: Type = 649;
1694 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1695 pub const dnnl_BcdA64b4a: Type = 650;
1696 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1697 pub const dnnl_aCdeb32c: Type = 651;
1698 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1699 pub const dnnl_aCdeB32c2b: Type = 652;
1700 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1701 pub const dnnl_aCdeB32c4b: Type = 653;
1702 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1703 pub const dnnl_aCdeb48c: Type = 654;
1704 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1705 pub const dnnl_aCdeB48c2b: Type = 655;
1706 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1707 pub const dnnl_aCdeB48c4b: Type = 656;
1708 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1709 pub const dnnl_aCdeb64c: Type = 657;
1710 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1711 pub const dnnl_aCdeB64c2b: Type = 658;
1712 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1713 pub const dnnl_aCdeB64c4b: Type = 659;
1714 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1715 pub const dnnl_Acb24a: Type = 660;
1716 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1717 pub const dnnl_Acdb24a: Type = 661;
1718 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1719 pub const dnnl_Acdeb24a: Type = 662;
1720 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1721 pub const dnnl_aBdc24b: Type = 663;
1722 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1723 pub const dnnl_aBdec24b: Type = 664;
1724 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1725 pub const dnnl_aBdefc24b: Type = 665;
1726 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1727 pub const dnnl_abDc16d: Type = 666;
1728 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1729 pub const dnnl_abdEc16e: Type = 667;
1730 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1731 pub const dnnl_abdCe16c: Type = 668;
1732 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1733 pub const dnnl_AcB24a2b: Type = 669;
1734 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1735 pub const dnnl_AcdB24a2b: Type = 670;
1736 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1737 pub const dnnl_AcdeB24a2b: Type = 671;
1738 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1739 pub const dnnl_aBdC24b2c: Type = 672;
1740 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1741 pub const dnnl_aBdeC24b2c: Type = 673;
1742 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1743 pub const dnnl_aBdefC24b2c: Type = 674;
1744 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1745 pub const dnnl_AcB8a2b: Type = 675;
1746 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1747 pub const dnnl_AcdB8a2b: Type = 676;
1748 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1749 pub const dnnl_AcdeB8a2b: Type = 677;
1750 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1751 pub const dnnl_aBdC8b2c: Type = 678;
1752 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1753 pub const dnnl_aBdeC8b2c: Type = 679;
1754 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1755 pub const dnnl_aBdefC8b2c: Type = 680;
1756 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1757 pub const dnnl_AB8b32a: Type = 681;
1758 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1759 pub const dnnl_ABc8b32a: Type = 682;
1760 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1761 pub const dnnl_ABcd8b32a: Type = 683;
1762 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1763 pub const dnnl_ABcde8b32a: Type = 684;
1764 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1765 pub const dnnl_AB8b24a: Type = 685;
1766 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1767 pub const dnnl_ABc8b24a: Type = 686;
1768 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1769 pub const dnnl_ABcd8b24a: Type = 687;
1770 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1771 pub const dnnl_ABcde8b24a: Type = 688;
1772 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1773 pub const dnnl_AB8b16a: Type = 689;
1774 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1775 pub const dnnl_ABc8b16a: Type = 690;
1776 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1777 pub const dnnl_ABcd8b16a: Type = 691;
1778 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1779 pub const dnnl_ABcde8b16a: Type = 692;
1780 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1781 pub const dnnl_AB8b8a: Type = 693;
1782 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1783 pub const dnnl_AB4b8a4b: Type = 694;
1784 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1785 pub const dnnl_AB4b24a4b: Type = 695;
1786 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1787 pub const dnnl_ABc4b8a4b: Type = 696;
1788 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1789 pub const dnnl_ABc4b24a4b: Type = 697;
1790 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1791 pub const dnnl_ABcd4b8a4b: Type = 698;
1792 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1793 pub const dnnl_ABcd4b24a4b: Type = 699;
1794 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1795 pub const dnnl_ABcde4b8a4b: Type = 700;
1796 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1797 pub const dnnl_ABcde4b24a4b: Type = 701;
1798 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1799 pub const dnnl_AB8b24a2b: Type = 702;
1800 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1801 pub const dnnl_ABc8b24a2b: Type = 703;
1802 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1803 pub const dnnl_ABcd8b24a2b: Type = 704;
1804 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1805 pub const dnnl_ABcde8b24a2b: Type = 705;
1806 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1807 pub const dnnl_AB8b8a2b: Type = 706;
1808 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1809 pub const dnnl_ABc8b8a2b: Type = 707;
1810 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1811 pub const dnnl_ABcd8b8a2b: Type = 708;
1812 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1813 pub const dnnl_ABcde8b8a2b: Type = 709;
1814 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1815 pub const dnnl_AcB24a4b: Type = 710;
1816 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1817 pub const dnnl_AcdB24a4b: Type = 711;
1818 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1819 pub const dnnl_AcdeB24a4b: Type = 712;
1820 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1821 pub const dnnl_aBdC24b4c: Type = 713;
1822 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1823 pub const dnnl_aBdeC24b4c: Type = 714;
1824 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1825 pub const dnnl_aBdefC24b4c: Type = 715;
1826 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1827 pub const dnnl_AcB8a4b: Type = 716;
1828 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1829 pub const dnnl_AcdB8a4b: Type = 717;
1830 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1831 pub const dnnl_AcdeB8a4b: Type = 718;
1832 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1833 pub const dnnl_aBdC8b4c: Type = 719;
1834 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1835 pub const dnnl_aBdeC8b4c: Type = 720;
1836 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1837 pub const dnnl_aBdefC8b4c: Type = 721;
1838 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1839 pub const dnnl_Bca8b: Type = 722;
1840 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1841 pub const dnnl_BcA8b2a: Type = 723;
1842 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1843 pub const dnnl_Bcda8b: Type = 724;
1844 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1845 pub const dnnl_BcdA8b2a: Type = 725;
1846 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1847 pub const dnnl_Bcdea8b: Type = 726;
1848 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1849 pub const dnnl_BcdeA8b2a: Type = 727;
1850 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1851 pub const dnnl_aCdb8c: Type = 728;
1852 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1853 pub const dnnl_aCdB8c2b: Type = 729;
1854 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1855 pub const dnnl_aCdeb8c: Type = 730;
1856 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1857 pub const dnnl_aCdeB8c2b: Type = 731;
1858 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1859 pub const dnnl_aCdefb8c: Type = 732;
1860 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1861 pub const dnnl_aCdefB8c2b: Type = 733;
1862 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1863 pub const dnnl_Bca24b: Type = 734;
1864 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1865 pub const dnnl_BcA24b2a: Type = 735;
1866 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1867 pub const dnnl_Bcda24b: Type = 736;
1868 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1869 pub const dnnl_BcdA24b2a: Type = 737;
1870 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1871 pub const dnnl_Bcdea24b: Type = 738;
1872 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1873 pub const dnnl_BcdeA24b2a: Type = 739;
1874 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1875 pub const dnnl_aCdb24c: Type = 740;
1876 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1877 pub const dnnl_aCdB24c2b: Type = 741;
1878 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1879 pub const dnnl_aCdeb24c: Type = 742;
1880 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1881 pub const dnnl_aCdeB24c2b: Type = 743;
1882 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1883 pub const dnnl_aCdefb24c: Type = 744;
1884 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1885 pub const dnnl_aCdefB24c2b: Type = 745;
1886 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1887 pub const dnnl_BcA8b4a: Type = 746;
1888 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1889 pub const dnnl_BcdA8b4a: Type = 747;
1890 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1891 pub const dnnl_BcdeA8b4a: Type = 748;
1892 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1893 pub const dnnl_aCdB8c4b: Type = 749;
1894 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1895 pub const dnnl_aCdeB8c4b: Type = 750;
1896 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1897 pub const dnnl_aCdefB8c4b: Type = 751;
1898 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1899 pub const dnnl_BcA24b4a: Type = 752;
1900 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1901 pub const dnnl_BcdA24b4a: Type = 753;
1902 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1903 pub const dnnl_BcdeA24b4a: Type = 754;
1904 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1905 pub const dnnl_aCdB24c4b: Type = 755;
1906 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1907 pub const dnnl_aCdeB24c4b: Type = 756;
1908 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1909 pub const dnnl_aCdefB24c4b: Type = 757;
1910 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1911 pub const dnnl_AB16b48a: Type = 758;
1912 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1913 pub const dnnl_ABc16b48a: Type = 759;
1914 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1915 pub const dnnl_ABcd16b48a: Type = 760;
1916 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1917 pub const dnnl_ABcde16b48a: Type = 761;
1918 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1919 pub const dnnl_ABc16a4b: Type = 762;
1920 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1921 pub const dnnl_ABcd16a4b: Type = 763;
1922 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1923 pub const dnnl_ABcde16a4b: Type = 764;
1924 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1925 pub const dnnl_defcbA16a: Type = 765;
1926 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1927 pub const dnnl_defcbA8a: Type = 766;
1928 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1929 pub const dnnl_AcB16b64a: Type = 767;
1930 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1931 pub const dnnl_AcdB16b64a: Type = 768;
1932 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1933 pub const dnnl_AcdeB16b64a: Type = 769;
1934 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1935 pub const dnnl_AcB16b48a: Type = 770;
1936 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1937 pub const dnnl_AcdB16b48a: Type = 771;
1938 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1939 pub const dnnl_AcdeB16b48a: Type = 772;
1940 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1941 pub const dnnl_AcB16b32a: Type = 773;
1942 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1943 pub const dnnl_AcdB16b32a: Type = 774;
1944 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1945 pub const dnnl_AcdeB16b32a: Type = 775;
1946 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1947 pub const dnnl_AcB16b16a: Type = 776;
1948 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1949 pub const dnnl_AcdB16b16a: Type = 777;
1950 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1951 pub const dnnl_AcdeB16b16a: Type = 778;
1952 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1953 pub const dnnl_AcB8b32a: Type = 779;
1954 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1955 pub const dnnl_AcdB8b32a: Type = 780;
1956 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1957 pub const dnnl_AcdeB8b32a: Type = 781;
1958 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1959 pub const dnnl_AcB8b24a: Type = 782;
1960 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1961 pub const dnnl_AcdB8b24a: Type = 783;
1962 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1963 pub const dnnl_AcdeB8b24a: Type = 784;
1964 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1965 pub const dnnl_AcB8b16a: Type = 785;
1966 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1967 pub const dnnl_AcdB8b16a: Type = 786;
1968 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1969 pub const dnnl_AcdeB8b16a: Type = 787;
1970 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1971 pub const dnnl_AcB8b8a: Type = 788;
1972 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1973 pub const dnnl_AcdB8b8a: Type = 789;
1974 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1975 pub const dnnl_AcdeB8b8a: Type = 790;
1976 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1977 pub const dnnl_AcB8b64a2b: Type = 791;
1978 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1979 pub const dnnl_AcdB8b64a2b: Type = 792;
1980 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1981 pub const dnnl_AcdeB8b64a2b: Type = 793;
1982 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1983 pub const dnnl_AcB8b32a2b: Type = 794;
1984 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1985 pub const dnnl_AcdB8b32a2b: Type = 795;
1986 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1987 pub const dnnl_AcdeB8b32a2b: Type = 796;
1988 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1989 pub const dnnl_AcB8b24a2b: Type = 797;
1990 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1991 pub const dnnl_AcdB8b24a2b: Type = 798;
1992 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1993 pub const dnnl_AcdeB8b24a2b: Type = 799;
1994 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1995 pub const dnnl_AcB8b16a2b: Type = 800;
1996 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1997 pub const dnnl_AcdB8b16a2b: Type = 801;
1998 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
1999 pub const dnnl_AcdeB8b16a2b: Type = 802;
2000 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2001 pub const dnnl_AcB8b8a2b: Type = 803;
2002 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2003 pub const dnnl_AcdB8b8a2b: Type = 804;
2004 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2005 pub const dnnl_AcdeB8b8a2b: Type = 805;
2006 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2007 pub const dnnl_AcB4b64a4b: Type = 806;
2008 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2009 pub const dnnl_AcdB4b64a4b: Type = 807;
2010 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2011 pub const dnnl_AcdeB4b64a4b: Type = 808;
2012 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2013 pub const dnnl_AcB4b32a4b: Type = 809;
2014 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2015 pub const dnnl_AcdB4b32a4b: Type = 810;
2016 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2017 pub const dnnl_AcdeB4b32a4b: Type = 811;
2018 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2019 pub const dnnl_AcB4b24a4b: Type = 812;
2020 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2021 pub const dnnl_AcdB4b24a4b: Type = 813;
2022 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2023 pub const dnnl_AcdeB4b24a4b: Type = 814;
2024 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2025 pub const dnnl_AcB4b16a4b: Type = 815;
2026 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2027 pub const dnnl_AcdB4b16a4b: Type = 816;
2028 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2029 pub const dnnl_AcdeB4b16a4b: Type = 817;
2030 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2031 pub const dnnl_AcB4b8a4b: Type = 818;
2032 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2033 pub const dnnl_AcdB4b8a4b: Type = 819;
2034 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2035 pub const dnnl_AcdeB4b8a4b: Type = 820;
2036 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2037 pub const dnnl_Ab4a: Type = 821;
2038 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2039 pub const dnnl_Ab8a: Type = 822;
2040 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2041 pub const dnnl_BA4b4a: Type = 823;
2042 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2043 pub const dnnl_BA8b4a: Type = 824;
2044 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2045 pub const dnnl_BA2a24b: Type = 825;
2046 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2047 pub const dnnl_aCB2b24c: Type = 826;
2048 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2049 pub const dnnl_BA2a8b: Type = 827;
2050 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2051 pub const dnnl_aCB2b8c: Type = 828;
2052 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2053 pub const dnnl_BA8a24b: Type = 829;
2054 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2055 pub const dnnl_aCB8b24c: Type = 830;
2056 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2057 pub const dnnl_BA8a16b: Type = 831;
2058 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2059 pub const dnnl_aCB8b16c: Type = 832;
2060 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2061 pub const dnnl_BA8a8b: Type = 833;
2062 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2063 pub const dnnl_aCB8b8c: Type = 834;
2064 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2065 pub const dnnl_bcad: Type = 835;
2066 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2067 pub const dnnl_cabd: Type = 836;
2068 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2069 pub const dnnl_dabc: Type = 837;
2070 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2071 pub const dnnl_Ab32a: Type = 838;
2072 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2073 pub const dnnl_aCBd8b8c: Type = 839;
2074 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2075 pub const dnnl_aCBde8b8c: Type = 840;
2076 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2077 pub const dnnl_BAc8a8b: Type = 841;
2078 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2079 pub const dnnl_BAcd8a8b: Type = 842;
2080 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2081 pub const dnnl_BAcde8a8b: Type = 843;
2082 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2083 pub const dnnl_aCBdef8b8c: Type = 844;
2084 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2085 pub const dnnl_abdEC16e4c: Type = 845;
2086 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2087 pub const dnnl_abDC16d4c: Type = 846;
2088 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2089 pub const dnnl_BA24b8a: Type = 847;
2090 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2091 pub const dnnl_aCB24c8b: Type = 848;
2092 #[doc = " 6D tensor blocked by 2nd dimension with block size 4"]
2093 pub const dnnl_abDC24d8c: Type = 849;
2094 #[doc = " Just a sentinel, not real memory format tag. Must be changed after new\n format tag is added."]
2095 pub const dnnl_format_tag_last: Type = 850;
2096 #[doc = " 1D tensor, an alias to #dnnl_a"]
2097 pub const dnnl_x: Type = 2;
2098 #[doc = " 2D CNN activations tensor, an alias to #dnnl_ab"]
2099 pub const dnnl_nc: Type = 3;
2100 #[doc = " 2D CNN activations tensor, an alias to #dnnl_ba"]
2101 pub const dnnl_cn: Type = 14;
2102 #[doc = " 2D RNN statistics tensor, an alias to #dnnl_ab"]
2103 pub const dnnl_tn: Type = 3;
2104 #[doc = " 2D RNN statistics tensor, an alias to #dnnl_ba"]
2105 pub const dnnl_nt: Type = 14;
2106 #[doc = " 3D CNN activations tensor, an alias to #dnnl_abc"]
2107 pub const dnnl_ncw: Type = 4;
2108 #[doc = " 3D CNN activations tensor, an alias to #dnnl_acb"]
2109 pub const dnnl_nwc: Type = 15;
2110 #[doc = " 4D CNN activations tensor, an alias to #dnnl_abcd"]
2111 pub const dnnl_nchw: Type = 5;
2112 #[doc = " 4D CNN activations tensor, an alias to #dnnl_acdb"]
2113 pub const dnnl_nhwc: Type = 22;
2114 #[doc = " 4D CNN activations tensor, an alias to #dnnl_bcda"]
2115 pub const dnnl_chwn: Type = 26;
2116 #[doc = " 5D CNN activations tensor, an alias to #dnnl_abcde"]
2117 pub const dnnl_ncdhw: Type = 6;
2118 #[doc = " 5D CNN activations tensor, an alias to #dnnl_acdeb"]
2119 pub const dnnl_ndhwc: Type = 33;
2120 #[doc = " 2D CNN weights tensor, an alias to #dnnl_ab"]
2121 pub const dnnl_oi: Type = 3;
2122 #[doc = " 2D CNN weights tensor, an alias to #dnnl_ba"]
2123 pub const dnnl_io: Type = 14;
2124 #[doc = " 3D CNN weights tensor, an alias to #dnnl_abc"]
2125 pub const dnnl_oiw: Type = 4;
2126 #[doc = " 3D CNN weights tensor, an alias to #dnnl_acb"]
2127 pub const dnnl_owi: Type = 15;
2128 #[doc = " 3D CNN weights tensor, an alias to #dnnl_cba"]
2129 pub const dnnl_wio: Type = 19;
2130 #[doc = " 3D CNN weights tensor, an alias to #dnnl_cab"]
2131 pub const dnnl_woi: Type = 18;
2132 #[doc = " 3D CNN weights tensor, an alias to #dnnl_bca"]
2133 pub const dnnl_iwo: Type = 17;
2134 #[doc = " 4D CNN weights tensor, an alias to #dnnl_abcd"]
2135 pub const dnnl_oihw: Type = 5;
2136 #[doc = " 4D CNN weights tensor, an alias to #dnnl_cdba"]
2137 pub const dnnl_hwio: Type = 28;
2138 #[doc = " 4D CNN weights tensor, an alias to #dnnl_cdab"]
2139 pub const dnnl_hwoi: Type = 27;
2140 #[doc = " 4D CNN weights tensor, an alias to #dnnl_acdb"]
2141 pub const dnnl_ohwi: Type = 22;
2142 #[doc = " 4D CNN weights tensor, an alias to #dnnl_bcda"]
2143 pub const dnnl_ihwo: Type = 26;
2144 #[doc = " 4D CNN weights tensor, an alias to #dnnl_bacd"]
2145 pub const dnnl_iohw: Type = 25;
2146 #[doc = " 5D CNN weights tensor, an alias to #dnnl_abcde"]
2147 pub const dnnl_oidhw: Type = 6;
2148 #[doc = " 5D CNN weights tensor, an alias to #dnnl_bacde"]
2149 pub const dnnl_iodhw: Type = 35;
2150 #[doc = " 5D CNN weights tensor, an alias to #dnnl_cdeba"]
2151 pub const dnnl_dhwio: Type = 38;
2152 #[doc = " 5D CNN weights tensor, an alias to #dnnl_cdeab"]
2153 pub const dnnl_dhwoi: Type = 37;
2154 #[doc = " 5D CNN weights tensor, an alias to #dnnl_acdeb"]
2155 pub const dnnl_odhwi: Type = 33;
2156 #[doc = " 5D CNN weights tensor, an alias to #dnnl_bcdea"]
2157 pub const dnnl_idhwo: Type = 36;
2158 #[doc = " 4D CNN weights tensor (incl. groups), an alias to #dnnl_abcd"]
2159 pub const dnnl_goiw: Type = 5;
2160 #[doc = " 4D CNN weights tensor (incl. groups), an alias to #dnnl_abdc"]
2161 pub const dnnl_gowi: Type = 20;
2162 #[doc = " 4D CNN weights tensor (incl. groups), an alias to #dnnl_dcab"]
2163 pub const dnnl_wigo: Type = 29;
2164 #[doc = " 5D CNN weights tensor (incl. groups), an alias to #dnnl_abcde"]
2165 pub const dnnl_goihw: Type = 6;
2166 #[doc = " 5D CNN weights tensor (incl. groups), an alias to #dnnl_abdec"]
2167 pub const dnnl_gohwi: Type = 31;
2168 #[doc = " 5D CNN weights tensor (incl. groups), an alias to #dnnl_decab"]
2169 pub const dnnl_hwigo: Type = 39;
2170 #[doc = " 5D CNN weights tensor (incl. groups), an alias to #dnnl_acbde"]
2171 pub const dnnl_giohw: Type = 32;
2172 #[doc = " 6D CNN weights tensor (incl. groups), an alias to #dnnl_abcdef"]
2173 pub const dnnl_goidhw: Type = 7;
2174 #[doc = " 6D CNN weights tensor (incl. groups), an alias to #dnnl_abdefc"]
2175 pub const dnnl_godhwi: Type = 41;
2176 #[doc = " 6D CNN weights tensor (incl. groups), an alias to #dnnl_acbdef"]
2177 pub const dnnl_giodhw: Type = 43;
2178 #[doc = " 6D CNN weights tensor (incl. groups), an alias to #dnnl_defcab"]
2179 pub const dnnl_dhwigo: Type = 45;
2180 #[doc = " 3D RNN data tensor in the format (seq_length, batch, input channels),\n an alias to #dnnl_abc."]
2181 pub const dnnl_tnc: Type = 4;
2182 #[doc = " 3D RNN data tensor in the format (batch, seq_length, input channels),\n an alias to #dnnl_bac."]
2183 pub const dnnl_ntc: Type = 16;
2184 #[doc = " 4D RNN states tensor in the format (num_layers, num_directions,\n batch, state channels), an alias to #dnnl_abcd."]
2185 pub const dnnl_ldnc: Type = 5;
2186 #[doc = " 5D RNN weights tensor in the format (num_layers, num_directions,\n input_channels, num_gates, output_channels), an alias to #dnnl_abcde.\n\n - For LSTM cells, the gates order is input, forget, candidate\n and output gate.\n - For GRU cells, the gates order is update, reset and output gate."]
2187 pub const dnnl_ldigo: Type = 6;
2188 #[doc = " 5D RNN weights tensor in the format (num_layers, num_directions,\n num_gates, output_channels, input_channels), an alias to #dnnl_abdec.\n\n - For LSTM cells, the gates order is input, forget, candidate\n and output gate.\n - For GRU cells, the gates order is update, reset and output gate."]
2189 pub const dnnl_ldgoi: Type = 31;
2190 #[doc = " 4D LSTM projection tensor in the format (num_layers, num_directions,\n num_channels_in_hidden_state, num_channels_in_recurrent_projection),\n an alias to #dnnl_abcd."]
2191 pub const dnnl_ldio: Type = 5;
2192 #[doc = " 4D LSTM projection tensor in the format (num_layers, num_directions,\n num_channels_in_recurrent_projection, num_channels_in_hidden_state),\n an alias to #dnnl_abdc."]
2193 pub const dnnl_ldoi: Type = 20;
2194 #[doc = " 4D RNN bias tensor in the format (num_layers, num_directions,\n num_gates, output_channels), an alias to #dnnl_abcd.\n\n - For LSTM cells, the gates order is input, forget, candidate\n and output gate.\n - For GRU cells, the gates order is update, reset and output gate."]
2195 pub const dnnl_ldgo: Type = 5;
2196 #[doc = " 5D LSTM projection tensor"]
2197 pub const dnnl_ldOi16o: Type = 666;
2198 #[doc = " 5D LSTM projection tensor"]
2199 pub const dnnl_ldOi32o: Type = 267;
2200 #[doc = " 5D LSTM projection tensor"]
2201 pub const dnnl_ldOI16o4i: Type = 846;
2202 #[doc = " 5D LSTM projection tensor"]
2203 pub const dnnl_ldOI32o4i: Type = 268;
2204 #[doc = " 5D LSTM projection tensor"]
2205 pub const dnnl_ldIo32i: Type = 378;
2206 #[doc = " 6D RNN weights tensor"]
2207 pub const dnnl_ldgOi16o: Type = 667;
2208 #[doc = " 6D RNN weights tensor"]
2209 pub const dnnl_ldgOI16o4i: Type = 845;
2210 #[doc = " 6D RNN weights tensor"]
2211 pub const dnnl_ldgOi32o: Type = 269;
2212 #[doc = " 6D RNN weights tensor"]
2213 pub const dnnl_ldgOI32o2i: Type = 270;
2214 #[doc = " 6D RNN weights tensor"]
2215 pub const dnnl_ldgOI32o4i: Type = 271;
2216 #[doc = " 6D RNN weights tensor"]
2217 pub const dnnl_ldgOI64o2i: Type = 519;
2218 #[doc = " 6D RNN weights tensor"]
2219 pub const dnnl_ldgOI64o4i: Type = 520;
2220 #[doc = " 6D RNN weights tensor"]
2221 pub const dnnl_ldgIo16i: Type = 668;
2222 #[doc = " 6D RNN weights tensor"]
2223 pub const dnnl_ldgIo32i: Type = 379;
2224 #[doc = " 6D RNN weights tensor"]
2225 pub const dnnl_ldgIO32i2o: Type = 380;
2226 #[doc = " 5D CNN activations tensor blocked by channels with block size 32,\n an alias to #dnnl_aBcde32b"]
2227 pub const dnnl_nCdhw32c: Type = 127;
2228 #[doc = " 5D CNN activations tensor blocked by channels with block size 16,\n an alias to #dnnl_aBcde16b"]
2229 pub const dnnl_nCdhw16c: Type = 121;
2230 #[doc = " 5D CNN activations tensor blocked by channels with block size 4,\n an alias to #dnnl_aBcde4b"]
2231 pub const dnnl_nCdhw4c: Type = 128;
2232 #[doc = " 5D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBcde8b"]
2233 pub const dnnl_nCdhw8c: Type = 142;
2234 #[doc = " 4D CNN activations tensor blocked by channels with block size 32,\n an alias to #dnnl_aBcd32b"]
2235 pub const dnnl_nChw32c: Type = 83;
2236 #[doc = " 4D CNN activations tensor blocked by channels with block size 16,\n an alias to #dnnl_aBcd16b"]
2237 pub const dnnl_nChw16c: Type = 78;
2238 #[doc = " 4D CNN activations tensor blocked by channels with block size 4,\n an alias to #dnnl_aBcd4b"]
2239 pub const dnnl_nChw4c: Type = 84;
2240 #[doc = " 4D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBcd8b"]
2241 pub const dnnl_nChw8c: Type = 102;
2242 #[doc = " 3D CNN activations tensor blocked by channels with block size 32,\n an alias to #dnnl_aBc32b"]
2243 pub const dnnl_nCw32c: Type = 59;
2244 #[doc = " 3D CNN activations tensor blocked by channels with block size 16,\n an alias to #dnnl_aBc16b"]
2245 pub const dnnl_nCw16c: Type = 56;
2246 #[doc = " 3D CNN activations tensor blocked by channels with block size 4,\n an alias to #dnnl_aBc4b"]
2247 pub const dnnl_nCw4c: Type = 60;
2248 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2249 pub const dnnl_nCw8c: Type = 69;
2250 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2251 pub const dnnl_NCw16n16c: Type = 53;
2252 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2253 pub const dnnl_NCdhw16n16c: Type = 116;
2254 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2255 pub const dnnl_NChw16n16c: Type = 75;
2256 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2257 pub const dnnl_NCw32n16c: Type = 367;
2258 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2259 pub const dnnl_NChw32n16c: Type = 368;
2260 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2261 pub const dnnl_NChw16n32c: Type = 538;
2262 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2263 pub const dnnl_NCdhw32n16c: Type = 369;
2264 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2265 pub const dnnl_NCw32n32c: Type = 54;
2266 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2267 pub const dnnl_NChw32n32c: Type = 77;
2268 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2269 pub const dnnl_NCdhw32n32c: Type = 148;
2270 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2271 pub const dnnl_OI16i16o: Type = 237;
2272 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2273 pub const dnnl_OI16i32o: Type = 238;
2274 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2275 pub const dnnl_OI16i48o: Type = 758;
2276 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2277 pub const dnnl_OI16i64o: Type = 239;
2278 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2279 pub const dnnl_OI8i8o2i: Type = 706;
2280 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2281 pub const dnnl_OI8i16o2i: Type = 240;
2282 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2283 pub const dnnl_OI8i24o2i: Type = 702;
2284 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2285 pub const dnnl_OI8i32o2i: Type = 241;
2286 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2287 pub const dnnl_OI8i64o2i: Type = 242;
2288 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2289 pub const dnnl_OI4i8o4i: Type = 694;
2290 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2291 pub const dnnl_OI4i16o4i: Type = 243;
2292 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2293 pub const dnnl_OI4i24o4i: Type = 695;
2294 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2295 pub const dnnl_OI4i32o4i: Type = 244;
2296 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2297 pub const dnnl_OI4i64o4i: Type = 245;
2298 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2299 pub const dnnl_OI16i16o4i: Type = 246;
2300 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2301 pub const dnnl_OI8i32o: Type = 681;
2302 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2303 pub const dnnl_OI8i24o: Type = 685;
2304 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2305 pub const dnnl_OI8i16o: Type = 689;
2306 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2307 pub const dnnl_OI8i8o: Type = 693;
2308 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2309 pub const dnnl_IOw8o8i: Type = 841;
2310 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2311 pub const dnnl_IOw16o16i: Type = 222;
2312 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2313 pub const dnnl_IOw16i16o: Type = 223;
2314 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2315 pub const dnnl_OIw16i16o: Type = 57;
2316 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2317 pub const dnnl_OwI16i16o: Type = 776;
2318 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2319 pub const dnnl_OIw16i32o: Type = 231;
2320 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2321 pub const dnnl_OwI16i32o: Type = 773;
2322 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2323 pub const dnnl_OIw16i48o: Type = 759;
2324 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2325 pub const dnnl_OwI16i48o: Type = 770;
2326 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2327 pub const dnnl_OIw16i64o: Type = 232;
2328 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2329 pub const dnnl_OwI16i64o: Type = 767;
2330 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2331 pub const dnnl_OIw16o16i: Type = 53;
2332 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2333 pub const dnnl_Oiw16o: Type = 52;
2334 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2335 pub const dnnl_OIw4i8o4i: Type = 696;
2336 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2337 pub const dnnl_OwI4i8o4i: Type = 818;
2338 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2339 pub const dnnl_OIw4i16o4i: Type = 61;
2340 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2341 pub const dnnl_OwI4i16o4i: Type = 815;
2342 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2343 pub const dnnl_OIw4i24o4i: Type = 697;
2344 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2345 pub const dnnl_OwI4i24o4i: Type = 812;
2346 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2347 pub const dnnl_OIw4i32o4i: Type = 233;
2348 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2349 pub const dnnl_OwI4i32o4i: Type = 809;
2350 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2351 pub const dnnl_OIw4i64o4i: Type = 234;
2352 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2353 pub const dnnl_OwI4i64o4i: Type = 806;
2354 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2355 pub const dnnl_OIw2i8o4i: Type = 62;
2356 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2357 pub const dnnl_OIw16i16o4i: Type = 63;
2358 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2359 pub const dnnl_OIw16i16o2i: Type = 64;
2360 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2361 pub const dnnl_OIw16o16i2o: Type = 275;
2362 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2363 pub const dnnl_OIw4i4o: Type = 65;
2364 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2365 pub const dnnl_OIw4o4i: Type = 55;
2366 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2367 pub const dnnl_Oiw4o: Type = 58;
2368 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2369 pub const dnnl_OIw8i8o2i: Type = 707;
2370 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2371 pub const dnnl_OwI8i8o2i: Type = 803;
2372 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2373 pub const dnnl_OIw8i16o2i: Type = 70;
2374 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2375 pub const dnnl_OwI8i16o2i: Type = 800;
2376 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2377 pub const dnnl_OIw8i24o2i: Type = 703;
2378 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2379 pub const dnnl_OwI8i24o2i: Type = 797;
2380 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2381 pub const dnnl_OIw8i32o2i: Type = 235;
2382 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2383 pub const dnnl_OwI8i32o2i: Type = 794;
2384 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2385 pub const dnnl_OIw8i64o2i: Type = 236;
2386 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2387 pub const dnnl_OwI8i64o2i: Type = 791;
2388 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2389 pub const dnnl_OIw8i8o: Type = 72;
2390 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2391 pub const dnnl_OwI8i8o: Type = 788;
2392 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2393 pub const dnnl_OIw8o16i2o: Type = 66;
2394 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2395 pub const dnnl_IOw8o16i2o: Type = 71;
2396 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2397 pub const dnnl_OIw8o8i: Type = 67;
2398 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2399 pub const dnnl_OIw8o4i: Type = 68;
2400 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2401 pub const dnnl_Owi16o: Type = 202;
2402 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2403 pub const dnnl_OwI16o2i: Type = 203;
2404 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2405 pub const dnnl_OwI16o4i: Type = 204;
2406 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2407 pub const dnnl_Iwo8i: Type = 722;
2408 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2409 pub const dnnl_IwO8i2o: Type = 723;
2410 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2411 pub const dnnl_IwO8i4o: Type = 746;
2412 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2413 pub const dnnl_Iwo16i: Type = 624;
2414 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2415 pub const dnnl_IwO16i2o: Type = 625;
2416 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2417 pub const dnnl_IwO16i4o: Type = 626;
2418 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2419 pub const dnnl_Iwo24i: Type = 734;
2420 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2421 pub const dnnl_IwO24i2o: Type = 735;
2422 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2423 pub const dnnl_IwO24i4o: Type = 752;
2424 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2425 pub const dnnl_Owi4o: Type = 205;
2426 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2427 pub const dnnl_Owi8o: Type = 206;
2428 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2429 pub const dnnl_OwI8o2i: Type = 675;
2430 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2431 pub const dnnl_OIw8i32o: Type = 682;
2432 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2433 pub const dnnl_OwI8i32o: Type = 779;
2434 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2435 pub const dnnl_OIw8i24o: Type = 686;
2436 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2437 pub const dnnl_OwI8i24o: Type = 782;
2438 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2439 pub const dnnl_OIw8i16o: Type = 690;
2440 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2441 pub const dnnl_OwI8i16o: Type = 785;
2442 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2443 pub const dnnl_OwI8o4i: Type = 716;
2444 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2445 pub const dnnl_IOhw16i16o: Type = 225;
2446 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2447 pub const dnnl_IOhw8o8i: Type = 842;
2448 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2449 pub const dnnl_IOhw16o16i: Type = 224;
2450 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2451 pub const dnnl_Ohwi16o: Type = 211;
2452 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2453 pub const dnnl_OhwI16o2i: Type = 212;
2454 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2455 pub const dnnl_OhwI16o4i: Type = 213;
2456 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2457 pub const dnnl_Ihwo8i: Type = 724;
2458 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2459 pub const dnnl_IhwO8i2o: Type = 725;
2460 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2461 pub const dnnl_IhwO8i4o: Type = 747;
2462 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2463 pub const dnnl_Ihwo16i: Type = 627;
2464 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2465 pub const dnnl_IhwO16i2o: Type = 628;
2466 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2467 pub const dnnl_IhwO16i4o: Type = 629;
2468 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2469 pub const dnnl_Ihwo24i: Type = 736;
2470 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2471 pub const dnnl_IhwO24i2o: Type = 737;
2472 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2473 pub const dnnl_IhwO24i4o: Type = 753;
2474 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2475 pub const dnnl_Ohwi24o: Type = 661;
2476 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2477 pub const dnnl_Ohwi32o: Type = 214;
2478 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2479 pub const dnnl_Ohwi4o: Type = 215;
2480 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2481 pub const dnnl_Ohwi8o: Type = 216;
2482 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2483 pub const dnnl_OhwI8o2i: Type = 676;
2484 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2485 pub const dnnl_OhwI8o4i: Type = 717;
2486 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2487 pub const dnnl_OIhw16i16o: Type = 79;
2488 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2489 pub const dnnl_OhwI16i16o: Type = 777;
2490 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2491 pub const dnnl_OIhw16i32o: Type = 247;
2492 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2493 pub const dnnl_OhwI16i32o: Type = 774;
2494 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2495 pub const dnnl_OIhw16i48o: Type = 760;
2496 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2497 pub const dnnl_OhwI16i48o: Type = 771;
2498 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2499 pub const dnnl_OIhw16i64o: Type = 248;
2500 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2501 pub const dnnl_OhwI16i64o: Type = 768;
2502 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2503 pub const dnnl_OIhw16o16i: Type = 75;
2504 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2505 pub const dnnl_Oihw16o: Type = 73;
2506 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2507 pub const dnnl_OIhw4i8o4i: Type = 698;
2508 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2509 pub const dnnl_OhwI4i8o4i: Type = 819;
2510 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2511 pub const dnnl_OIhw4i16o4i: Type = 85;
2512 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2513 pub const dnnl_OhwI4i16o4i: Type = 816;
2514 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2515 pub const dnnl_OIhw4i24o4i: Type = 699;
2516 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2517 pub const dnnl_OhwI4i24o4i: Type = 813;
2518 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2519 pub const dnnl_OIhw4i32o4i: Type = 249;
2520 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2521 pub const dnnl_OhwI4i32o4i: Type = 810;
2522 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2523 pub const dnnl_OIhw4i64o4i: Type = 250;
2524 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2525 pub const dnnl_OhwI4i64o4i: Type = 807;
2526 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2527 pub const dnnl_OIhw16i16o4i: Type = 86;
2528 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2529 pub const dnnl_OIhw16i16o2i: Type = 87;
2530 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2531 pub const dnnl_OIhw16o16i2o: Type = 274;
2532 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2533 pub const dnnl_OIhw4i4o: Type = 88;
2534 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2535 pub const dnnl_OIhw4o4i: Type = 89;
2536 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2537 pub const dnnl_Oihw4o: Type = 82;
2538 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2539 pub const dnnl_OIhw8i8o2i: Type = 708;
2540 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2541 pub const dnnl_OhwI8i8o2i: Type = 804;
2542 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2543 pub const dnnl_OIhw8i16o2i: Type = 104;
2544 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2545 pub const dnnl_OhwI8i16o2i: Type = 801;
2546 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2547 pub const dnnl_OIhw8i32o2i: Type = 251;
2548 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2549 pub const dnnl_OhwI8i32o2i: Type = 795;
2550 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2551 pub const dnnl_OIhw8i24o2i: Type = 704;
2552 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2553 pub const dnnl_OhwI8i24o2i: Type = 798;
2554 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2555 pub const dnnl_OIhw8i64o2i: Type = 252;
2556 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2557 pub const dnnl_OhwI8i64o2i: Type = 792;
2558 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2559 pub const dnnl_OIhw8i8o: Type = 107;
2560 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2561 pub const dnnl_OhwI8i8o: Type = 789;
2562 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2563 pub const dnnl_OIhw8o16i2o: Type = 98;
2564 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2565 pub const dnnl_OIhw2i8o4i: Type = 99;
2566 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2567 pub const dnnl_IOhw8o16i2o: Type = 106;
2568 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2569 pub const dnnl_OIhw8o8i: Type = 100;
2570 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2571 pub const dnnl_OIhw8o4i: Type = 101;
2572 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2573 pub const dnnl_Owhi16o: Type = 221;
2574 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2575 pub const dnnl_OIhw8i32o: Type = 683;
2576 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2577 pub const dnnl_OhwI8i32o: Type = 780;
2578 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2579 pub const dnnl_OIhw8i24o: Type = 687;
2580 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2581 pub const dnnl_OhwI8i24o: Type = 783;
2582 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2583 pub const dnnl_OIhw8i16o: Type = 691;
2584 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2585 pub const dnnl_OhwI8i16o: Type = 786;
2586 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2587 pub const dnnl_Odhwi16o: Type = 217;
2588 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2589 pub const dnnl_OdhwI16o2i: Type = 218;
2590 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2591 pub const dnnl_OdhwI16o4i: Type = 273;
2592 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2593 pub const dnnl_Idhwo8i: Type = 726;
2594 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2595 pub const dnnl_IdhwO8i2o: Type = 727;
2596 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2597 pub const dnnl_IdhwO8i4o: Type = 748;
2598 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2599 pub const dnnl_Idhwo16i: Type = 630;
2600 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2601 pub const dnnl_IdhwO16i2o: Type = 631;
2602 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2603 pub const dnnl_IdhwO16i4o: Type = 632;
2604 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2605 pub const dnnl_Idhwo24i: Type = 738;
2606 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2607 pub const dnnl_IdhwO24i2o: Type = 739;
2608 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2609 pub const dnnl_IdhwO24i4o: Type = 754;
2610 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2611 pub const dnnl_Odhwi4o: Type = 219;
2612 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2613 pub const dnnl_Odhwi8o: Type = 220;
2614 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2615 pub const dnnl_OdhwI8o2i: Type = 677;
2616 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2617 pub const dnnl_OdhwI8o4i: Type = 718;
2618 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2619 pub const dnnl_Odwhi16o: Type = 517;
2620 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2621 pub const dnnl_OIdhw16i16o: Type = 122;
2622 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2623 pub const dnnl_OdhwI16i16o: Type = 778;
2624 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2625 pub const dnnl_OIdhw16i32o: Type = 257;
2626 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2627 pub const dnnl_OdhwI16i32o: Type = 775;
2628 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2629 pub const dnnl_OIdhw16i48o: Type = 761;
2630 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2631 pub const dnnl_OdhwI16i48o: Type = 772;
2632 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2633 pub const dnnl_OIdhw16i64o: Type = 258;
2634 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2635 pub const dnnl_OdhwI16i64o: Type = 769;
2636 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2637 pub const dnnl_OIdhw16o16i: Type = 116;
2638 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2639 pub const dnnl_Oidhw16o: Type = 114;
2640 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2641 pub const dnnl_OIdhw4i4o: Type = 129;
2642 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2643 pub const dnnl_OIdhw4o4i: Type = 130;
2644 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2645 pub const dnnl_Oidhw4o: Type = 126;
2646 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2647 pub const dnnl_OIdhw8i8o2i: Type = 709;
2648 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2649 pub const dnnl_OdhwI8i8o2i: Type = 805;
2650 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2651 pub const dnnl_OIdhw8i16o2i: Type = 143;
2652 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2653 pub const dnnl_OdhwI8i16o2i: Type = 802;
2654 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2655 pub const dnnl_OIdhw8i32o2i: Type = 259;
2656 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2657 pub const dnnl_OdhwI8i32o2i: Type = 796;
2658 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2659 pub const dnnl_OIdhw8i24o2i: Type = 705;
2660 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2661 pub const dnnl_OdhwI8i24o2i: Type = 799;
2662 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2663 pub const dnnl_OIdhw8i64o2i: Type = 260;
2664 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2665 pub const dnnl_OdhwI8i64o2i: Type = 793;
2666 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2667 pub const dnnl_OIdhw8i8o: Type = 147;
2668 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2669 pub const dnnl_OdhwI8i8o: Type = 790;
2670 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2671 pub const dnnl_OIdhw8o16i2o: Type = 111;
2672 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2673 pub const dnnl_IOdhw8o16i2o: Type = 117;
2674 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2675 pub const dnnl_OIdhw4i8o4i: Type = 700;
2676 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2677 pub const dnnl_OdhwI4i8o4i: Type = 820;
2678 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2679 pub const dnnl_OIdhw4i16o4i: Type = 119;
2680 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2681 pub const dnnl_OdhwI4i16o4i: Type = 817;
2682 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2683 pub const dnnl_OIdhw4i24o4i: Type = 701;
2684 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2685 pub const dnnl_OdhwI4i24o4i: Type = 814;
2686 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2687 pub const dnnl_OIdhw4i32o4i: Type = 253;
2688 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2689 pub const dnnl_OdhwI4i32o4i: Type = 811;
2690 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2691 pub const dnnl_OIdhw4i64o4i: Type = 254;
2692 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2693 pub const dnnl_OdhwI4i64o4i: Type = 808;
2694 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2695 pub const dnnl_OIdhw16i16o4i: Type = 255;
2696 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2697 pub const dnnl_OIdhw16i16o2i: Type = 256;
2698 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2699 pub const dnnl_OIdhw2i8o4i: Type = 120;
2700 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2701 pub const dnnl_OIdhw8o8i: Type = 139;
2702 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2703 pub const dnnl_OIdhw8o4i: Type = 140;
2704 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2705 pub const dnnl_IOdhw16i16o: Type = 141;
2706 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2707 pub const dnnl_OIdhw4o8i8o4i: Type = 153;
2708 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2709 pub const dnnl_IOdhw8o8i: Type = 843;
2710 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2711 pub const dnnl_IOdhw16o16i: Type = 229;
2712 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2713 pub const dnnl_OIdhw16o16i2o: Type = 515;
2714 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2715 pub const dnnl_OIdhw8i32o: Type = 684;
2716 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2717 pub const dnnl_OdhwI8i32o: Type = 781;
2718 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2719 pub const dnnl_OIdhw8i24o: Type = 688;
2720 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2721 pub const dnnl_OdhwI8i24o: Type = 784;
2722 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2723 pub const dnnl_OIdhw8i16o: Type = 692;
2724 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2725 pub const dnnl_OdhwI8i16o: Type = 787;
2726 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2727 pub const dnnl_Goiw16g: Type = 73;
2728 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2729 pub const dnnl_Goiw8g: Type = 74;
2730 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2731 pub const dnnl_Goiw4g: Type = 82;
2732 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2733 pub const dnnl_gIOw8o8i: Type = 839;
2734 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2735 pub const dnnl_gIOw16o16i: Type = 207;
2736 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2737 pub const dnnl_gIOw16i16o: Type = 208;
2738 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2739 pub const dnnl_gOIw16i16o: Type = 81;
2740 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2741 pub const dnnl_gOIw16o16i: Type = 80;
2742 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2743 pub const dnnl_gOiw16o: Type = 78;
2744 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2745 pub const dnnl_gOIw4i16o4i: Type = 92;
2746 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2747 pub const dnnl_gOIw2i8o4i: Type = 93;
2748 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2749 pub const dnnl_gOIw16i16o4i: Type = 94;
2750 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2751 pub const dnnl_gOIw16i16o2i: Type = 95;
2752 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2753 pub const dnnl_gOIw16o16i2o: Type = 276;
2754 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2755 pub const dnnl_gOIw4i4o: Type = 96;
2756 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2757 pub const dnnl_gOIw4o4i: Type = 97;
2758 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2759 pub const dnnl_gOiw4o: Type = 84;
2760 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2761 pub const dnnl_gOIw8i16o2i: Type = 110;
2762 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2763 pub const dnnl_gOIw8i8o: Type = 113;
2764 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2765 pub const dnnl_gOIw8o16i2o: Type = 105;
2766 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2767 pub const dnnl_gIOw8o16i2o: Type = 112;
2768 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2769 pub const dnnl_gOIw8o8i: Type = 108;
2770 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2771 pub const dnnl_gOIw8o4i: Type = 109;
2772 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2773 pub const dnnl_gOwi16o: Type = 183;
2774 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2775 pub const dnnl_gOwI16o2i: Type = 184;
2776 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2777 pub const dnnl_gOwI16o4i: Type = 185;
2778 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2779 pub const dnnl_gIwo8i: Type = 728;
2780 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2781 pub const dnnl_gIwO8i2o: Type = 729;
2782 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2783 pub const dnnl_gIwO8i4o: Type = 749;
2784 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2785 pub const dnnl_gIwo16i: Type = 633;
2786 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2787 pub const dnnl_gIwO16i2o: Type = 634;
2788 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2789 pub const dnnl_gIwO16i4o: Type = 635;
2790 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2791 pub const dnnl_gIwo24i: Type = 740;
2792 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2793 pub const dnnl_gIwO24i2o: Type = 741;
2794 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2795 pub const dnnl_gIwO24i4o: Type = 755;
2796 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2797 pub const dnnl_gOwi4o: Type = 186;
2798 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2799 pub const dnnl_gOwi8o: Type = 187;
2800 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2801 pub const dnnl_gOwI8o2i: Type = 678;
2802 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2803 pub const dnnl_gOwI8o4i: Type = 719;
2804 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2805 pub const dnnl_Goiw32g: Type = 76;
2806 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2807 pub const dnnl_gOIw2i4o2i: Type = 90;
2808 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2809 pub const dnnl_gOIw2o4i2o: Type = 118;
2810 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2811 pub const dnnl_gOIw4i8o2i: Type = 103;
2812 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2813 pub const dnnl_gOIw4o8i2o: Type = 91;
2814 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2815 pub const dnnl_goIw4i: Type = 533;
2816 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2817 pub const dnnl_goIw32i: Type = 378;
2818 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2819 pub const dnnl_gIOhw16i16o: Type = 210;
2820 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2821 pub const dnnl_gIOhw8o8i: Type = 840;
2822 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2823 pub const dnnl_gIOhw16o16i: Type = 209;
2824 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2825 pub const dnnl_gOhwi16o: Type = 188;
2826 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2827 pub const dnnl_gOhwI16o2i: Type = 189;
2828 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2829 pub const dnnl_gOhwI16o4i: Type = 190;
2830 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2831 pub const dnnl_gIhwo8i: Type = 730;
2832 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2833 pub const dnnl_gIhwO8i2o: Type = 731;
2834 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2835 pub const dnnl_gIhwO8i4o: Type = 750;
2836 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2837 pub const dnnl_gIhwo16i: Type = 636;
2838 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2839 pub const dnnl_gIhwO16i2o: Type = 637;
2840 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2841 pub const dnnl_gIhwO16i4o: Type = 638;
2842 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2843 pub const dnnl_gIhwo24i: Type = 742;
2844 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2845 pub const dnnl_gIhwO24i2o: Type = 743;
2846 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2847 pub const dnnl_gIhwO24i4o: Type = 756;
2848 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2849 pub const dnnl_gOhwi32o: Type = 191;
2850 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2851 pub const dnnl_gOhwi24o: Type = 664;
2852 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2853 pub const dnnl_gOhwI24o2i: Type = 673;
2854 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2855 pub const dnnl_gOhwI24o4i: Type = 714;
2856 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2857 pub const dnnl_gOhwi4o: Type = 192;
2858 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2859 pub const dnnl_gOhwi8o: Type = 193;
2860 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2861 pub const dnnl_gOhwI8o2i: Type = 679;
2862 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2863 pub const dnnl_gOhwI8o4i: Type = 720;
2864 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2865 pub const dnnl_Goihw16g: Type = 114;
2866 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2867 pub const dnnl_gOIhw16i16o: Type = 124;
2868 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2869 pub const dnnl_gOIhw16o16i: Type = 123;
2870 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2871 pub const dnnl_gOihw16o: Type = 121;
2872 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2873 pub const dnnl_gOIhw2i8o4i: Type = 125;
2874 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2875 pub const dnnl_gOIhw4i16o4i: Type = 134;
2876 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2877 pub const dnnl_gOIhw16i16o4i: Type = 135;
2878 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2879 pub const dnnl_gOIhw16i16o2i: Type = 136;
2880 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2881 pub const dnnl_gOIhw16o16i2o: Type = 277;
2882 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2883 pub const dnnl_gOIhw4i4o: Type = 137;
2884 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2885 pub const dnnl_gOIhw4o4i: Type = 131;
2886 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2887 pub const dnnl_gOihw4o: Type = 128;
2888 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2889 pub const dnnl_Goihw8g: Type = 138;
2890 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2891 pub const dnnl_Goihw4g: Type = 126;
2892 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2893 pub const dnnl_gOIhw8i16o2i: Type = 161;
2894 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2895 pub const dnnl_gOIhw8i8o: Type = 162;
2896 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2897 pub const dnnl_gOIhw8o16i2o: Type = 144;
2898 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2899 pub const dnnl_gIOhw8o16i2o: Type = 146;
2900 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2901 pub const dnnl_gOIhw8o8i: Type = 149;
2902 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2903 pub const dnnl_gOIhw8o4i: Type = 150;
2904 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2905 pub const dnnl_Goihw32g: Type = 115;
2906 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2907 pub const dnnl_gOwhi16o: Type = 201;
2908 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2909 pub const dnnl_goIhw4i: Type = 534;
2910 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2911 pub const dnnl_goIhw32i: Type = 536;
2912 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2913 pub const dnnl_OIw4o8i8o4i: Type = 151;
2914 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2915 pub const dnnl_OIhw4o8i8o4i: Type = 152;
2916 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2917 pub const dnnl_IOw4i8o8i4o: Type = 154;
2918 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2919 pub const dnnl_IOhw4i8o8i4o: Type = 155;
2920 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2921 pub const dnnl_IOdhw4i8o8i4o: Type = 156;
2922 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2923 pub const dnnl_OIhw2o8i8o2i: Type = 157;
2924 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2925 pub const dnnl_gOIw4o8i8o4i: Type = 158;
2926 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2927 pub const dnnl_gOIhw4o8i8o4i: Type = 159;
2928 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2929 pub const dnnl_gOIdhw4o8i8o4i: Type = 179;
2930 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2931 pub const dnnl_gIOw4i8o8i4o: Type = 226;
2932 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2933 pub const dnnl_gIOhw4i8o8i4o: Type = 227;
2934 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2935 pub const dnnl_gIOdhw4i8o8i4o: Type = 228;
2936 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2937 pub const dnnl_gOIhw2o8i8o2i: Type = 160;
2938 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2939 pub const dnnl_gOIhw2i4o2i: Type = 132;
2940 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2941 pub const dnnl_gOIhw2o4i2o: Type = 163;
2942 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2943 pub const dnnl_gOIhw4i8o2i: Type = 145;
2944 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2945 pub const dnnl_gOIhw4o8i2o: Type = 133;
2946 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2947 pub const dnnl_gIOdhw16i16o: Type = 196;
2948 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2949 pub const dnnl_gIOdhw8o8i: Type = 844;
2950 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2951 pub const dnnl_gIOdhw16o16i: Type = 230;
2952 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2953 pub const dnnl_gOdhwi16o: Type = 194;
2954 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2955 pub const dnnl_gOdhwI16o2i: Type = 195;
2956 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2957 pub const dnnl_gOdhwI16o4i: Type = 272;
2958 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2959 pub const dnnl_gIdhwo8i: Type = 732;
2960 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2961 pub const dnnl_gIdhwO8i2o: Type = 733;
2962 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2963 pub const dnnl_gIdhwO8i4o: Type = 751;
2964 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2965 pub const dnnl_gIdhwo16i: Type = 639;
2966 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2967 pub const dnnl_gIdhwO16i2o: Type = 640;
2968 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2969 pub const dnnl_gIdhwO16i4o: Type = 641;
2970 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2971 pub const dnnl_gIdhwo24i: Type = 744;
2972 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2973 pub const dnnl_gIdhwO24i2o: Type = 745;
2974 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2975 pub const dnnl_gIdhwO24i4o: Type = 757;
2976 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2977 pub const dnnl_gOdhwi4o: Type = 197;
2978 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2979 pub const dnnl_gOdhwi8o: Type = 198;
2980 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2981 pub const dnnl_gOdhwI8o2i: Type = 680;
2982 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2983 pub const dnnl_gOdhwI8o4i: Type = 721;
2984 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2985 pub const dnnl_gOdwhi16o: Type = 518;
2986 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2987 pub const dnnl_gOIdhw16i16o: Type = 166;
2988 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2989 pub const dnnl_gOIdhw4i16o4i: Type = 167;
2990 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2991 pub const dnnl_gOIdhw16i16o4i: Type = 261;
2992 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2993 pub const dnnl_gOIdhw2i8o4i: Type = 168;
2994 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2995 pub const dnnl_gOIdhw16i16o2i: Type = 262;
2996 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2997 pub const dnnl_gOIdhw16o16i: Type = 165;
2998 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
2999 pub const dnnl_gOIdhw16o16i2o: Type = 516;
3000 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3001 pub const dnnl_gOidhw16o: Type = 164;
3002 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3003 pub const dnnl_gOIdhw4i4o: Type = 172;
3004 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3005 pub const dnnl_gOIdhw4o4i: Type = 173;
3006 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3007 pub const dnnl_gOidhw4o: Type = 171;
3008 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3009 pub const dnnl_gOIdhw8i16o2i: Type = 178;
3010 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3011 pub const dnnl_gOIdhw8i8o: Type = 182;
3012 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3013 pub const dnnl_gOIdhw8o16i2o: Type = 180;
3014 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3015 pub const dnnl_gIOdhw8o16i2o: Type = 181;
3016 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3017 pub const dnnl_gOIdhw8o8i: Type = 176;
3018 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3019 pub const dnnl_gOIdhw8o4i: Type = 177;
3020 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3021 pub const dnnl_Goidhw16g: Type = 199;
3022 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3023 pub const dnnl_Goidhw32g: Type = 200;
3024 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3025 pub const dnnl_gOIdhw2i4o2i: Type = 174;
3026 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3027 pub const dnnl_gOIdhw4i8o2i: Type = 169;
3028 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3029 pub const dnnl_gOIdhw2o4i2o: Type = 170;
3030 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3031 pub const dnnl_gOIdhw4o8i2o: Type = 175;
3032 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3033 pub const dnnl_goIdhw4i: Type = 535;
3034 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3035 pub const dnnl_goIdhw32i: Type = 537;
3036 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3037 pub const dnnl_Owi24o: Type = 660;
3038 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3039 pub const dnnl_OwI24o2i: Type = 669;
3040 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3041 pub const dnnl_OwI24o4i: Type = 710;
3042 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3043 pub const dnnl_Owi32o: Type = 278;
3044 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3045 pub const dnnl_OwI32o2i: Type = 279;
3046 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3047 pub const dnnl_OwI32o4i: Type = 280;
3048 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3049 pub const dnnl_Owi48o: Type = 281;
3050 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3051 pub const dnnl_OwI48o2i: Type = 282;
3052 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3053 pub const dnnl_OwI48o4i: Type = 283;
3054 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3055 pub const dnnl_Owi64o: Type = 284;
3056 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3057 pub const dnnl_OwI64o2i: Type = 285;
3058 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3059 pub const dnnl_OwI64o4i: Type = 286;
3060 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3061 pub const dnnl_Iwo32i: Type = 570;
3062 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3063 pub const dnnl_IwO32i2o: Type = 571;
3064 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3065 pub const dnnl_IwO32i4o: Type = 572;
3066 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3067 pub const dnnl_Iwo48i: Type = 573;
3068 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3069 pub const dnnl_IwO48i2o: Type = 574;
3070 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3071 pub const dnnl_IwO48i4o: Type = 575;
3072 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3073 pub const dnnl_Iwo64i: Type = 576;
3074 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3075 pub const dnnl_IwO64i2o: Type = 577;
3076 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3077 pub const dnnl_IwO64i4o: Type = 578;
3078 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3079 pub const dnnl_wIo2i: Type = 287;
3080 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3081 pub const dnnl_wIo4i: Type = 288;
3082 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3083 pub const dnnl_gOwi24o: Type = 663;
3084 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3085 pub const dnnl_gOwI24o2i: Type = 672;
3086 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3087 pub const dnnl_gOwI24o4i: Type = 713;
3088 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3089 pub const dnnl_gOwi32o: Type = 289;
3090 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3091 pub const dnnl_gOwI32o2i: Type = 290;
3092 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3093 pub const dnnl_gOwI32o4i: Type = 291;
3094 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3095 pub const dnnl_gOwi48o: Type = 292;
3096 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3097 pub const dnnl_gOwI48o2i: Type = 293;
3098 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3099 pub const dnnl_gOwI48o4i: Type = 294;
3100 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3101 pub const dnnl_gOwi64o: Type = 295;
3102 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3103 pub const dnnl_gOwI64o2i: Type = 296;
3104 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3105 pub const dnnl_gOwI64o4i: Type = 297;
3106 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3107 pub const dnnl_gIwo32i: Type = 579;
3108 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3109 pub const dnnl_gIwO32i2o: Type = 580;
3110 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3111 pub const dnnl_gIwO32i4o: Type = 581;
3112 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3113 pub const dnnl_gIwo48i: Type = 582;
3114 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3115 pub const dnnl_gIwO48i2o: Type = 583;
3116 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3117 pub const dnnl_gIwO48i4o: Type = 584;
3118 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3119 pub const dnnl_gIwo64i: Type = 585;
3120 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3121 pub const dnnl_gIwO64i2o: Type = 586;
3122 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3123 pub const dnnl_gIwO64i4o: Type = 587;
3124 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3125 pub const dnnl_gwio: Type = 24;
3126 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3127 pub const dnnl_gwIo2i: Type = 298;
3128 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3129 pub const dnnl_gwIo4i: Type = 299;
3130 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3131 pub const dnnl_OhwI24o: Type = 661;
3132 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3133 pub const dnnl_OhwI24o2i: Type = 670;
3134 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3135 pub const dnnl_OhwI24o4i: Type = 711;
3136 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3137 pub const dnnl_OhwI32o: Type = 214;
3138 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3139 pub const dnnl_OhwI32o2i: Type = 300;
3140 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3141 pub const dnnl_OhwI32o4i: Type = 301;
3142 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3143 pub const dnnl_Ohwi48o: Type = 302;
3144 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3145 pub const dnnl_OhwI48o2i: Type = 303;
3146 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3147 pub const dnnl_OhwI48o4i: Type = 304;
3148 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3149 pub const dnnl_Ohwi64o: Type = 305;
3150 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3151 pub const dnnl_OhwI64o2i: Type = 306;
3152 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3153 pub const dnnl_OhwI64o4i: Type = 307;
3154 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3155 pub const dnnl_Ihwo32i: Type = 642;
3156 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3157 pub const dnnl_IhwO32i2o: Type = 643;
3158 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3159 pub const dnnl_IhwO32i4o: Type = 644;
3160 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3161 pub const dnnl_Ihwo48i: Type = 645;
3162 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3163 pub const dnnl_IhwO48i2o: Type = 646;
3164 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3165 pub const dnnl_IhwO48i4o: Type = 647;
3166 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3167 pub const dnnl_Ihwo64i: Type = 648;
3168 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3169 pub const dnnl_IhwO64i2o: Type = 649;
3170 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3171 pub const dnnl_IhwO64i4o: Type = 650;
3172 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3173 pub const dnnl_hwIo2i: Type = 308;
3174 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3175 pub const dnnl_hwIo4i: Type = 309;
3176 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3177 pub const dnnl_gOhwI24o: Type = 664;
3178 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3179 pub const dnnl_gOhwI32o: Type = 191;
3180 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3181 pub const dnnl_gOhwI32o2i: Type = 310;
3182 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3183 pub const dnnl_gOhwI32o4i: Type = 311;
3184 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3185 pub const dnnl_gOhwi48o: Type = 312;
3186 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3187 pub const dnnl_gOhwI48o2i: Type = 313;
3188 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3189 pub const dnnl_gOhwI48o4i: Type = 314;
3190 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3191 pub const dnnl_gOhwi64o: Type = 315;
3192 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3193 pub const dnnl_gOhwI64o2i: Type = 316;
3194 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3195 pub const dnnl_gOhwI64o4i: Type = 317;
3196 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3197 pub const dnnl_gIhwo32i: Type = 651;
3198 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3199 pub const dnnl_gIhwO32i2o: Type = 652;
3200 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3201 pub const dnnl_gIhwO32i4o: Type = 653;
3202 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3203 pub const dnnl_gIhwo48i: Type = 654;
3204 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3205 pub const dnnl_gIhwO48i2o: Type = 655;
3206 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3207 pub const dnnl_gIhwO48i4o: Type = 656;
3208 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3209 pub const dnnl_gIhwo64i: Type = 657;
3210 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3211 pub const dnnl_gIhwO64i2o: Type = 658;
3212 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3213 pub const dnnl_gIhwO64i4o: Type = 659;
3214 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3215 pub const dnnl_ghwio: Type = 34;
3216 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3217 pub const dnnl_ghwIo2i: Type = 318;
3218 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3219 pub const dnnl_ghwIo4i: Type = 319;
3220 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3221 pub const dnnl_Odhwi24o: Type = 662;
3222 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3223 pub const dnnl_OdhwI24o2i: Type = 671;
3224 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3225 pub const dnnl_OdhwI24o4i: Type = 712;
3226 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3227 pub const dnnl_Odhwi32o: Type = 320;
3228 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3229 pub const dnnl_OdhwI32o2i: Type = 321;
3230 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3231 pub const dnnl_OdhwI32o4i: Type = 322;
3232 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3233 pub const dnnl_Odhwi48o: Type = 323;
3234 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3235 pub const dnnl_OdhwI48o2i: Type = 324;
3236 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3237 pub const dnnl_OdhwI48o4i: Type = 325;
3238 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3239 pub const dnnl_Odhwi64o: Type = 326;
3240 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3241 pub const dnnl_OdhwI64o2i: Type = 327;
3242 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3243 pub const dnnl_OdhwI64o4i: Type = 328;
3244 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3245 pub const dnnl_Idhwo32i: Type = 561;
3246 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3247 pub const dnnl_IdhwO32i2o: Type = 562;
3248 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3249 pub const dnnl_IdhwO32i4o: Type = 563;
3250 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3251 pub const dnnl_Idhwo48i: Type = 564;
3252 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3253 pub const dnnl_IdhwO48i2o: Type = 565;
3254 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3255 pub const dnnl_IdhwO48i4o: Type = 566;
3256 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3257 pub const dnnl_Idhwo64i: Type = 567;
3258 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3259 pub const dnnl_IdhwO64i2o: Type = 568;
3260 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3261 pub const dnnl_IdhwO64i4o: Type = 569;
3262 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3263 pub const dnnl_dhwIo2i: Type = 329;
3264 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3265 pub const dnnl_dhwIo4i: Type = 330;
3266 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3267 pub const dnnl_gOdhwi24o: Type = 665;
3268 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3269 pub const dnnl_gOdhwI24o2i: Type = 674;
3270 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3271 pub const dnnl_gOdhwI24o4i: Type = 715;
3272 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3273 pub const dnnl_gOdhwi32o: Type = 331;
3274 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3275 pub const dnnl_gOdhwI32o2i: Type = 332;
3276 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3277 pub const dnnl_gOdhwI32o4i: Type = 333;
3278 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3279 pub const dnnl_gOdhwi48o: Type = 334;
3280 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3281 pub const dnnl_gOdhwI48o2i: Type = 335;
3282 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3283 pub const dnnl_gOdhwI48o4i: Type = 336;
3284 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3285 pub const dnnl_gOdhwi64o: Type = 337;
3286 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3287 pub const dnnl_gOdhwI64o2i: Type = 338;
3288 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3289 pub const dnnl_gOdhwI64o4i: Type = 339;
3290 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3291 pub const dnnl_gIdhwo32i: Type = 552;
3292 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3293 pub const dnnl_gIdhwO32i2o: Type = 553;
3294 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3295 pub const dnnl_gIdhwO32i4o: Type = 554;
3296 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3297 pub const dnnl_gIdhwo48i: Type = 555;
3298 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3299 pub const dnnl_gIdhwO48i2o: Type = 556;
3300 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3301 pub const dnnl_gIdhwO48i4o: Type = 557;
3302 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3303 pub const dnnl_gIdhwo64i: Type = 558;
3304 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3305 pub const dnnl_gIdhwO64i2o: Type = 559;
3306 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3307 pub const dnnl_gIdhwO64i4o: Type = 560;
3308 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3309 pub const dnnl_gdhwio: Type = 44;
3310 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3311 pub const dnnl_gdhwIo2i: Type = 340;
3312 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3313 pub const dnnl_gdhwIo4i: Type = 341;
3314 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3315 pub const dnnl_OI16i32o4i: Type = 342;
3316 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3317 pub const dnnl_OI16i48o4i: Type = 343;
3318 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3319 pub const dnnl_OI16i64o4i: Type = 344;
3320 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3321 pub const dnnl_OI16i16o2i: Type = 345;
3322 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3323 pub const dnnl_OI16i32o2i: Type = 346;
3324 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3325 pub const dnnl_OI16i48o2i: Type = 347;
3326 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3327 pub const dnnl_OI16i64o2i: Type = 348;
3328 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3329 pub const dnnl_OIw16i32o4i: Type = 349;
3330 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3331 pub const dnnl_OIw16i48o4i: Type = 350;
3332 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3333 pub const dnnl_OIw16i64o4i: Type = 351;
3334 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3335 pub const dnnl_OIw16i32o2i: Type = 352;
3336 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3337 pub const dnnl_OIw16i48o2i: Type = 353;
3338 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3339 pub const dnnl_OIw16i64o2i: Type = 354;
3340 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3341 pub const dnnl_OIhw16i32o4i: Type = 355;
3342 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3343 pub const dnnl_OIhw16i48o4i: Type = 356;
3344 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3345 pub const dnnl_OIhw16i64o4i: Type = 357;
3346 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3347 pub const dnnl_OIhw16i32o2i: Type = 358;
3348 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3349 pub const dnnl_OIhw16i48o2i: Type = 359;
3350 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3351 pub const dnnl_OIhw16i64o2i: Type = 360;
3352 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3353 pub const dnnl_OIdhw16i32o4i: Type = 361;
3354 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3355 pub const dnnl_OIdhw16i48o4i: Type = 362;
3356 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3357 pub const dnnl_OIdhw16i64o4i: Type = 363;
3358 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3359 pub const dnnl_OIdhw16i32o2i: Type = 364;
3360 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3361 pub const dnnl_OIdhw16i48o2i: Type = 365;
3362 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3363 pub const dnnl_OIdhw16i64o2i: Type = 366;
3364 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3365 pub const dnnl_OwI16i16o2i: Type = 393;
3366 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3367 pub const dnnl_OwI16i16o4i: Type = 394;
3368 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3369 pub const dnnl_OhwI16i16o2i: Type = 395;
3370 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3371 pub const dnnl_OhwI16i16o4i: Type = 396;
3372 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3373 pub const dnnl_OdhwI16i16o2i: Type = 397;
3374 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3375 pub const dnnl_OdhwI16i16o4i: Type = 399;
3376 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3377 pub const dnnl_IwO16o16i2o: Type = 588;
3378 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3379 pub const dnnl_IwO16o16i4o: Type = 589;
3380 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3381 pub const dnnl_IhwO16o16i2o: Type = 590;
3382 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3383 pub const dnnl_IhwO16o16i4o: Type = 591;
3384 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3385 pub const dnnl_IdhwO16o16i2o: Type = 592;
3386 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3387 pub const dnnl_IdhwO16o16i4o: Type = 593;
3388 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3389 pub const dnnl_gOwI16i16o2i: Type = 406;
3390 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3391 pub const dnnl_gOwI16i16o4i: Type = 407;
3392 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3393 pub const dnnl_gOhwI16i16o2i: Type = 390;
3394 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3395 pub const dnnl_gOhwI16i16o4i: Type = 391;
3396 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3397 pub const dnnl_gOdhwI16i16o2i: Type = 392;
3398 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3399 pub const dnnl_gOdhwI16i16o4i: Type = 398;
3400 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3401 pub const dnnl_gIwO16o16i2o: Type = 594;
3402 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3403 pub const dnnl_gIwO16o16i4o: Type = 595;
3404 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3405 pub const dnnl_gIhwO16o16i2o: Type = 596;
3406 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3407 pub const dnnl_gIhwO16o16i4o: Type = 597;
3408 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3409 pub const dnnl_gIdhwO16o16i2o: Type = 598;
3410 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3411 pub const dnnl_gIdhwO16o16i4o: Type = 599;
3412 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3413 pub const dnnl_OwI16i32o2i: Type = 400;
3414 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3415 pub const dnnl_OwI16i32o4i: Type = 401;
3416 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3417 pub const dnnl_OwI16i48o2i: Type = 402;
3418 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3419 pub const dnnl_OwI16i48o4i: Type = 403;
3420 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3421 pub const dnnl_OwI16i64o2i: Type = 404;
3422 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3423 pub const dnnl_OwI16i64o4i: Type = 405;
3424 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3425 pub const dnnl_IwO16o32i2o: Type = 600;
3426 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3427 pub const dnnl_IwO16o32i4o: Type = 601;
3428 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3429 pub const dnnl_IwO16o48i2o: Type = 602;
3430 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3431 pub const dnnl_IwO16o48i4o: Type = 603;
3432 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3433 pub const dnnl_IwO16o64i2o: Type = 604;
3434 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3435 pub const dnnl_IwO16o64i4o: Type = 605;
3436 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3437 pub const dnnl_gOwI16i32o2i: Type = 408;
3438 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3439 pub const dnnl_gOwI16i32o4i: Type = 409;
3440 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3441 pub const dnnl_gOwI16i48o2i: Type = 410;
3442 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3443 pub const dnnl_gOwI16i48o4i: Type = 411;
3444 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3445 pub const dnnl_gOwI16i64o2i: Type = 412;
3446 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3447 pub const dnnl_gOwI16i64o4i: Type = 413;
3448 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3449 pub const dnnl_gIwO16o32i2o: Type = 606;
3450 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3451 pub const dnnl_gIwO16o32i4o: Type = 607;
3452 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3453 pub const dnnl_gIwO16o48i2o: Type = 608;
3454 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3455 pub const dnnl_gIwO16o48i4o: Type = 609;
3456 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3457 pub const dnnl_gIwO16o64i2o: Type = 610;
3458 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3459 pub const dnnl_gIwO16o64i4o: Type = 611;
3460 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3461 pub const dnnl_OhwI16i32o2i: Type = 414;
3462 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3463 pub const dnnl_OhwI16i32o4i: Type = 415;
3464 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3465 pub const dnnl_OhwI16i48o2i: Type = 416;
3466 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3467 pub const dnnl_OhwI16i48o4i: Type = 417;
3468 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3469 pub const dnnl_OhwI16i64o2i: Type = 418;
3470 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3471 pub const dnnl_OhwI16i64o4i: Type = 419;
3472 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3473 pub const dnnl_IhwO16o32i2o: Type = 612;
3474 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3475 pub const dnnl_IhwO16o32i4o: Type = 613;
3476 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3477 pub const dnnl_IhwO16o48i2o: Type = 614;
3478 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3479 pub const dnnl_IhwO16o48i4o: Type = 615;
3480 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3481 pub const dnnl_IhwO16o64i2o: Type = 616;
3482 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3483 pub const dnnl_IhwO16o64i4o: Type = 617;
3484 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3485 pub const dnnl_gOhwI16i32o2i: Type = 420;
3486 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3487 pub const dnnl_gOhwI16i32o4i: Type = 421;
3488 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3489 pub const dnnl_gOhwI16i48o2i: Type = 422;
3490 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3491 pub const dnnl_gOhwI16i48o4i: Type = 423;
3492 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3493 pub const dnnl_gOhwI16i64o2i: Type = 424;
3494 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3495 pub const dnnl_gOhwI16i64o4i: Type = 425;
3496 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3497 pub const dnnl_gIhwO16o32i2o: Type = 618;
3498 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3499 pub const dnnl_gIhwO16o32i4o: Type = 619;
3500 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3501 pub const dnnl_gIhwO16o48i2o: Type = 620;
3502 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3503 pub const dnnl_gIhwO16o48i4o: Type = 621;
3504 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3505 pub const dnnl_gIhwO16o64i2o: Type = 622;
3506 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3507 pub const dnnl_gIhwO16o64i4o: Type = 623;
3508 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3509 pub const dnnl_OdhwI16i32o2i: Type = 426;
3510 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3511 pub const dnnl_OdhwI16i32o4i: Type = 427;
3512 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3513 pub const dnnl_OdhwI16i48o2i: Type = 428;
3514 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3515 pub const dnnl_OdhwI16i48o4i: Type = 429;
3516 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3517 pub const dnnl_OdhwI16i64o2i: Type = 430;
3518 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3519 pub const dnnl_OdhwI16i64o4i: Type = 431;
3520 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3521 pub const dnnl_IdhwO16o32i2o: Type = 546;
3522 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3523 pub const dnnl_IdhwO16o32i4o: Type = 547;
3524 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3525 pub const dnnl_IdhwO16o48i2o: Type = 548;
3526 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3527 pub const dnnl_IdhwO16o48i4o: Type = 549;
3528 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3529 pub const dnnl_IdhwO16o64i2o: Type = 550;
3530 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3531 pub const dnnl_IdhwO16o64i4o: Type = 551;
3532 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3533 pub const dnnl_gOdhwI16i32o2i: Type = 432;
3534 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3535 pub const dnnl_gOdhwI16i32o4i: Type = 433;
3536 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3537 pub const dnnl_gOdhwI16i48o2i: Type = 434;
3538 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3539 pub const dnnl_gOdhwI16i48o4i: Type = 435;
3540 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3541 pub const dnnl_gOdhwI16i64o2i: Type = 436;
3542 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3543 pub const dnnl_gOdhwI16i64o4i: Type = 437;
3544 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3545 pub const dnnl_gIdhwO16o32i2o: Type = 540;
3546 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3547 pub const dnnl_gIdhwO16o32i4o: Type = 541;
3548 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3549 pub const dnnl_gIdhwO16o48i2o: Type = 542;
3550 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3551 pub const dnnl_gIdhwO16o48i4o: Type = 543;
3552 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3553 pub const dnnl_gIdhwO16o64i2o: Type = 544;
3554 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3555 pub const dnnl_gIdhwO16o64i4o: Type = 545;
3556 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3557 pub const dnnl_hwioG16g: Type = 438;
3558 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3559 pub const dnnl_hwioG8g: Type = 539;
3560 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3561 pub const dnnl_dhwioG16g: Type = 765;
3562 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3563 pub const dnnl_dhwioG8g: Type = 766;
3564 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3565 pub const dnnl_NCdhw40n16c: Type = 444;
3566 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3567 pub const dnnl_NCw40n16c: Type = 372;
3568 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3569 pub const dnnl_NChw40n16c: Type = 376;
3570 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3571 pub const dnnl_NCw40n32c: Type = 373;
3572 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3573 pub const dnnl_NChw40n32c: Type = 377;
3574 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3575 pub const dnnl_NCdhw40n32c: Type = 445;
3576 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3577 pub const dnnl_OIdhw4o8i8o2i: Type = 447;
3578 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3579 pub const dnnl_OIhw4o8i8o2i: Type = 448;
3580 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3581 pub const dnnl_OIw4o8i8o2i: Type = 449;
3582 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3583 pub const dnnl_gOIdhw4o8i8o2i: Type = 450;
3584 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3585 pub const dnnl_gOIhw4o8i8o2i: Type = 451;
3586 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3587 pub const dnnl_gOIw4o8i8o2i: Type = 452;
3588 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3589 pub const dnnl_IOdhw4i8o8i2o: Type = 453;
3590 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3591 pub const dnnl_IOhw4i8o8i2o: Type = 454;
3592 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3593 pub const dnnl_IOw4i8o8i2o: Type = 455;
3594 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3595 pub const dnnl_gIOdhw4i8o8i2o: Type = 456;
3596 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3597 pub const dnnl_gIOhw4i8o8i2o: Type = 457;
3598 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3599 pub const dnnl_gIOw4i8o8i2o: Type = 458;
3600 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3601 pub const dnnl_NCw2c32n8c: Type = 474;
3602 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3603 pub const dnnl_NChw2c32n8c: Type = 478;
3604 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3605 pub const dnnl_NCdhw2c32n8c: Type = 483;
3606 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3607 pub const dnnl_OIw2i8o16i4o: Type = 469;
3608 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3609 pub const dnnl_OIhw2i8o16i4o: Type = 470;
3610 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3611 pub const dnnl_OIdhw2i8o16i4o: Type = 471;
3612 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3613 pub const dnnl_OIw2o8i16o4i: Type = 472;
3614 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3615 pub const dnnl_OIw2o8i16o2i: Type = 473;
3616 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3617 pub const dnnl_IOw2i8o16i4o: Type = 490;
3618 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3619 pub const dnnl_IOw2i8o16i2o: Type = 493;
3620 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3621 pub const dnnl_OIhw2o8i16o4i: Type = 475;
3622 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3623 pub const dnnl_OIhw2o8i16o2i: Type = 476;
3624 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3625 pub const dnnl_IOhw2i8o16i4o: Type = 489;
3626 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3627 pub const dnnl_IOhw2i8o16i2o: Type = 492;
3628 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3629 pub const dnnl_OIdhw2o8i16o4i: Type = 480;
3630 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3631 pub const dnnl_OIdhw2o8i16o2i: Type = 481;
3632 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3633 pub const dnnl_IOdhw2i8o16i4o: Type = 488;
3634 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3635 pub const dnnl_IOdhw2i8o16i2o: Type = 491;
3636 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3637 pub const dnnl_gOIw2o8i16o2i: Type = 485;
3638 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3639 pub const dnnl_gIOw2i8o16i2o: Type = 477;
3640 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3641 pub const dnnl_gIOhw2i8o16i2o: Type = 494;
3642 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3643 pub const dnnl_gIOdhw2i8o16i2o: Type = 495;
3644 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3645 pub const dnnl_gOIhw2o8i16o2i: Type = 486;
3646 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3647 pub const dnnl_gOIdhw2o8i16o2i: Type = 487;
3648 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3649 pub const dnnl_gOIw2o8i16o4i: Type = 497;
3650 #[doc = " 3D CNN activations tensor blocked by channels with block size 8,\n an alias to #dnnl_aBc8b"]
3651 pub const dnnl_gOIhw2o8i16o4i: Type = 498;
3652}
3653pub mod dnnl_prop_kind_t {
3654 #[doc = " Kinds of propagation."]
3655 pub type Type = ::std::os::raw::c_uint;
3656 #[doc = " Undefined propagation type."]
3657 pub const dnnl_prop_kind_undef: Type = 0;
3658 #[doc = " Forward data propagation (training mode). In this mode primitives\n perform computations necessary for subsequent backward propagation."]
3659 pub const dnnl_forward_training: Type = 64;
3660 #[doc = " Forward data propagation (inference mode). In this mode primitives\n perform only computations that are necessary for inference and omit\n computations that are necessary only for backward propagation."]
3661 pub const dnnl_forward_inference: Type = 96;
3662 #[doc = " Forward data propagation (alias for @c dnnl_forward_training)."]
3663 pub const dnnl_forward: Type = 64;
3664 #[doc = " Backward propagation (with respect to all parameters)."]
3665 pub const dnnl_backward: Type = 128;
3666 #[doc = " Backward data propagation."]
3667 pub const dnnl_backward_data: Type = 160;
3668 #[doc = " Backward weights propagation."]
3669 pub const dnnl_backward_weights: Type = 192;
3670 #[doc = " Backward bias propagation."]
3671 pub const dnnl_backward_bias: Type = 193;
3672}
3673pub mod dnnl_primitive_kind_t {
3674 #[doc = " Kinds of primitives. Used to implement a way to extend the library with new\n primitives without changing the ABI."]
3675 pub type Type = ::std::os::raw::c_uint;
3676 #[doc = " Undefined primitive"]
3677 pub const dnnl_undefined_primitive: Type = 0;
3678 #[doc = " A reorder primitive."]
3679 pub const dnnl_reorder: Type = 1;
3680 #[doc = " A shuffle primitive."]
3681 pub const dnnl_shuffle: Type = 2;
3682 #[doc = " A (out-of-place) concat primitive."]
3683 pub const dnnl_concat: Type = 3;
3684 #[doc = " A sum primitive."]
3685 pub const dnnl_sum: Type = 4;
3686 #[doc = " A convolution primitive."]
3687 pub const dnnl_convolution: Type = 5;
3688 #[doc = " A deconvolution primitive."]
3689 pub const dnnl_deconvolution: Type = 6;
3690 #[doc = " An element-wise primitive."]
3691 pub const dnnl_eltwise: Type = 7;
3692 #[doc = " An LRN primitive."]
3693 pub const dnnl_lrn: Type = 8;
3694 #[doc = " A batch normalization primitive."]
3695 pub const dnnl_batch_normalization: Type = 9;
3696 #[doc = " An inner product primitive."]
3697 pub const dnnl_inner_product: Type = 10;
3698 #[doc = " A rnn primitive."]
3699 pub const dnnl_rnn: Type = 11;
3700 #[doc = " A matrix multiplication primitive (internal)."]
3701 pub const dnnl_gemm: Type = 12;
3702 #[doc = " A binary primitive."]
3703 pub const dnnl_binary: Type = 13;
3704 #[doc = " A matrix multiplication primitive."]
3705 pub const dnnl_matmul: Type = 14;
3706 #[doc = " A resampling primitive."]
3707 pub const dnnl_resampling: Type = 15;
3708 #[doc = " A pooling primitive."]
3709 pub const dnnl_pooling: Type = 16;
3710 #[doc = " A reduction primitive."]
3711 pub const dnnl_reduction: Type = 17;
3712 #[doc = " A PReLU primitive."]
3713 pub const dnnl_prelu: Type = 18;
3714 #[doc = " A softmax primitive."]
3715 pub const dnnl_softmax: Type = 19;
3716 #[doc = " A layer normalization primitive."]
3717 pub const dnnl_layer_normalization: Type = 20;
3718 #[doc = " A group normalization primitive."]
3719 pub const dnnl_group_normalization: Type = 21;
3720 #[doc = " Parameter to allow internal only primitives without undefined behavior.\n This parameter is chosen to be valid for so long as sizeof(int) >= 2."]
3721 pub const dnnl_primitive_kind_max: Type = 32767;
3722}
3723pub mod dnnl_alg_kind_t {
3724 #[doc = " Kinds of algorithms."]
3725 pub type Type = ::std::os::raw::c_uint;
3726 pub const dnnl_alg_kind_undef: Type = 0;
3727 #[doc = " Direct convolution"]
3728 pub const dnnl_convolution_direct: Type = 1;
3729 #[doc = " Winograd convolution"]
3730 pub const dnnl_convolution_winograd: Type = 2;
3731 #[doc = " Convolution algorithm(either direct or Winograd) is chosen just in time"]
3732 pub const dnnl_convolution_auto: Type = 3;
3733 #[doc = " Direct deconvolution"]
3734 pub const dnnl_deconvolution_direct: Type = 10;
3735 #[doc = " Winograd deconvolution"]
3736 pub const dnnl_deconvolution_winograd: Type = 11;
3737 #[doc = " Eltwise: ReLU"]
3738 pub const dnnl_eltwise_relu: Type = 32;
3739 #[doc = " Eltwise: hyperbolic tangent non-linearity (tanh)"]
3740 pub const dnnl_eltwise_tanh: Type = 33;
3741 #[doc = " Eltwise: exponential linear unit (elu)"]
3742 pub const dnnl_eltwise_elu: Type = 34;
3743 #[doc = " Eltwise: square"]
3744 pub const dnnl_eltwise_square: Type = 35;
3745 #[doc = " Eltwise: abs"]
3746 pub const dnnl_eltwise_abs: Type = 36;
3747 #[doc = " Eltwise: square root"]
3748 pub const dnnl_eltwise_sqrt: Type = 37;
3749 #[doc = " Eltwise: linear"]
3750 pub const dnnl_eltwise_linear: Type = 38;
3751 #[doc = " Eltwise: soft_relu"]
3752 pub const dnnl_eltwise_soft_relu: Type = 39;
3753 #[doc = " Eltwise: hardsigmoid"]
3754 pub const dnnl_eltwise_hardsigmoid: Type = 40;
3755 #[doc = " Eltwise: logistic"]
3756 pub const dnnl_eltwise_logistic: Type = 41;
3757 #[doc = " Eltwise: exponent"]
3758 pub const dnnl_eltwise_exp: Type = 42;
3759 #[doc = " Eltwise: gelu\n\n @note Tanh approximation formula is used to approximate\n the cumulative distribution function of a Gaussian here"]
3760 pub const dnnl_eltwise_gelu_tanh: Type = 43;
3761 #[doc = " Eltwise: swish"]
3762 pub const dnnl_eltwise_swish: Type = 44;
3763 #[doc = " Eltwise: natural logarithm"]
3764 pub const dnnl_eltwise_log: Type = 45;
3765 #[doc = " Eltwise: clip"]
3766 pub const dnnl_eltwise_clip: Type = 46;
3767 #[doc = " Eltwise: clip version 2"]
3768 pub const dnnl_eltwise_clip_v2: Type = 47;
3769 #[doc = " Eltwise: pow"]
3770 pub const dnnl_eltwise_pow: Type = 48;
3771 #[doc = " Eltwise: erf-based gelu"]
3772 pub const dnnl_eltwise_gelu_erf: Type = 49;
3773 #[doc = " Eltwise: round"]
3774 pub const dnnl_eltwise_round: Type = 50;
3775 #[doc = " Eltwise: mish"]
3776 pub const dnnl_eltwise_mish: Type = 51;
3777 #[doc = " Eltwise: hardswish"]
3778 pub const dnnl_eltwise_hardswish: Type = 52;
3779 #[doc = " Eltwise: ReLU (dst for backward)"]
3780 pub const dnnl_eltwise_relu_use_dst_for_bwd: Type = 256;
3781 #[doc = " Eltwise: hyperbolic tangent non-linearity (tanh) (dst for backward)"]
3782 pub const dnnl_eltwise_tanh_use_dst_for_bwd: Type = 257;
3783 #[doc = " Eltwise: exponential linear unit (elu) (dst for backward)"]
3784 pub const dnnl_eltwise_elu_use_dst_for_bwd: Type = 258;
3785 #[doc = " Eltwise: square root (dst for backward)"]
3786 pub const dnnl_eltwise_sqrt_use_dst_for_bwd: Type = 259;
3787 #[doc = " Eltwise: logistic (dst for backward)"]
3788 pub const dnnl_eltwise_logistic_use_dst_for_bwd: Type = 260;
3789 #[doc = " Eltwise: exp (dst for backward)"]
3790 pub const dnnl_eltwise_exp_use_dst_for_bwd: Type = 261;
3791 #[doc = " Eltwise: clip version 2 (dst for backward)"]
3792 pub const dnnl_eltwise_clip_v2_use_dst_for_bwd: Type = 262;
3793 #[doc = " Max pooling"]
3794 pub const dnnl_pooling_max: Type = 511;
3795 #[doc = " Average pooling include padding"]
3796 pub const dnnl_pooling_avg_include_padding: Type = 767;
3797 #[doc = " Average pooling exclude padding"]
3798 pub const dnnl_pooling_avg_exclude_padding: Type = 1023;
3799 #[doc = " Local response normalization (LRN) across multiple channels"]
3800 pub const dnnl_lrn_across_channels: Type = 2815;
3801 #[doc = " LRN within a single channel"]
3802 pub const dnnl_lrn_within_channel: Type = 3071;
3803 #[doc = " RNN cell"]
3804 pub const dnnl_vanilla_rnn: Type = 8191;
3805 #[doc = " LSTM cell"]
3806 pub const dnnl_vanilla_lstm: Type = 12287;
3807 #[doc = " GRU cell"]
3808 pub const dnnl_vanilla_gru: Type = 16383;
3809 #[doc = " GRU cell with linear before reset\n\n Modification of original GRU cell. Differs from #dnnl_vanilla_gru\n in how the new memory gate is calculated:\n \\f[ c_t = tanh(W_c*x_t + b_{c_x} + r_t*(U_c*h_{t-1}+b_{c_h})) \\f]\n Primitive expects 4 biases on input:\n \\f$[b_{u}, b_{r}, b_{c_x}, b_{c_h}]\\f$"]
3810 pub const dnnl_lbr_gru: Type = 20479;
3811 #[doc = " AUGRU cell"]
3812 pub const dnnl_vanilla_augru: Type = 24575;
3813 #[doc = " AUGRU cell with linear before reset"]
3814 pub const dnnl_lbr_augru: Type = 28671;
3815 #[doc = " Binary add"]
3816 pub const dnnl_binary_add: Type = 131056;
3817 #[doc = " Binary mul"]
3818 pub const dnnl_binary_mul: Type = 131057;
3819 #[doc = " Binary max"]
3820 pub const dnnl_binary_max: Type = 131058;
3821 #[doc = " Binary min"]
3822 pub const dnnl_binary_min: Type = 131059;
3823 #[doc = " Binary div"]
3824 pub const dnnl_binary_div: Type = 131060;
3825 #[doc = " Binary sub"]
3826 pub const dnnl_binary_sub: Type = 131061;
3827 #[doc = " Binary greater or equal"]
3828 pub const dnnl_binary_ge: Type = 131062;
3829 #[doc = " Binary greater than"]
3830 pub const dnnl_binary_gt: Type = 131063;
3831 #[doc = " Binary less or equal"]
3832 pub const dnnl_binary_le: Type = 131064;
3833 #[doc = " Binary less than"]
3834 pub const dnnl_binary_lt: Type = 131065;
3835 #[doc = " Binary equal"]
3836 pub const dnnl_binary_eq: Type = 131066;
3837 #[doc = " Binary not equal"]
3838 pub const dnnl_binary_ne: Type = 131067;
3839 #[doc = " Binary select"]
3840 pub const dnnl_binary_select: Type = 131068;
3841 #[doc = " Nearest Neighbor Resampling Method"]
3842 pub const dnnl_resampling_nearest: Type = 196592;
3843 #[doc = " Linear Resampling Method"]
3844 pub const dnnl_resampling_linear: Type = 196593;
3845 #[doc = " Reduction using max"]
3846 pub const dnnl_reduction_max: Type = 196594;
3847 #[doc = " Reduction using min"]
3848 pub const dnnl_reduction_min: Type = 196595;
3849 #[doc = " Reduction using sum"]
3850 pub const dnnl_reduction_sum: Type = 196596;
3851 #[doc = " Reduction using mul"]
3852 pub const dnnl_reduction_mul: Type = 196597;
3853 #[doc = " Reduction using mean"]
3854 pub const dnnl_reduction_mean: Type = 196598;
3855 #[doc = " Reduction using lp norm"]
3856 pub const dnnl_reduction_norm_lp_max: Type = 196599;
3857 #[doc = " Reduction using lp norm"]
3858 pub const dnnl_reduction_norm_lp_sum: Type = 196600;
3859 #[doc = " Reduction using lp norm without final pth-root"]
3860 pub const dnnl_reduction_norm_lp_power_p_max: Type = 196601;
3861 #[doc = " Reduction using lp norm without final pth-root"]
3862 pub const dnnl_reduction_norm_lp_power_p_sum: Type = 196602;
3863 #[doc = " Softmax"]
3864 pub const dnnl_softmax_accurate: Type = 196608;
3865 #[doc = " Logsoftmax"]
3866 pub const dnnl_softmax_log: Type = 196609;
3867}
3868pub mod dnnl_normalization_flags_t {
3869 #[doc = " Flags for normalization primitives."]
3870 pub type Type = ::std::os::raw::c_uint;
3871 #[doc = " Use no normalization flags\n\n If specified\n - on forward training propagation mean and variance are computed and\n stored as output\n - on backward propagation compute full derivative wrt data\n - on backward propagation prop_kind == #dnnl_backward_data has the same\n behavior as prop_kind == #dnnl_backward"]
3872 pub const dnnl_normalization_flags_none: Type = 0;
3873 #[doc = " Use global statistics\n\n If specified\n - on forward propagation use mean and variance provided by user (input)\n - on backward propagation reduces the amount of computations, since\n mean and variance are considered as constants\n\n If not specified:\n - on forward propagation mean and variance are computed and stored as\n output\n - on backward propagation compute full derivative wrt data"]
3874 pub const dnnl_use_global_stats: Type = 1;
3875 #[doc = " Use scale parameter\n\n If specified:\n - on forward propagation use scale for the normalization results\n - on backward propagation (for prop_kind == #dnnl_backward) compute\n diff wrt scale (hence one extra output used)"]
3876 pub const dnnl_use_scale: Type = 2;
3877 #[doc = " Use shift parameter\n\n If specified:\n - on forward propagation use shift (aka bias) for the normalization\n results\n - on backward propagation (for prop_kind == #dnnl_backward) compute\n diff wrt shift (hence one extra output used)"]
3878 pub const dnnl_use_shift: Type = 4;
3879 #[doc = " Fuse with ReLU\n\n The flag implies negative slope being 0. On training this is the only\n configuration supported. For inference, to use non-zero negative slope\n consider using @ref dev_guide_attributes_post_ops.\n\n If specified:\n - on inference this option behaves the same as if the primitive were\n fused with ReLU using post ops API with zero negative slope.\n - on training primitive requires workspace (required to be able to\n perform backward pass)"]
3880 pub const dnnl_fuse_norm_relu: Type = 8;
3881 #[doc = " Fuse with Add and then fuse with ReLU\n\n If specified:\n\n - on forward propagation apply element-wise binary Add operation to\n to the normalization results with an additional input tensor and then\n apply ReLU with negative slope being 0.\n - on training primitive requires workspace (required to be able to\n perform backward pass).\n - on backward propagation save the result of backward ReLU operation\n with input tensor and workspace from forward pass to extra output\n tensor and then perform backward normalization."]
3882 pub const dnnl_fuse_norm_add_relu: Type = 16;
3883}
3884#[doc = " @cond DO_NOT_DOCUMENT_THIS\n Hex representation for a **special** quiet NAN (!= NAN from math.h)"]
3885#[repr(C)]
3886#[derive(Copy, Clone)]
3887pub union _bindgen_ty_1 {
3888 pub u: ::std::os::raw::c_uint,
3889 pub f: f32,
3890}
3891#[allow(clippy::unnecessary_operation, clippy::identity_op)]
3892const _: () = {
3893 ["Size of _bindgen_ty_1"][::std::mem::size_of::<_bindgen_ty_1>() - 4usize];
3894 ["Alignment of _bindgen_ty_1"][::std::mem::align_of::<_bindgen_ty_1>() - 4usize];
3895 ["Offset of field: _bindgen_ty_1::u"][::std::mem::offset_of!(_bindgen_ty_1, u) - 0usize];
3896 ["Offset of field: _bindgen_ty_1::f"][::std::mem::offset_of!(_bindgen_ty_1, f) - 0usize];
3897};
3898unsafe extern "C" {
3899 pub static DNNL_RUNTIME_F32_VAL_REP: _bindgen_ty_1;
3900}
3901#[doc = " @cond DO_NOT_DOCUMENT_THIS"]
3902pub const DNNL_RUNTIME_S32_VAL_REP: ::std::os::raw::c_int = -2147483648;
3903#[doc = " @struct dnnl_memory_desc\n An opaque structure to describe a memory descriptor."]
3904#[repr(C)]
3905#[derive(Debug, Copy, Clone)]
3906pub struct dnnl_memory_desc {
3907 _unused: [u8; 0],
3908}
3909#[doc = " A memory descriptor handle."]
3910pub type dnnl_memory_desc_t = *mut dnnl_memory_desc;
3911#[doc = " A memory descriptor handle."]
3912pub type const_dnnl_memory_desc_t = *const dnnl_memory_desc;
3913#[doc = " @struct dnnl_memory\n An opaque structure to describe a memory."]
3914#[repr(C)]
3915#[derive(Debug, Copy, Clone)]
3916pub struct dnnl_memory {
3917 _unused: [u8; 0],
3918}
3919#[doc = " A memory handle."]
3920pub type dnnl_memory_t = *mut dnnl_memory;
3921#[doc = " A constant memory handle."]
3922pub type const_dnnl_memory_t = *const dnnl_memory;
3923pub mod dnnl_rnn_flags_t {
3924 #[doc = " Flags for RNN cell."]
3925 pub type Type = ::std::os::raw::c_uint;
3926 #[doc = " Undefined RNN flags"]
3927 pub const dnnl_rnn_flags_undef: Type = 0;
3928 #[doc = " Do not add weights gradient to existing diff_weights memory"]
3929 pub const dnnl_rnn_flags_diff_weights_overwrite: Type = 1;
3930}
3931pub mod dnnl_rnn_direction_t {
3932 #[doc = " A direction of RNN primitive execution."]
3933 pub type Type = ::std::os::raw::c_uint;
3934 #[doc = " Undefined RNN direction."]
3935 pub const dnnl_rnn_direction_undef: Type = 0;
3936 #[doc = " Unidirectional execution of RNN primitive from left to right."]
3937 pub const dnnl_unidirectional_left2right: Type = 1;
3938 #[doc = " Unidirectional execution of RNN primitive from right to left."]
3939 pub const dnnl_unidirectional_right2left: Type = 2;
3940 #[doc = " Bidirectional execution of RNN primitive with concatenation of the\n results."]
3941 pub const dnnl_bidirectional_concat: Type = 3;
3942 #[doc = " Bidirectional execution of RNN primitive with summation of the\n results."]
3943 pub const dnnl_bidirectional_sum: Type = 4;
3944}
3945#[doc = " @struct dnnl_primitive_desc\n @brief An opaque structure to describe a primitive descriptor."]
3946#[repr(C)]
3947#[derive(Debug, Copy, Clone)]
3948pub struct dnnl_primitive_desc {
3949 _unused: [u8; 0],
3950}
3951#[doc = " @brief A primitive descriptor handle."]
3952pub type dnnl_primitive_desc_t = *mut dnnl_primitive_desc;
3953#[doc = " @brief A constant primitive descriptor handle."]
3954pub type const_dnnl_primitive_desc_t = *const dnnl_primitive_desc;
3955pub mod dnnl_scratchpad_mode_t {
3956 #[doc = " Scratchpad mode"]
3957 pub type Type = ::std::os::raw::c_uint;
3958 #[doc = " The library manages the scratchpad allocation according to the policy\n specified by the `DNNL_ENABLE_CONCURRENT_EXEC`\n [build option](@ref dev_guide_build_options) (default).\n\n When `DNNL_ENABLE_CONCURRENT_EXEC=OFF` (default), the library\n scratchpad is common to all primitives to reduce the memory footprint.\n This configuration comes with limited thread-safety properties, namely\n primitives can be created and executed in parallel but cannot migrate\n between threads (in other words, each primitive should be executed in\n the same thread it was created in).\n\n When `DNNL_ENABLE_CONCURRENT_EXEC=ON`, the library scratchpad is\n private to each primitive. The memory footprint is larger than when\n using `DNNL_ENABLE_CONCURRENT_EXEC=OFF` but different primitives can be\n created and run concurrently (the same primitive cannot be run\n concurrently from two different threads though)."]
3959 pub const dnnl_scratchpad_mode_library: Type = 0;
3960 #[doc = " The user manages the scratchpad allocation by querying and providing\n the scratchpad memory to primitives. This mode is thread-safe as long\n as the scratchpad buffers are not used concurrently by two primitive\n executions."]
3961 pub const dnnl_scratchpad_mode_user: Type = 1;
3962}
3963pub mod dnnl_rounding_mode_t {
3964 #[doc = " Rounding mode"]
3965 pub type Type = ::std::os::raw::c_uint;
3966 #[doc = " rounding mode dictated by the floating-point environment"]
3967 pub const dnnl_rounding_mode_environment: Type = 0;
3968 #[doc = " stochastic rounding mode where a random bias is added to the\n trailing mantissa bits before conversion."]
3969 pub const dnnl_rounding_mode_stochastic: Type = 1;
3970}
3971#[doc = " @struct dnnl_primitive_attr\n @brief An opaque structure for primitive descriptor attributes.\n\n Attributes may contain:\n - output scales (to scale the result prior to storing it to the memory)"]
3972#[repr(C)]
3973#[derive(Debug, Copy, Clone)]
3974pub struct dnnl_primitive_attr {
3975 _unused: [u8; 0],
3976}
3977#[doc = " @brief A primitive descriptor attributes handle that controls primitive\n behavior."]
3978pub type dnnl_primitive_attr_t = *mut dnnl_primitive_attr;
3979#[doc = " @brief A constant primitive descriptor attributes handle."]
3980pub type const_dnnl_primitive_attr_t = *const dnnl_primitive_attr;
3981#[doc = " @struct dnnl_post_ops\n @brief An opaque structure for a chain of post operations.\n\n dnnl_post_ops can be used to perform some (trivial) operations like\n accumulation or eltwise after certain primitives like convolution.\n\n Post operations might be combined together, making a chain of post\n operations. For instance one can configure convolution followed by\n accumulation followed by eltwise. This might be especially beneficial\n for residual learning blocks.\n\n @warning\n Of course not all combinations are supported, so the user should handle\n errors accordingly.\n\n Supported post operations:\n - accumulation (base primitive: convolution)\n - eltwise (base primitive: convolution)"]
3982#[repr(C)]
3983#[derive(Debug, Copy, Clone)]
3984pub struct dnnl_post_ops {
3985 _unused: [u8; 0],
3986}
3987#[doc = " @brief A post operation chain handle."]
3988pub type dnnl_post_ops_t = *mut dnnl_post_ops;
3989#[doc = " @brief A constant post operation chain handle."]
3990pub type const_dnnl_post_ops_t = *const dnnl_post_ops;
3991#[doc = " @struct dnnl_primitive\n An opaque structure to describe a primitive."]
3992#[repr(C)]
3993#[derive(Debug, Copy, Clone)]
3994pub struct dnnl_primitive {
3995 _unused: [u8; 0],
3996}
3997#[doc = " A primitive handle."]
3998pub type dnnl_primitive_t = *mut dnnl_primitive;
3999#[doc = " A constant primitive handle."]
4000pub type const_dnnl_primitive_t = *const dnnl_primitive;
4001#[doc = " A structure that contains an index and a memory object, and is used to pass\n arguments to dnnl_primitive_execute()."]
4002#[repr(C)]
4003#[derive(Debug, Copy, Clone)]
4004pub struct dnnl_exec_arg_t {
4005 #[doc = "< An argument index, e.g. DNNL_ARG_SRC"]
4006 pub arg: ::std::os::raw::c_int,
4007 #[doc = "< Input/output memory"]
4008 pub memory: dnnl_memory_t,
4009}
4010#[allow(clippy::unnecessary_operation, clippy::identity_op)]
4011const _: () = {
4012 ["Size of dnnl_exec_arg_t"][::std::mem::size_of::<dnnl_exec_arg_t>() - 16usize];
4013 ["Alignment of dnnl_exec_arg_t"][::std::mem::align_of::<dnnl_exec_arg_t>() - 8usize];
4014 ["Offset of field: dnnl_exec_arg_t::arg"]
4015 [::std::mem::offset_of!(dnnl_exec_arg_t, arg) - 0usize];
4016 ["Offset of field: dnnl_exec_arg_t::memory"]
4017 [::std::mem::offset_of!(dnnl_exec_arg_t, memory) - 8usize];
4018};
4019pub mod dnnl_query_t {
4020 #[doc = " Primitive descriptor query specification\n\n For generic function dnnl_primitive_desc_query(), the type of result must\n agree with the queried argument. The correspondence table:\n\n Query kind | Type of query result\n --------------------------------|-----------------------------\n dnnl_query_*_engine | #dnnl_engine_t *\n #dnnl_query_primitive_kind | #dnnl_primitive_kind_t *\n dnnl_query_*_s32 | int *\n dnnl_query_*_s64 | #dnnl_dim_t * (same as int64_t *)\n dnnl_query_*_f32 | float *\n dnnl_query_*_f64 | double *\n dnnl_query_*_str | const char **\n dnnl_query_*_md | #const_dnnl_memory_desc_t *\n dnnl_query_*_pd | #const_dnnl_primitive_desc_t *\n dnnl_query_cache_blob_id | const uint8_t **\n dnnl_query_strides | const #dnnl_dims_t **\n dnnl_query_dilations | const #dnnl_dims_t **\n dnnl_query_padding_l | const #dnnl_dims_t **\n dnnl_query_padding_r | const #dnnl_dims_t **\n dnnl_query_flags | unsigned *\n dnnl_query_alg_kind | #dnnl_alg_kind_t *\n dnnl_query_factors | const float **\n dnnl_query_cell_kind | #dnnl_alg_kind_t *\n dnnl_query_direction | #dnnl_rnn_direction_t *\n dnnl_query_activation_kind | #dnnl_alg_kind_t *\n dnnl_query_kernel | const #dnnl_dims_t **\n dnnl_query_dims | const #dnnl_dims_t **\n dnnl_query_data_type | #dnnl_data_type_t *\n dnnl_query_padded_dims | const #dnnl_dims_t **\n dnnl_query_padded_offsets | const #dnnl_dims_t **\n dnnl_query_format_kind | #dnnl_format_kind_t *\n dnnl_query_inner_blks | const #dnnl_dims_t **\n dnnl_query_inner_idxs | const #dnnl_dims_t **\n dnnl_query_sparse_encoding | #dnnl_sparse_encoding_t *\n\n @note\n Rule of thumb: all opaque types and structures are returned by\n reference. All numbers are returned by value.\n\n @warning\n All returned references point to constant objects and are valid only\n during the lifetime of the queried primitive descriptor. Returned objects\n must not be destroyed by the user. If you need to keep the object longer\n than the lifetime of the queried primitive descriptor, use\n dnnl_primitive_desc_clone() to make a copy."]
4021 pub type Type = ::std::os::raw::c_uint;
4022 #[doc = "< no query"]
4023 pub const dnnl_query_undef: Type = 0;
4024 #[doc = "< execution engine"]
4025 pub const dnnl_query_engine: Type = 1;
4026 #[doc = "< primitive kind"]
4027 pub const dnnl_query_primitive_kind: Type = 2;
4028 #[doc = "< number of inputs expected"]
4029 pub const dnnl_query_num_of_inputs_s32: Type = 3;
4030 #[doc = "< number of outputs expected"]
4031 pub const dnnl_query_num_of_outputs_s32: Type = 4;
4032 #[doc = "< runtime estimation (seconds)"]
4033 pub const dnnl_query_time_estimate_f64: Type = 5;
4034 #[doc = "< memory consumption -- extra"]
4035 pub const dnnl_query_memory_consumption_s64: Type = 6;
4036 #[doc = "< scratchpad engine -- engine to be used"]
4037 pub const dnnl_query_scratchpad_engine: Type = 7;
4038 #[doc = "< implementation name"]
4039 pub const dnnl_query_impl_info_str: Type = 8;
4040 #[doc = "< source engine"]
4041 pub const dnnl_query_reorder_src_engine: Type = 9;
4042 #[doc = "< destination engine"]
4043 pub const dnnl_query_reorder_dst_engine: Type = 10;
4044 #[doc = "< propagation kind"]
4045 pub const dnnl_query_prop_kind: Type = 11;
4046 #[doc = "< size of cache blob ID in bytes"]
4047 pub const dnnl_query_cache_blob_id_size_s64: Type = 12;
4048 #[doc = "< cache blob ID (pointer to array)"]
4049 pub const dnnl_query_cache_blob_id: Type = 13;
4050 #[doc = "< strides"]
4051 pub const dnnl_query_strides: Type = 14;
4052 #[doc = "< dilations"]
4053 pub const dnnl_query_dilations: Type = 15;
4054 #[doc = "< left padding"]
4055 pub const dnnl_query_padding_l: Type = 16;
4056 #[doc = "< right padding"]
4057 pub const dnnl_query_padding_r: Type = 17;
4058 #[doc = "< epsilon"]
4059 pub const dnnl_query_epsilon_f32: Type = 18;
4060 #[doc = "< flags"]
4061 pub const dnnl_query_flags: Type = 19;
4062 #[doc = "< algorithm kind"]
4063 pub const dnnl_query_alg_kind: Type = 20;
4064 #[doc = "< alpha"]
4065 pub const dnnl_query_alpha_f32: Type = 21;
4066 #[doc = "< beta"]
4067 pub const dnnl_query_beta_f32: Type = 22;
4068 #[doc = "< axis"]
4069 pub const dnnl_query_axis_s32: Type = 23;
4070 #[doc = "< LRN parameter local size"]
4071 pub const dnnl_query_local_size_s64: Type = 24;
4072 #[doc = "< LRN parameter K"]
4073 pub const dnnl_query_k_f32: Type = 25;
4074 #[doc = "< Reduction parameter P"]
4075 pub const dnnl_query_p_f32: Type = 26;
4076 #[doc = "< Resampling parameter factors"]
4077 pub const dnnl_query_factors: Type = 27;
4078 #[doc = "< RNN parameter cell kind"]
4079 pub const dnnl_query_cell_kind: Type = 28;
4080 #[doc = "< RNN parameter direction"]
4081 pub const dnnl_query_direction: Type = 29;
4082 #[doc = "< RNN parameter activation kind"]
4083 pub const dnnl_query_activation_kind: Type = 30;
4084 #[doc = "< Pooling parameter kernel"]
4085 pub const dnnl_query_kernel: Type = 31;
4086 #[doc = "< Shuffle parameter group size"]
4087 pub const dnnl_query_group_size_s64: Type = 32;
4088 #[doc = "< stub"]
4089 pub const dnnl_query_some_md: Type = 128;
4090 #[doc = "< source memory desc"]
4091 pub const dnnl_query_src_md: Type = 129;
4092 #[doc = "< source gradient memory desc"]
4093 pub const dnnl_query_diff_src_md: Type = 130;
4094 #[doc = "< weights memory descriptor desc"]
4095 pub const dnnl_query_weights_md: Type = 131;
4096 #[doc = "< weights grad. memory desc"]
4097 pub const dnnl_query_diff_weights_md: Type = 132;
4098 #[doc = "< destination memory desc"]
4099 pub const dnnl_query_dst_md: Type = 133;
4100 #[doc = "< destination grad. memory desc"]
4101 pub const dnnl_query_diff_dst_md: Type = 134;
4102 #[doc = "< workspace memory desc"]
4103 pub const dnnl_query_workspace_md: Type = 135;
4104 #[doc = "< scratchpad memory desc"]
4105 pub const dnnl_query_scratchpad_md: Type = 136;
4106 #[doc = "< memory desc of an execute argument"]
4107 pub const dnnl_query_exec_arg_md: Type = 255;
4108 #[doc = "< number of dimensions"]
4109 pub const dnnl_query_ndims_s32: Type = 256;
4110 #[doc = "< vector of dimensions"]
4111 pub const dnnl_query_dims: Type = 257;
4112 #[doc = "< data type"]
4113 pub const dnnl_query_data_type: Type = 258;
4114 #[doc = "< submemory offset"]
4115 pub const dnnl_query_submemory_offset_s64: Type = 259;
4116 #[doc = "< vector of padded dimensions"]
4117 pub const dnnl_query_padded_dims: Type = 260;
4118 #[doc = "< vector of padded offsets"]
4119 pub const dnnl_query_padded_offsets: Type = 261;
4120 #[doc = "< format kind"]
4121 pub const dnnl_query_format_kind: Type = 262;
4122 #[doc = "< number of innermost blocks"]
4123 pub const dnnl_query_inner_nblks_s32: Type = 263;
4124 #[doc = "< vector of sizes of the innermost blocks"]
4125 pub const dnnl_query_inner_blks: Type = 264;
4126 #[doc = "< vector of logical indices of the blocks"]
4127 pub const dnnl_query_inner_idxs: Type = 265;
4128 pub const dnnl_query_max: Type = 32767;
4129}
4130pub mod dnnl_cpu_isa_t {
4131 #[doc = " CPU instruction set flags"]
4132 pub type Type = ::std::os::raw::c_uint;
4133 #[doc = " Library choice of ISA (excepting those listed as initial support)"]
4134 pub const dnnl_cpu_isa_default: Type = 0;
4135 #[doc = " Intel Streaming SIMD Extensions 4.1 (Intel SSE4.1)"]
4136 pub const dnnl_cpu_isa_sse41: Type = 1;
4137 #[doc = " Intel Advanced Vector Extensions (Intel AVX)"]
4138 pub const dnnl_cpu_isa_avx: Type = 3;
4139 #[doc = " Intel Advanced Vector Extensions 2 (Intel AVX2)"]
4140 pub const dnnl_cpu_isa_avx2: Type = 7;
4141 #[doc = " Intel AVX2 and Intel Deep Learning Boost (Intel DL Boost) support"]
4142 pub const dnnl_cpu_isa_avx2_vnni: Type = 15;
4143 #[doc = " Intel AVX2 and Intel Deep Learning Boost (Intel DL Boost)\n with 8-bit integer, float16 and bfloat16 support"]
4144 pub const dnnl_cpu_isa_avx2_vnni_2: Type = 31;
4145 #[doc = " Intel AVX-512 subset for Intel Xeon Scalable processor family\n and Intel Core processor family."]
4146 pub const dnnl_cpu_isa_avx512_core: Type = 39;
4147 #[doc = " Intel AVX-512 and Intel Deep Learning Boost (Intel DL Boost) support\n for Intel Xeon Scalable processor family\n and Intel Core processor family."]
4148 pub const dnnl_cpu_isa_avx512_core_vnni: Type = 103;
4149 #[doc = " Intel AVX-512, Intel DL Boost and bfloat16 support\n for Intel Xeon Scalable processor family\n and Intel Core processor family."]
4150 pub const dnnl_cpu_isa_avx512_core_bf16: Type = 231;
4151 #[doc = " Intel AVX-512 with float16, Intel DL Boost and bfloat16 support\n for Intel Xeon Scalable processor family\n and Intel Core processor family."]
4152 pub const dnnl_cpu_isa_avx10_1_512: Type = 495;
4153 #[doc = " @copydoc dnnl_cpu_isa_avx10_1_512"]
4154 pub const dnnl_cpu_isa_avx512_core_fp16: Type = 495;
4155 #[doc = " Intel AVX-512 with float16, Intel DL Boost and bfloat16 support and\n Intel AMX with 8-bit integer and bfloat16 support"]
4156 pub const dnnl_cpu_isa_avx10_1_512_amx: Type = 4079;
4157 #[doc = " @copydoc dnnl_cpu_isa_avx10_1_512_amx"]
4158 pub const dnnl_cpu_isa_avx512_core_amx: Type = 4079;
4159 #[doc = " Intel AVX-512 with float16, Intel DL Boost and bfloat16 support and\n Intel AMX with 8-bit integer, bfloat16 and float16 support"]
4160 pub const dnnl_cpu_isa_avx10_1_512_amx_fp16: Type = 8175;
4161 #[doc = " @copydoc dnnl_cpu_isa_avx10_1_512_amx_fp16"]
4162 pub const dnnl_cpu_isa_avx512_core_amx_fp16: Type = 8175;
4163}
4164pub mod dnnl_cpu_isa_hints_t {
4165 #[doc = " CPU ISA hints flags"]
4166 pub type Type = ::std::os::raw::c_uint;
4167 #[doc = " No hints (use default features)"]
4168 pub const dnnl_cpu_isa_no_hints: Type = 0;
4169 #[doc = " Prefer to exclusively use Ymm registers for computations"]
4170 pub const dnnl_cpu_isa_prefer_ymm: Type = 1;
4171}
4172unsafe extern "C" {
4173 #[doc = " Changes the primitive descriptor to point to the next available\n implementation.\n\n @param primitive_desc A primitive descriptor to change.\n @returns #dnnl_success on success and a status describing the error\n otherwise.\n @returns #dnnl_last_impl_reached if no more implementations available,\n in which case the primitive descriptor itself is kept unchanged."]
4174 pub fn dnnl_primitive_desc_next_impl(
4175 primitive_desc: dnnl_primitive_desc_t,
4176 ) -> dnnl_status_t::Type;
4177}
4178unsafe extern "C" {
4179 #[doc = " Clones a primitive descriptor. The resulting primitive descriptor must be\n destroyed separately.\n\n @param primitive_desc Output primitive descriptor.\n @param existing_primitive_desc Primitive descriptor to clone.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4180 pub fn dnnl_primitive_desc_clone(
4181 primitive_desc: *mut dnnl_primitive_desc_t,
4182 existing_primitive_desc: const_dnnl_primitive_desc_t,
4183 ) -> dnnl_status_t::Type;
4184}
4185unsafe extern "C" {
4186 #[doc = " Returns a constant reference to the attributes of a primitive descriptor.\n\n @warning\n It is an error to destroy the resulting @p attr.\n\n @warning\n The lifetime of an @p attr is the same as that of a @p\n primitive_desc, so it is an error to use the @p attr once the @p\n primitive_desc has been destroyed.\n\n @param primitive_desc Primitive descriptor.\n @param attr Output primitive attributes.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4187 pub fn dnnl_primitive_desc_get_attr(
4188 primitive_desc: const_dnnl_primitive_desc_t,
4189 attr: *mut const_dnnl_primitive_attr_t,
4190 ) -> dnnl_status_t::Type;
4191}
4192unsafe extern "C" {
4193 #[doc = " Destroys a primitive descriptor.\n\n @param primitive_desc Primitive descriptor to destroy.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4194 pub fn dnnl_primitive_desc_destroy(
4195 primitive_desc: dnnl_primitive_desc_t,
4196 ) -> dnnl_status_t::Type;
4197}
4198unsafe extern "C" {
4199 #[doc = " Queries a primitive descriptor for various pieces of information.\n\n The most common use case is to query a primitive descriptor, created with\n source, weights, and destination memory descriptors with format tags set\n to #dnnl_format_tag_any, for the corresponding memory descriptors (in this\n case the @p what is set to #dnnl_query_src_md, #dnnl_query_weights_md, and\n #dnnl_query_dst_md respectively) so that it is possible to create memory\n objects and reorder primitives if necessary.\n\n Another typical use case is to query a primitive descriptor for workspace\n memory descriptor (with @p what set to #dnnl_query_workspace_md). If this\n query returns #dnnl_not_required status, then workspace memory is not\n required.\n\n @note\n When querying for a memory descriptor for a scratchpad, a workspace,\n or an optional parameter, the query will return a pointer to a zero\n memory descriptor if the parameter is not needed.\n\n A few other use cases:\n - query a primitive descriptor for the implementation information string\n (#dnnl_query_impl_info_str)\n - query a primitive descriptor for the number of inputs and outputs\n (#dnnl_query_num_of_inputs_s32 and #dnnl_query_num_of_outputs_s32\n respectively)\n\n @sa dnnl_query_t for more options\n\n @param primitive_desc Primitive descriptor.\n @param what Parameter to query.\n @param index Index of the parameter to query for.\n @param result Output result. The type depends on the query. For example,\n it must be a @c dnnl_memory_desc_t* if querying for a memory\n descriptor.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4200 pub fn dnnl_primitive_desc_query(
4201 primitive_desc: const_dnnl_primitive_desc_t,
4202 what: dnnl_query_t::Type,
4203 index: ::std::os::raw::c_int,
4204 result: *mut ::std::os::raw::c_void,
4205 ) -> dnnl_status_t::Type;
4206}
4207unsafe extern "C" {
4208 #[doc = " Queries primitive descriptor for a memory descriptor.\n\n @note\n This function is a convenience version of\n #dnnl_primitive_desc_query().\n\n @param primitive_desc Primitive descriptor.\n @param what Kind of memory descriptor parameter to query for.\n @param index Index of the parameter to query.\n @returns A pointer to the requested memory descriptor.\n @returns A pointer to a zero memory descriptor if the parameter is not\n needed.\n @returns NULL in case of any error.\n"]
4209 pub fn dnnl_primitive_desc_query_md(
4210 primitive_desc: const_dnnl_primitive_desc_t,
4211 what: dnnl_query_t::Type,
4212 index: ::std::os::raw::c_int,
4213 ) -> const_dnnl_memory_desc_t;
4214}
4215unsafe extern "C" {
4216 #[doc = " Queries primitive descriptor for a signed 32bit int.\n\n @note\n This function is a convenience version of\n #dnnl_primitive_desc_query().\n\n @param primitive_desc Primitive descriptor.\n @param what Kind of the value to query for.\n @param index Index of the parameter to query.\n @returns The requested value.\n @returns 0 in case of any error (in particular if the queried entity is\n not of type int32_t). Note that 0 may also be the actual returned\n value."]
4217 pub fn dnnl_primitive_desc_query_s32(
4218 primitive_desc: const_dnnl_primitive_desc_t,
4219 what: dnnl_query_t::Type,
4220 index: ::std::os::raw::c_int,
4221 ) -> ::std::os::raw::c_int;
4222}
4223unsafe extern "C" {
4224 #[doc = " Creates a primitive.\n\n @param primitive Output primitive.\n @param primitive_desc Primitive descriptor used to create the primitive.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4225 pub fn dnnl_primitive_create(
4226 primitive: *mut dnnl_primitive_t,
4227 primitive_desc: const_dnnl_primitive_desc_t,
4228 ) -> dnnl_status_t::Type;
4229}
4230unsafe extern "C" {
4231 #[doc = " Creates a primitive from a cache blob.\n\n @param primitive Output primitive.\n @param primitive_desc Primitive descriptor used to create the primitive.\n @param size Size of the cache blob in bytes.\n @param cache_blob Cache blob of size @p size.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4232 pub fn dnnl_primitive_create_from_cache_blob(
4233 primitive: *mut dnnl_primitive_t,
4234 primitive_desc: const_dnnl_primitive_desc_t,
4235 size: usize,
4236 cache_blob: *const u8,
4237 ) -> dnnl_status_t::Type;
4238}
4239unsafe extern "C" {
4240 #[doc = " @note If any argument in @p args is padded (padded_dims >\n dims), the primitive execution will assume properly zero-padded\n input arguments, and produce zero-padded output arguments."]
4241 pub fn dnnl_primitive_execute(
4242 primitive: const_dnnl_primitive_t,
4243 stream: dnnl_stream_t,
4244 nargs: ::std::os::raw::c_int,
4245 args: *const dnnl_exec_arg_t,
4246 ) -> dnnl_status_t::Type;
4247}
4248unsafe extern "C" {
4249 #[doc = " Retrieves a constant reference to the primitive descriptor of a given\n primitive.\n\n @warning\n It is an error to destroy the returned object. It is owned by the\n primitive. The @c const qualifier of the returned object prevents\n such attempts.\n\n @param primitive Primitive to query for the primitive descriptor.\n @param primitive_desc Output primitive descriptor.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4250 pub fn dnnl_primitive_get_primitive_desc(
4251 primitive: const_dnnl_primitive_t,
4252 primitive_desc: *mut const_dnnl_primitive_desc_t,
4253 ) -> dnnl_status_t::Type;
4254}
4255unsafe extern "C" {
4256 #[doc = " Retrieves a cache blob associated with the given primitive.\n\n @param primitive Primitive to query for the cache blob.\n @param size Size of the cache blob in bytes.\n @param cache_blob Cache blob of size @p size. If the @p cache_blob is\n nullptr then the size of the cache blob is returned in @p size.\n @returns #dnnl_success on success and a status describing the error\n otherwise.\n\n @note The cache blob can be empty. It's the user's responsibility to check\n whether it's empty prior to passing it to\n #dnnl_primitive_create_from_cache_blob()."]
4257 pub fn dnnl_primitive_get_cache_blob(
4258 primitive: const_dnnl_primitive_t,
4259 size: *mut usize,
4260 cache_blob: *mut u8,
4261 ) -> dnnl_status_t::Type;
4262}
4263unsafe extern "C" {
4264 #[doc = " Destroys a primitive.\n\n @param primitive The primitive to destroy.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4265 pub fn dnnl_primitive_destroy(primitive: dnnl_primitive_t) -> dnnl_status_t::Type;
4266}
4267unsafe extern "C" {
4268 #[doc = " Creates an empty (default) primitive attributes with all the parameters\n set to their default values.\n\n Empty attributes are implied whenever the respective argument is NULL.\n\n @param attr Output primitive attributes.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4269 pub fn dnnl_primitive_attr_create(attr: *mut dnnl_primitive_attr_t) -> dnnl_status_t::Type;
4270}
4271unsafe extern "C" {
4272 #[doc = " Clones primitive attributes.\n\n @param attr Output primitive attributes.\n @param existing_attr Primitive attributes to clone.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4273 pub fn dnnl_primitive_attr_clone(
4274 attr: *mut dnnl_primitive_attr_t,
4275 existing_attr: const_dnnl_primitive_attr_t,
4276 ) -> dnnl_status_t::Type;
4277}
4278unsafe extern "C" {
4279 #[doc = " Destroys primitive attributes.\n\n @param attr Primitive attributes to destroy.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4280 pub fn dnnl_primitive_attr_destroy(attr: dnnl_primitive_attr_t) -> dnnl_status_t::Type;
4281}
4282unsafe extern "C" {
4283 #[doc = " Returns probability for output dropout primitive attribute.\n\n @param attr Primitive attributes.\n @param dropout_desc Output dropout memory descriptor\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4284 pub fn dnnl_primitive_attr_get_dropout(
4285 attr: const_dnnl_primitive_attr_t,
4286 dropout_desc: *mut const_dnnl_memory_desc_t,
4287 ) -> dnnl_status_t::Type;
4288}
4289unsafe extern "C" {
4290 #[doc = " Sets probability for output dropout primitive attribute.\n\n @param attr Primitive attributes.\n @param dropout_desc Output dropout memory descriptor\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4291 pub fn dnnl_primitive_attr_set_dropout(
4292 attr: dnnl_primitive_attr_t,
4293 dropout_desc: const_dnnl_memory_desc_t,
4294 ) -> dnnl_status_t::Type;
4295}
4296unsafe extern "C" {
4297 #[doc = " Returns the floating-point math mode primitive attribute.\n\n @param attr Primitive attributes.\n @param mode Output FP math mode.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4298 pub fn dnnl_primitive_attr_get_fpmath_mode(
4299 attr: const_dnnl_primitive_attr_t,
4300 mode: *mut dnnl_fpmath_mode_t::Type,
4301 ) -> dnnl_status_t::Type;
4302}
4303unsafe extern "C" {
4304 #[doc = " Sets the floating-point math mode primitive attributes.\n\n @param attr Primitive attributes.\n @param mode FP math mode. The possible values are:\n #dnnl_fpmath_mode_strict (default),\n #dnnl_fpmath_mode_bf16,\n #dnnl_fpmath_mode_f16,\n #dnnl_fpmath_mode_tf32,\n #dnnl_fpmath_mode_any.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4305 pub fn dnnl_primitive_attr_set_fpmath_mode(
4306 attr: dnnl_primitive_attr_t,
4307 mode: dnnl_fpmath_mode_t::Type,
4308 ) -> dnnl_status_t::Type;
4309}
4310unsafe extern "C" {
4311 #[doc = " Returns the floating-point math mode primitive attribute.\n\n @param attr Primitive attributes.\n @param mode Output FP math mode.\n @param apply_to_int Output use floating-point arithmetic for integer primitives.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4312 pub fn dnnl_primitive_attr_get_fpmath_mode_v2(
4313 attr: const_dnnl_primitive_attr_t,
4314 mode: *mut dnnl_fpmath_mode_t::Type,
4315 apply_to_int: *mut ::std::os::raw::c_int,
4316 ) -> dnnl_status_t::Type;
4317}
4318unsafe extern "C" {
4319 #[doc = " Sets the floating-point math mode primitive attributes.\n\n @param attr Primitive attributes.\n @param mode FP math mode. The possible values are:\n #dnnl_fpmath_mode_strict (default),\n #dnnl_fpmath_mode_bf16,\n #dnnl_fpmath_mode_f16,\n #dnnl_fpmath_mode_tf32,\n #dnnl_fpmath_mode_any.\n @param apply_to_int Boolean. Use of floating-point arithmetic for integer primitives.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4320 pub fn dnnl_primitive_attr_set_fpmath_mode_v2(
4321 attr: dnnl_primitive_attr_t,
4322 mode: dnnl_fpmath_mode_t::Type,
4323 apply_to_int: ::std::os::raw::c_int,
4324 ) -> dnnl_status_t::Type;
4325}
4326unsafe extern "C" {
4327 #[doc = " Returns the deterministic primitive attribute value.\n\n @param attr Primitive attributes.\n @param value Output deterministic attribute value\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4328 pub fn dnnl_primitive_attr_get_deterministic(
4329 attr: const_dnnl_primitive_attr_t,
4330 value: *mut ::std::os::raw::c_int,
4331 ) -> dnnl_status_t::Type;
4332}
4333unsafe extern "C" {
4334 #[doc = " Sets the deterministic primitive attribute value.\n\n @param attr Primitive attributes.\n @param value Boolean value to set deterministic attribute.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4335 pub fn dnnl_primitive_attr_set_deterministic(
4336 attr: dnnl_primitive_attr_t,
4337 value: ::std::os::raw::c_int,
4338 ) -> dnnl_status_t::Type;
4339}
4340unsafe extern "C" {
4341 #[doc = " Returns the accumulation mode primitive attribute.\n\n @param attr Primitive attributes.\n @param mode Output accumulation mode.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4342 pub fn dnnl_primitive_attr_get_accumulation_mode(
4343 attr: const_dnnl_primitive_attr_t,
4344 mode: *mut dnnl_accumulation_mode_t::Type,
4345 ) -> dnnl_status_t::Type;
4346}
4347unsafe extern "C" {
4348 #[doc = " Sets the accumulation mode primitive attribute.\n\n @param attr Primitive attributes.\n @param mode Accumulation mode. The possible values are:\n #dnnl_accumulation_mode_strict (default), which is s32 for quantized primitives, f32/f64 otherwise\n #dnnl_accumulation_mode_relaxed, which is same as strict but allows intermediate accumulators to be in src/dst datatype\n #dnnl_accumulation_mode_any, which allows accumulators to be src/dst datatype or any wider type.\n #dnnl_accumulation_mode_f32,\n #dnnl_accumulation_mode_s32,\n #dnnl_accumulation_mode_f16.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4349 pub fn dnnl_primitive_attr_set_accumulation_mode(
4350 attr: dnnl_primitive_attr_t,
4351 mode: dnnl_accumulation_mode_t::Type,
4352 ) -> dnnl_status_t::Type;
4353}
4354unsafe extern "C" {
4355 #[doc = " Returns the primitive attributes scratchpad mode.\n\n @param attr Primitive attributes.\n @param mode Output scratchpad mode.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4356 pub fn dnnl_primitive_attr_get_scratchpad_mode(
4357 attr: const_dnnl_primitive_attr_t,
4358 mode: *mut dnnl_scratchpad_mode_t::Type,
4359 ) -> dnnl_status_t::Type;
4360}
4361unsafe extern "C" {
4362 #[doc = " Sets primitive attributes scratchpad mode.\n\n @param attr Primitive attributes.\n @param mode Scratchpad mode. The possible values are:\n #dnnl_scratchpad_mode_library (default) and\n #dnnl_scratchpad_mode_user.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4363 pub fn dnnl_primitive_attr_set_scratchpad_mode(
4364 attr: dnnl_primitive_attr_t,
4365 mode: dnnl_scratchpad_mode_t::Type,
4366 ) -> dnnl_status_t::Type;
4367}
4368unsafe extern "C" {
4369 #[doc = " Sets primitive attributes scaling factors for primitive operations for a\n given memory argument. The scaling factors must be passed at execution time\n as an argument with index #DNNL_ARG_ATTR_SCALES | arg.\n\n @sa dnnl_primitive_attr_set_scales_mask\n\n\n @param attr Primitive attributes.\n @param arg Parameter argument index as passed to the\n dnnl_primitive_execute() call.\n @param mask Scaling factors correspondence mask that defines the\n correspondence between the tensor dimensions and the @p scales array.\n The set i-th bit indicates that a dedicated scaling factor is used for\n each index along that dimension. Set the mask to 0 to use a common\n scaling factor for the whole output tensor.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4370 pub fn dnnl_primitive_attr_set_scales_mask(
4371 attr: dnnl_primitive_attr_t,
4372 arg: ::std::os::raw::c_int,
4373 mask: ::std::os::raw::c_int,
4374 ) -> dnnl_status_t::Type;
4375}
4376unsafe extern "C" {
4377 #[doc = " Sets primitive attributes scaling factors for primitive operations for a\n given memory argument. The scaling factors must be passed at execution time\n as an argument with index #DNNL_ARG_ATTR_SCALES | arg.\n\n @sa dnnl_primitive_attr_set_scales\n\n\n @param attr Primitive attributes.\n @param arg Parameter argument index as passed to the\n dnnl_primitive_execute() call.\n @param mask Scaling factors correspondence mask that defines the\n correspondence between the tensor dimensions and the @p scales array.\n The set i-th bit indicates that a dedicated scaling factor is used for\n each index along that dimension. Set the mask to 0 to use a common\n scaling factor for the whole output tensor.\n @param ndims Number of group dimensions.\n @param group_dims Scaling factors correspondence groups that define the\n correspondence between the tensor dimensions and the scales array.\n The group dimensions should only be provided for each logical dimension\n that has correspondence mask @p mask set.\n @param data_type Scaling factors data_type.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4378 pub fn dnnl_primitive_attr_set_scales(
4379 attr: dnnl_primitive_attr_t,
4380 arg: ::std::os::raw::c_int,
4381 mask: ::std::os::raw::c_int,
4382 ndims: ::std::os::raw::c_int,
4383 group_dims: *const dnnl_dim_t,
4384 data_type: dnnl_data_type_t::Type,
4385 ) -> dnnl_status_t::Type;
4386}
4387unsafe extern "C" {
4388 #[doc = " Sets primitive attributes zero points for primitive operations for a given\n memory argument. The zero points must be passed at execution time\n as an argument with index #DNNL_ARG_ATTR_ZERO_POINTS | arg.\n\n @sa dnnl_primitive_attr_set_zero_points_mask\n\n\n @param attr Primitive attributes.\n @param arg Parameter argument index as passed to the\n dnnl_primitive_execute() call.\n @param mask Zero point correspondence mask that defines the\n correspondence between the tensor dimensions and the @p\n zero_points array. The set i-th bit indicates that a dedicated\n zero point is used for each index along that dimension. Set the\n mask to 0 to use a common zero point for the whole output tensor.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4389 pub fn dnnl_primitive_attr_set_zero_points_mask(
4390 attr: dnnl_primitive_attr_t,
4391 arg: ::std::os::raw::c_int,
4392 mask: ::std::os::raw::c_int,
4393 ) -> dnnl_status_t::Type;
4394}
4395unsafe extern "C" {
4396 #[doc = " Sets primitive attributes zero points for primitive operations for a given\n memory argument. The zero points must be passed at execution time\n as an argument with index #DNNL_ARG_ATTR_ZERO_POINTS | arg.\n\n @sa dnnl_primitive_attr_set_zero_points\n\n\n @param attr Primitive attributes.\n @param arg Parameter argument index as passed to the\n dnnl_primitive_execute() call.\n @param mask Zero point correspondence mask that defines the\n correspondence between the tensor dimensions and the @p\n zero_points array. The set i-th bit indicates that a dedicated\n zero point is used for each index along that dimension. Set the\n mask to 0 to use a common zero point for the whole output tensor.\n @param ndims Number of group dimensions.\n @param group_dims Zero point factors correspondence groups that define the\n correspondence between the tensor dimensions and the zero_points array.\n The group dimensions should be only provided for each logical dimension\n that has the bit set correspondence mask @p mask set.\n @param data_type Zero points factors data_type.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4397 pub fn dnnl_primitive_attr_set_zero_points(
4398 attr: dnnl_primitive_attr_t,
4399 arg: ::std::os::raw::c_int,
4400 mask: ::std::os::raw::c_int,
4401 ndims: ::std::os::raw::c_int,
4402 group_dims: *const dnnl_dim_t,
4403 data_type: dnnl_data_type_t::Type,
4404 ) -> dnnl_status_t::Type;
4405}
4406unsafe extern "C" {
4407 #[doc = " Sets the rounding mode attribute value for a given argument\n\n @param attr Primitive attributes.\n @param arg Argument for which rounding mode should be set.\n @param mode Rounding mode to apply to the argument.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4408 pub fn dnnl_primitive_attr_set_rounding(
4409 attr: dnnl_primitive_attr_t,
4410 arg: ::std::os::raw::c_int,
4411 mode: dnnl_rounding_mode_t::Type,
4412 ) -> dnnl_status_t::Type;
4413}
4414unsafe extern "C" {
4415 #[doc = " Returns the rounding mode attribute value for a given argument\n\n @param attr Primitive attributes.\n @param arg Argument for which rounding mode query applies.\n @param mode Output rounding mode.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4416 pub fn dnnl_primitive_attr_get_rounding(
4417 attr: dnnl_primitive_attr_t,
4418 arg: ::std::os::raw::c_int,
4419 mode: *mut dnnl_rounding_mode_t::Type,
4420 ) -> dnnl_status_t::Type;
4421}
4422unsafe extern "C" {
4423 #[doc = " Returns primitive attributes post-ops.\n\n @warning\n The output @p post_ops points to the internal @p attr field, so it is\n an error to modify or destroy them. The lifetime of @p post_ops is\n the same as that of the @p attr it belongs to, so it is an error to\n use @p post_ops after @p attr has been destroyed.\n\n @param attr Primitive attributes.\n @param post_ops Output post-ops.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4424 pub fn dnnl_primitive_attr_get_post_ops(
4425 attr: const_dnnl_primitive_attr_t,
4426 post_ops: *mut const_dnnl_post_ops_t,
4427 ) -> dnnl_status_t::Type;
4428}
4429unsafe extern "C" {
4430 #[doc = " Sets primitive attributes post-ops.\n\n @note\n There is no way to check whether the post-ops would be supported by\n the target primitive. Any error will be reported by the\n dnnl_<primitive name>_[propagation kind]_primitive_desc_create() function call.\n\n @param attr Primitive attributes.\n @param post_ops Post-ops to set.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4431 pub fn dnnl_primitive_attr_set_post_ops(
4432 attr: dnnl_primitive_attr_t,
4433 post_ops: const_dnnl_post_ops_t,
4434 ) -> dnnl_status_t::Type;
4435}
4436unsafe extern "C" {
4437 #[doc = " Creates empty post-ops sequence.\n\n @param post_ops Output post-ops.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4438 pub fn dnnl_post_ops_create(post_ops: *mut dnnl_post_ops_t) -> dnnl_status_t::Type;
4439}
4440unsafe extern "C" {
4441 #[doc = " Clones post-ops primitive attribute.\n\n @param post_ops Output post-ops primitive attribute.\n @param existing_post_ops Post-ops primitive attribute to clone.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4442 pub fn dnnl_post_ops_clone(
4443 post_ops: *mut dnnl_post_ops_t,
4444 existing_post_ops: const_dnnl_post_ops_t,
4445 ) -> dnnl_status_t::Type;
4446}
4447unsafe extern "C" {
4448 #[doc = " Destroys post-ops.\n\n @param post_ops Post-ops to destroy.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4449 pub fn dnnl_post_ops_destroy(post_ops: dnnl_post_ops_t) -> dnnl_status_t::Type;
4450}
4451unsafe extern "C" {
4452 #[doc = " Returns the length of post-ops.\n\n @param post_ops Post-ops.\n @returns The number of post-ops entries."]
4453 pub fn dnnl_post_ops_len(post_ops: const_dnnl_post_ops_t) -> ::std::os::raw::c_int;
4454}
4455unsafe extern "C" {
4456 #[doc = " Returns the kind of a post-op entry.\n\n @param post_ops Post-ops.\n @param index Post-op entry index.\n @returns The kind of the post-op with the specified index.\n @returns #dnnl_undefined_primitive if there is no post-op at the specified\n index."]
4457 pub fn dnnl_post_ops_get_kind(
4458 post_ops: const_dnnl_post_ops_t,
4459 index: ::std::os::raw::c_int,
4460 ) -> dnnl_primitive_kind_t::Type;
4461}
4462unsafe extern "C" {
4463 #[doc = " Appends an accumulation v3 (sum) to post-ops. Prior to accumulating the\n result, a zero point is subtracted from the previous value and is\n multiplied by the scale.\n\n The kind of this post-op is #dnnl_sum.\n\n This feature may improve performance for cases like dequantize the\n asymmetrically quantized sum's src1 tensor to f32 domain before performing\n the sum operation by subtracting the @p zero_point before the scaling.\n\n In the simplest case where accumulation is the only post-op, the\n computations will be:\n\n dst[:] <- scale * (dst[:] - zero_point) + op(...)\n // instead of dst[:] <- op(...)\n\n If @p data_type is specified, original dst tensor will be reinterpreted\n as a tensor with provided data type. Since it is reinterpretation,\n data_type and dst data type should have the same size.\n As a result, computations will be:\n\n dst[:] <- scale * (as_data_type(dst[:]) - zero_point) + op(...)\n // instead of dst[:] <- op(...)\n @note\n This post-op executes in-place and does not change the\n destination layout.\n\n @param post_ops Post-ops.\n @param scale Accumulation scaling factor.\n @param zero_point Single scalar int32_t value of zero point.\n @param data_type Accumulation data_type.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4464 pub fn dnnl_post_ops_append_sum(
4465 post_ops: dnnl_post_ops_t,
4466 scale: f32,
4467 zero_point: i32,
4468 data_type: dnnl_data_type_t::Type,
4469 ) -> dnnl_status_t::Type;
4470}
4471unsafe extern "C" {
4472 #[doc = " Returns the parameters of an accumulation (sum) post-op with\n zero point and data type parameter.\n\n @param post_ops Post-ops.\n @param index Index of the sum post-op.\n @param scale Output accumulation scaling factor.\n @param zero_point Zero point.\n @param data_type Data type for accumulation.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4473 pub fn dnnl_post_ops_get_params_sum(
4474 post_ops: const_dnnl_post_ops_t,
4475 index: ::std::os::raw::c_int,
4476 scale: *mut f32,
4477 zero_point: *mut i32,
4478 data_type: *mut dnnl_data_type_t::Type,
4479 ) -> dnnl_status_t::Type;
4480}
4481unsafe extern "C" {
4482 #[doc = " Appends an elementwise post-op.\n\n The kind of this post operation is #dnnl_eltwise.\n\n In the simplest case when the elementwise is the only post operation, the\n computations would be:\n\n dst[:] <- eltwise_op (op(...)) // instead of dst[:] <- op(...)\n\n where eltwise_op is configured with the given parameters.\n\n @param post_ops Post-ops.\n @param alg_kind Elementwise algorithm for the post-op.\n @param alpha Alpha parameter for the elementwise algorithm.\n @param beta Beta parameter for the elementwise algorithm.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4483 pub fn dnnl_post_ops_append_eltwise(
4484 post_ops: dnnl_post_ops_t,
4485 alg_kind: dnnl_alg_kind_t::Type,
4486 alpha: f32,
4487 beta: f32,
4488 ) -> dnnl_status_t::Type;
4489}
4490unsafe extern "C" {
4491 #[doc = " Returns the parameters of an elementwise post-op.\n\n @param post_ops Post-ops.\n @param index Index of the elementwise post-op.\n @param alg_kind Output elementwise algorithm kind.\n @param alpha Output alpha parameter for the elementwise algorithm.\n @param beta Output beta parameter for the elementwise algorithm.\n @returns #dnnl_success on success and a status describing the error\n otherwise.\n @returns #dnnl_invalid_arguments if @p index does not refer to an\n elementwise post-op."]
4492 pub fn dnnl_post_ops_get_params_eltwise(
4493 post_ops: const_dnnl_post_ops_t,
4494 index: ::std::os::raw::c_int,
4495 alg_kind: *mut dnnl_alg_kind_t::Type,
4496 alpha: *mut f32,
4497 beta: *mut f32,
4498 ) -> dnnl_status_t::Type;
4499}
4500unsafe extern "C" {
4501 #[doc = " Appends a depthwise post-op convolution.\n\n This post-op can only be fused with a 2D 1x1 convolution (convolution with\n weights spatial dimensions equal to 1 i.e., kh=kw=1).\n\n The kind of this post-op is #dnnl_convolution.\n\n The number of outputs for primitive with fusion is one. The output spatial\n size can be derived as below:\n\n output_height = ceil(output_height_1x1_convolution, stride)\n output_width = ceil(output_width_1x1_convolution, stride)\n\n See @ref dev_guide_attributes_post_ops_depthwise and\n @ref dev_guide_attributes_post_ops_depthwise_fusion for more info.\n\n @param post_ops Post-ops.\n @param weights_data_type Weights data type of depthwise post-op\n @param bias_data_type Bias data type of depthwise post-op\n @param dst_data_type Output data type of depthwise post-op\n @param kernel_size Size of kernel of depthwise post-op\n @param stride_size Size of stride of depthwise post-op\n @param padding_l_size Size of left and top paddings of depthwise post-op\n @returns #dnnl_success on success and a status describing the error\n otherwise"]
4502 pub fn dnnl_post_ops_append_dw(
4503 post_ops: dnnl_post_ops_t,
4504 weights_data_type: dnnl_data_type_t::Type,
4505 bias_data_type: dnnl_data_type_t::Type,
4506 dst_data_type: dnnl_data_type_t::Type,
4507 kernel_size: dnnl_dim_t,
4508 stride_size: dnnl_dim_t,
4509 padding_l_size: dnnl_dim_t,
4510 ) -> dnnl_status_t::Type;
4511}
4512unsafe extern "C" {
4513 #[doc = " Returns the parameters of an depthwise post-op.\n\n @param post_ops Post-ops.\n @param index Index of the elementwise post-op.\n @param weights_data_type Weights data type of depthwise post-op\n @param bias_data_type Bias data type of depthwise post-op\n @param dst_data_type Output data type of depthwise post-op\n @param kernel_size Size of kernel of depthwise post-op\n @param stride_size Size of stride of depthwise post-op\n @param padding_l_size Size of left and top paddings of depthwise post-op\n @returns #dnnl_success on success and a status describing the error\n otherwise"]
4514 pub fn dnnl_post_ops_get_params_dw(
4515 post_ops: const_dnnl_post_ops_t,
4516 index: ::std::os::raw::c_int,
4517 weights_data_type: *mut dnnl_data_type_t::Type,
4518 bias_data_type: *mut dnnl_data_type_t::Type,
4519 dst_data_type: *mut dnnl_data_type_t::Type,
4520 kernel_size: *mut dnnl_dim_t,
4521 stride_size: *mut dnnl_dim_t,
4522 padding_l_size: *mut dnnl_dim_t,
4523 ) -> dnnl_status_t::Type;
4524}
4525unsafe extern "C" {
4526 #[doc = " Appends a binary post-op.\n\n The kind of this post operation is #dnnl_binary.\n\n In the simplest case when the binary is the only post operation, the\n computations would be:\n\n dst[:] <- binary_op (dst[:], another_input[:])\n\n where binary_op is configured with the given parameters. binary_op supports\n broadcast semantics for a second operand.\n\n @param post_ops Post-ops.\n @param alg_kind Binary algorithm for the post-op.\n @param src1_desc Memory descriptor of a second operand.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4527 pub fn dnnl_post_ops_append_binary(
4528 post_ops: dnnl_post_ops_t,
4529 alg_kind: dnnl_alg_kind_t::Type,
4530 src1_desc: const_dnnl_memory_desc_t,
4531 ) -> dnnl_status_t::Type;
4532}
4533unsafe extern "C" {
4534 #[doc = " Returns the parameters of a binary post-op.\n\n @param post_ops Post-ops.\n @param index Index of the binary post-op.\n @param alg_kind Output binary algorithm kind.\n @param src1_desc Output memory descriptor of a second operand.\n @returns #dnnl_success on success and a status describing the error\n otherwise.\n @returns #dnnl_invalid_arguments if @p index does not refer to a binary\n post-op."]
4535 pub fn dnnl_post_ops_get_params_binary(
4536 post_ops: const_dnnl_post_ops_t,
4537 index: ::std::os::raw::c_int,
4538 alg_kind: *mut dnnl_alg_kind_t::Type,
4539 src1_desc: *mut const_dnnl_memory_desc_t,
4540 ) -> dnnl_status_t::Type;
4541}
4542unsafe extern "C" {
4543 #[doc = " Appends a prelu forward post-op.\n\n The kind of this post-op is #dnnl::primitive::kind::prelu.\n\n The post-op can be defined as:\n\n dst[:] <- prelu(dst[:], weights[:])\n prelu:\n dst[:] <- dst[:] if dst[:] > 0\n dst[:] <- dst[:] * weights[:] if dst[:] <= 0\n\n\n @note\n The order of dimensions does not depend on how elements are laid\n out in memory. For example:\n - for a 2D CNN activations tensor the order is always (n, c)\n - for a 4D CNN activations tensor the order is always (n, c, h, w)\n - for a 5D CNN weights tensor the order is always\n (g, oc, ic, kh, kw)\n\n Prelu weights tensor is passed in runtime execution phase. Prelu\n weights tensor data type is implicitly assumed as f32 using plain\n layout (a, ab, acb, acdb, acdeb)\n\n @param post_ops Post-ops.\n @param mask Defines the correspondence between the output tensor\n dimensions and the prelu weights tensor. The set i-th bit indicates\n that a dedicated weights value is used for each index along that\n dimension. Set the mask to 0 to use a common weights value\n for the whole output tensor.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4544 pub fn dnnl_post_ops_append_prelu(
4545 post_ops: dnnl_post_ops_t,
4546 mask: ::std::os::raw::c_int,
4547 ) -> dnnl_status_t::Type;
4548}
4549unsafe extern "C" {
4550 #[doc = " Returns the parameters of a prelu post-op.\n\n @param post_ops Post-ops.\n @param index Index of the prelu post-op.\n @param mask Mask of the prelu post-op.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4551 pub fn dnnl_post_ops_get_params_prelu(
4552 post_ops: const_dnnl_post_ops_t,
4553 index: ::std::os::raw::c_int,
4554 mask: *mut ::std::os::raw::c_int,
4555 ) -> dnnl_status_t::Type;
4556}
4557unsafe extern "C" {
4558 #[doc = " Destroys a memory descriptor.\n\n @param memory_desc Memory descriptor to destroy.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4559 pub fn dnnl_memory_desc_destroy(memory_desc: dnnl_memory_desc_t) -> dnnl_status_t::Type;
4560}
4561unsafe extern "C" {
4562 #[doc = " Clones a memory descriptor. The resulting memory descriptor must be\n destroyed separately.\n\n @param memory_desc Output memory descriptor.\n @param existing_memory_desc Memory descriptor to clone.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4563 pub fn dnnl_memory_desc_clone(
4564 memory_desc: *mut dnnl_memory_desc_t,
4565 existing_memory_desc: const_dnnl_memory_desc_t,
4566 ) -> dnnl_status_t::Type;
4567}
4568unsafe extern "C" {
4569 #[doc = " Retrieves a binary blob associated with the given memory descriptor\n\n @param blob Output pointer to binary blob.\n If not nullptr, size bytes of the memory descriptor blob are written.\n @param size Output pointer to the size of the binary blob in bytes.\n Size is written if blob is nullptr.\n @param memory_desc input memory descriptor to serialize\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4570 pub fn dnnl_memory_desc_get_blob(
4571 blob: *mut u8,
4572 size: *mut usize,
4573 memory_desc: const_dnnl_memory_desc_t,
4574 ) -> dnnl_status_t::Type;
4575}
4576unsafe extern "C" {
4577 #[doc = " Creates a memory descriptor from a memory descriptor binary blob.\n\n @param memory_desc Output pointer to a newly allocated memory descriptor.\n @param blob Pointer to a memory descriptor binary blob.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4578 pub fn dnnl_memory_desc_create_with_blob(
4579 memory_desc: *mut dnnl_memory_desc_t,
4580 blob: *const u8,
4581 ) -> dnnl_status_t::Type;
4582}
4583unsafe extern "C" {
4584 #[doc = " Creates a memory descriptor using dimensions and strides.\n\n @note\n As always, the logical order of dimensions corresponds to the `abc...`\n format tag, and the physical meaning of the dimensions depends on both\n the primitive that consumes the memory and the context of that\n consumption.\n\n @param memory_desc Output memory descriptor.\n @param ndims Number of dimensions\n @param dims Array of dimensions.\n @param data_type Elements data type.\n @param strides Strides in each dimension.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4585 pub fn dnnl_memory_desc_create_with_strides(
4586 memory_desc: *mut dnnl_memory_desc_t,
4587 ndims: ::std::os::raw::c_int,
4588 dims: *const dnnl_dim_t,
4589 data_type: dnnl_data_type_t::Type,
4590 strides: *const dnnl_dim_t,
4591 ) -> dnnl_status_t::Type;
4592}
4593unsafe extern "C" {
4594 #[doc = " Creates a memory descriptor using dimensions and memory format tag.\n\n @note\n As always, the logical order of dimensions corresponds to the `abc...`\n format tag, and the physical meaning of the dimensions depends on both\n the primitive that consumes the memory and the context of that\n consumption.\n\n @param memory_desc Output memory descriptor.\n @param ndims Number of dimensions\n @param dims Array of dimensions.\n @param data_type Elements data type.\n @param tag Memory format tag. Can be #dnnl_format_tag_any which would\n allow a primitive to chose the final memory format. In this case the\n format_kind field of the memory descriptor would be set to\n #dnnl_format_kind_any.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4595 pub fn dnnl_memory_desc_create_with_tag(
4596 memory_desc: *mut dnnl_memory_desc_t,
4597 ndims: ::std::os::raw::c_int,
4598 dims: *const dnnl_dim_t,
4599 data_type: dnnl_data_type_t::Type,
4600 tag: dnnl_format_tag_t::Type,
4601 ) -> dnnl_status_t::Type;
4602}
4603unsafe extern "C" {
4604 #[doc = " @param memory_desc Output memory descriptor.\n @param parent_memory_desc An existing memory descriptor.\n @param dims Sizes of the region.\n @param offsets Offsets to the region from the encompassing\n memory object in each dimension\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4605 pub fn dnnl_memory_desc_create_submemory(
4606 memory_desc: *mut dnnl_memory_desc_t,
4607 parent_memory_desc: const_dnnl_memory_desc_t,
4608 dims: *const dnnl_dim_t,
4609 offsets: *const dnnl_dim_t,
4610 ) -> dnnl_status_t::Type;
4611}
4612unsafe extern "C" {
4613 #[doc = " Creates a memory descriptor by reshaping an existing one. The new\n memory descriptor inherits the data type. This operation is valid only for\n memory descriptors that have format_kind #dnnl_blocked or\n #dnnl_format_kind_any.\n\n The resulting memory descriptor must be destroyed separately.\n\n The operation ensures the transformation of the physical memory format\n corresponds to the transformation of the logical dimensions. If such\n transformation is impossible, the function returns #dnnl_invalid_arguments.\n\n The reshape operation can be described as a combination of the following\n basic operations:\n 1. Add a dimension of size `1`. This is always possible.\n 2. Remove a dimension of size `1`. This is possible only if the dimension\n has no padding (i.e. `padded_dims[dim] == dims[dim] && dims[dim] == 1`).\n 3. Split a dimension into multiple ones. This is possible only if the size\n of the dimension is exactly equal to the product of the split ones and\n the dimension does not have padding (i.e.\n `padded_dims[dim] = dims[dim]`).\n 4. Joining multiple consecutive dimensions into a single one. As in the\n cases above, this requires that the dimensions do not have padding and\n that the memory format is such that in physical memory these dimensions\n are dense and have the same order as their logical counterparts. This\n also assumes that these dimensions are not blocked.\n - Here, dense means:\n `stride for dim[i] == (stride for dim[i + 1]) * dim[i + 1]`;\n - And same order means:\n `i < j` if and only if `stride for dim[j] <= stride for dim[i]`.\n\n @warning\n Some combinations of physical memory layout and/or offsets or\n dimensions may result in a failure to make a reshape.\n\n @param out_memory_desc Output memory descriptor.\n @param in_memory_desc An existing memory descriptor. Must have format_kind\n set to #dnnl_blocked or #dnnl_format_kind_any.\n @param ndims Number of dimensions for the output memory descriptor.\n @param dims Dimensions for the output memory descriptor.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4614 pub fn dnnl_memory_desc_reshape(
4615 out_memory_desc: *mut dnnl_memory_desc_t,
4616 in_memory_desc: const_dnnl_memory_desc_t,
4617 ndims: ::std::os::raw::c_int,
4618 dims: *const dnnl_dim_t,
4619 ) -> dnnl_status_t::Type;
4620}
4621unsafe extern "C" {
4622 #[doc = " Creates a memory descriptor by permuting axes in an existing one.\n\n The physical memory layout representation is adjusted accordingly to\n maintain the consistency between the logical and physical parts of the\n memory descriptor.\n\n The resulting memory descriptor must be destroyed separately.\n\n The new memory descriptor inherits the data type. This operation is valid\n only for memory descriptors that have format_kind set to #dnnl_blocked or\n #dnnl_format_kind_any.\n\n The logical axes will be permuted in the following manner:\n ```\n for (i: 0 .. in_memory_desc->ndims)\n out_memory_desc->dims[permutation[i]] = in_memory_desc->dims[i];\n ```\n\n Example:\n @code\n dnnl_memory_desc_t in_md, out_md, expect_out_md;\n\n const int permutation[] = {1, 0}; // swap the first and the second axes\n\n dnnl_dims_t in_dims = {2, 3}, out_dims = {3, 2};\n dnnl_format_tag_t in_tag = dnnl_ab, out_tag = dnnl_ba;\n\n dnnl_memory_desc_create_with_tag(\n &in_md, 2, in_dims, data_type, in_tag);\n dnnl_memory_desc_create_with_tag(\n &expect_out_md, 2, out_dims, data_type, out_tag);\n\n dnnl_memory_desc_permute_axes(&out_md, in_md, permutation);\n assert(dnnl_memory_desc_equal(out_md, expect_out_md));\n\n dnnl_memory_desc_destroy(in_md);\n dnnl_memory_desc_destroy(out_md);\n dnnl_memory_desc_destroy(expect_out_md);\n @endcode\n\n @param out_memory_desc Output memory descriptor.\n @param in_memory_desc An existing memory descriptor. Must have format_kind\n set to #dnnl_blocked or #dnnl_format_kind_any.\n @param permutation Axes permutation (of size `in_memory_desc->ndims`).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4623 pub fn dnnl_memory_desc_permute_axes(
4624 out_memory_desc: *mut dnnl_memory_desc_t,
4625 in_memory_desc: const_dnnl_memory_desc_t,
4626 permutation: *const ::std::os::raw::c_int,
4627 ) -> dnnl_status_t::Type;
4628}
4629unsafe extern "C" {
4630 #[doc = " Queries a memory descriptor for various pieces of information.\n\n The following information can be queried:\n - Number of dimensions (#dnnl_query_ndims_s32)\n - Dimensions (#dnnl_query_dims) in the following order:\n - CNN data tensors: mini-batch, channel, spatial\n (<code>{N, C, [[D,] H,] W}</code>)\n - CNN weight tensors: group (optional), output channel, input channel,\n spatial (<code>{[G,] O, I, [[D,] H,] W}</code>)\n - RNN data tensors: time, mini-batch, channels (<code>{T, N, C}</code>)\n or layers, directions, states, mini-batch, channels\n (<code>{L, D, S, N, C}</code>)\n - RNN weight tensor: layers, directions, input channel, gates, output\n channels (<code>{L, D, I, G, O}</code>)\n - Data type of the tensor elements (#dnnl_query_data_type)\n - Padded dimensions (#dnnl_query_padded_dims) - size of the data including\n padding in each dimension\n - Padded offsets (#dnnl_query_padded_offsets) - per-dimension offset from\n the padding to actual data, the top-level tensor with offsets applied\n must lie within the padding area.\n - Submemory offset (#dnnl_query_submemory_offset_s64) - offset from memory\n origin to the current block, non-zero only in a description of a memory\n sub-block.\n - Format kind (#dnnl_query_format_kind) - memory format kind\n\n @note\n The order of dimensions does not depend on the memory format, so\n whether the data is laid out in #dnnl_nchw or #dnnl_nhwc\n the dims for 4D CN data tensor would be <code>{N, C, H, W}</code>.\n\n The following queries are applicable only to format kind #dnnl_blocked.\n - Strides (#dnnl_query_strides) between the outermost blocks or in case\n of plain (non-blocked) formats the strides between dimensions\n - Number of innermost blocks (#dnnl_query_inner_nblks_s32), e.g.\n `{4, 16, 4}` in case of `OIhw_4i16o4i`\n - Size of the innermost blocks (#dnnl_query_inner_blks), e.g. 3 in case\n of `OIhw_4i16o4i_`\n - Logical indices of the blocks (#dnnl_query_inner_idxs), e.g. `{1, 0, 1}`\n in case of `4i16o4i`, because `i` is the 1st dim and `o` is the 0st dim\n\n @param memory_desc Memory descriptor.\n @param what Parameter to query.\n @param result Output result. The type depends on the query. For example,\n it must be a @c dnnl_dims_t** if querying for a strides.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4631 pub fn dnnl_memory_desc_query(
4632 memory_desc: const_dnnl_memory_desc_t,
4633 what: dnnl_query_t::Type,
4634 result: *mut ::std::os::raw::c_void,
4635 ) -> dnnl_status_t::Type;
4636}
4637unsafe extern "C" {
4638 #[doc = " Compares two memory descriptors.\n\n Use this function to identify whether a reorder is required between the\n two memories\n\n @param lhs Left-hand side of the comparison.\n @param rhs Right-hand side of the comparison.\n @returns 1 if the descriptors are the same.\n @returns 0 if the descriptors are different."]
4639 pub fn dnnl_memory_desc_equal(
4640 lhs: const_dnnl_memory_desc_t,
4641 rhs: const_dnnl_memory_desc_t,
4642 ) -> ::std::os::raw::c_int;
4643}
4644unsafe extern "C" {
4645 #[doc = " Returns the size of a memory descriptor.\n\n @param memory_desc Memory descriptor.\n @returns The number of bytes required for memory described by a memory\n descriptor."]
4646 pub fn dnnl_memory_desc_get_size(memory_desc: const_dnnl_memory_desc_t) -> usize;
4647}
4648unsafe extern "C" {
4649 #[doc = " Returns the size of data type.\n\n @param data_type Data type.\n @returns The number of bytes occupied by data type."]
4650 pub fn dnnl_data_type_size(data_type: dnnl_data_type_t::Type) -> usize;
4651}
4652unsafe extern "C" {
4653 #[doc = " Creates a memory object.\n\n Unless @p handle is equal to DNNL_MEMORY_NONE, the constructed memory\n object will have the underlying buffer set. In this case, the buffer will\n be initialized as if dnnl_memory_set_data_handle() had been called.\n\n @sa dnnl_memory_set_data_handle()\n\n @param memory Output memory object.\n @param memory_desc Memory descriptor.\n @param engine Engine to use.\n @param handle Handle of the memory buffer to use as an underlying storage.\n - A pointer to the user-allocated buffer. In this case the library\n doesn't own the buffer.\n - The DNNL_MEMORY_ALLOCATE special value. Instructs the library to\n allocate the buffer for the memory object. In this case the library\n owns the buffer.\n - DNNL_MEMORY_NONE to create dnnl_memory without an underlying buffer.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4654 pub fn dnnl_memory_create(
4655 memory: *mut dnnl_memory_t,
4656 memory_desc: const_dnnl_memory_desc_t,
4657 engine: dnnl_engine_t,
4658 handle: *mut ::std::os::raw::c_void,
4659 ) -> dnnl_status_t::Type;
4660}
4661unsafe extern "C" {
4662 #[doc = " Returns the memory descriptor for a memory object.\n\n @param memory Memory object.\n @param memory_desc Output memory descriptor (a copy).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4663 pub fn dnnl_memory_get_memory_desc(
4664 memory: const_dnnl_memory_t,
4665 memory_desc: *mut const_dnnl_memory_desc_t,
4666 ) -> dnnl_status_t::Type;
4667}
4668unsafe extern "C" {
4669 #[doc = " Returns the engine of a memory object.\n\n @param memory Memory object.\n @param engine Output engine on which the memory is located.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4670 pub fn dnnl_memory_get_engine(
4671 memory: const_dnnl_memory_t,
4672 engine: *mut dnnl_engine_t,
4673 ) -> dnnl_status_t::Type;
4674}
4675unsafe extern "C" {
4676 #[doc = " Maps a memory object and returns a host-side pointer to a memory buffer\n with a copy of its contents.\n\n Mapping enables explicit direct access to memory contents for the engines\n that do not support it implicitly.\n\n Mapping is an exclusive operation - a memory object cannot be used in\n other operations until this memory object is unmapped.\n\n @note\n Any primitives working with @p memory should be completed before\n the memory is mapped. Use dnnl_stream_wait to synchronize the\n corresponding execution stream.\n\n @note\n The dnnl_memory_map_data() and dnnl_memory_unmap_data() functions are\n mainly provided for debug and testing purposes, and their performance\n may be suboptimal.\n\n @param memory Memory object.\n @param mapped_ptr Output pointer to the mapped buffer.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4677 pub fn dnnl_memory_map_data(
4678 memory: const_dnnl_memory_t,
4679 mapped_ptr: *mut *mut ::std::os::raw::c_void,
4680 ) -> dnnl_status_t::Type;
4681}
4682unsafe extern "C" {
4683 #[doc = " Unmaps a memory object and writes back any changes made to the previously\n mapped memory buffer. The pointer to the mapped buffer must be obtained\n via the dnnl_memory_map_data() call.\n\n @note\n The dnnl_memory_map_data() and dnnl_memory_unmap_data() functions are\n mainly provided for debug and testing purposes, and their performance\n may be suboptimal.\n\n @param memory Memory object.\n @param mapped_ptr Pointer to the mapped buffer that must have been\n obtained using the dnnl_memory_map_data() function.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4684 pub fn dnnl_memory_unmap_data(
4685 memory: const_dnnl_memory_t,
4686 mapped_ptr: *mut ::std::os::raw::c_void,
4687 ) -> dnnl_status_t::Type;
4688}
4689unsafe extern "C" {
4690 #[doc = " Returns memory object's data handle.\n\n @param memory Memory object.\n @param handle Output data handle. For the CPU engine, the data handle is a\n pointer to the actual data. For OpenCL it is a cl_mem.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4691 pub fn dnnl_memory_get_data_handle(
4692 memory: const_dnnl_memory_t,
4693 handle: *mut *mut ::std::os::raw::c_void,
4694 ) -> dnnl_status_t::Type;
4695}
4696unsafe extern "C" {
4697 #[doc = " Sets the underlying memory buffer.\n\n @param memory Memory object.\n @param handle Data handle. For the CPU engine or when USM is used, the\n memory buffer is a pointer to the actual data. For OpenCL it is a\n `cl_mem`.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4698 pub fn dnnl_memory_set_data_handle(
4699 memory: dnnl_memory_t,
4700 handle: *mut ::std::os::raw::c_void,
4701 ) -> dnnl_status_t::Type;
4702}
4703unsafe extern "C" {
4704 #[doc = " Destroys a memory object.\n\n @param memory Memory object to destroy.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4705 pub fn dnnl_memory_destroy(memory: dnnl_memory_t) -> dnnl_status_t::Type;
4706}
4707unsafe extern "C" {
4708 #[doc = " Creates a primitive descriptor for a reorder primitive.\n\n @param reorder_primitive_desc Output primitive descriptor.\n @param src_desc Source memory descriptor.\n @param src_engine Engine on which the source memory object will be\n located.\n @param dst_desc Destination memory descriptor.\n @param dst_engine Engine on which the destination memory object\n will be located.\n @param attr Primitive attributes to use (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4709 pub fn dnnl_reorder_primitive_desc_create(
4710 reorder_primitive_desc: *mut dnnl_primitive_desc_t,
4711 src_desc: const_dnnl_memory_desc_t,
4712 src_engine: dnnl_engine_t,
4713 dst_desc: const_dnnl_memory_desc_t,
4714 dst_engine: dnnl_engine_t,
4715 attr: const_dnnl_primitive_attr_t,
4716 ) -> dnnl_status_t::Type;
4717}
4718unsafe extern "C" {
4719 #[doc = " Creates a primitive descriptor for an out-of-place concatenation\n primitive.\n\n @param concat_primitive_desc Output primitive descriptor.\n @param dst_desc Destination memory descriptor.\n @param n Number of source parameters.\n @param concat_dimension Source tensors will be concatenated over\n dimension with this index. Note that order of dimensions does\n not depend on memory format.\n @param src_descs Array of source memory descriptors with @p n elements.\n @param attr Primitive attributes to use (can be NULL).\n @param engine Engine to use.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4720 pub fn dnnl_concat_primitive_desc_create(
4721 concat_primitive_desc: *mut dnnl_primitive_desc_t,
4722 engine: dnnl_engine_t,
4723 dst_desc: const_dnnl_memory_desc_t,
4724 n: ::std::os::raw::c_int,
4725 concat_dimension: ::std::os::raw::c_int,
4726 src_descs: *const const_dnnl_memory_desc_t,
4727 attr: const_dnnl_primitive_attr_t,
4728 ) -> dnnl_status_t::Type;
4729}
4730unsafe extern "C" {
4731 #[doc = " Creates a primitive descriptor for an (out-of-place) sum primitive.\n\n @param sum_primitive_desc Output primitive descriptor.\n @param dst_desc Destination memory descriptor.\n @param n Number of source parameters.\n @param scales Vector of scales to multiply data in each source\n memory by.\n @param src_descs Array of source memory descriptors having @p n elements.\n @param attr Primitive attributes to use (can be NULL).\n @param engine Engine to use.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4732 pub fn dnnl_sum_primitive_desc_create(
4733 sum_primitive_desc: *mut dnnl_primitive_desc_t,
4734 engine: dnnl_engine_t,
4735 dst_desc: const_dnnl_memory_desc_t,
4736 n: ::std::os::raw::c_int,
4737 scales: *const f32,
4738 src_descs: *const const_dnnl_memory_desc_t,
4739 attr: const_dnnl_primitive_attr_t,
4740 ) -> dnnl_status_t::Type;
4741}
4742unsafe extern "C" {
4743 #[doc = " Creates a primitive descriptor for a binary primitive.\n\n @note\n Memory descriptors @p src1_desc and @p dst_desc are allowed to be\n initialized with #dnnl_format_tag_any or with format_kind set to\n #dnnl_format_kind_any.\n\n @note\n Both memory descriptors must have the same number of dimensions.\n Element broadcasting is supported for memory descriptor @p src1_desc\n and are applied to @p src1_desc dimensions that have size equal to 1.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param alg_kind Algorithm kind. Valid values are #dnnl_binary_add,\n #dnnl_binary_mul, #dnnl_binary_max, #dnnl_binary_min, #dnnl_binary_div,\n #dnnl_binary_sub, #dnnl_binary_ge, #dnnl_binary_gt, #dnnl_binary_le,\n #dnnl_binary_lt, #dnnl_binary_eq and #dnnl_binary_ne.\n @param src0_desc Source 0 memory descriptor.\n @param src1_desc Source 1 memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4744 pub fn dnnl_binary_primitive_desc_create(
4745 primitive_desc: *mut dnnl_primitive_desc_t,
4746 engine: dnnl_engine_t,
4747 alg_kind: dnnl_alg_kind_t::Type,
4748 src0_desc: const_dnnl_memory_desc_t,
4749 src1_desc: const_dnnl_memory_desc_t,
4750 dst_desc: const_dnnl_memory_desc_t,
4751 attr: const_dnnl_primitive_attr_t,
4752 ) -> dnnl_status_t::Type;
4753}
4754unsafe extern "C" {
4755 #[doc = " Creates a primitive descriptor for a binary primitive with support of\n ternary operators.\n\n @note\n Memory descriptors @p src1_desc, @p src2_desc and @p dst_desc are\n allowed to be initialized with #dnnl_format_tag_any or with format_kind\n set to #dnnl_format_kind_any.\n\n @note\n All memory descriptors must have the same number of dimensions.\n Element broadcasting is supported for memory descriptor @p src1_desc\n and is applied to @p src1_desc dimensions that have a size equal to 1.\n There is no broadcasting support for @p src2_desc.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param alg_kind Algorithm kind.\n @param src0_desc Source 0 memory descriptor.\n @param src1_desc Source 1 memory descriptor.\n @param src2_desc Source memory descriptor for ternary operations. Might\n be empty.\n @param dst_desc Destination memory descriptor.\n @param attr Primitive attributes.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4756 pub fn dnnl_binary_primitive_desc_create_v2(
4757 primitive_desc: *mut dnnl_primitive_desc_t,
4758 engine: dnnl_engine_t,
4759 alg_kind: dnnl_alg_kind_t::Type,
4760 src0_desc: const_dnnl_memory_desc_t,
4761 src1_desc: const_dnnl_memory_desc_t,
4762 src2_desc: const_dnnl_memory_desc_t,
4763 dst_desc: const_dnnl_memory_desc_t,
4764 attr: const_dnnl_primitive_attr_t,
4765 ) -> dnnl_status_t::Type;
4766}
4767unsafe extern "C" {
4768 #[doc = " Creates a primitive descriptor for a convolution forward propagation\n primitive.\n\n @note\n Memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n Arrays @p strides, @p dilates, @p padding_l, and @p padding_r contain\n values for spatial dimensions only and hence must have the same number of\n elements as there are spatial dimensions. The order of values is the same\n as in the tensor: depth (for 3D tensors), height (for 3D and 2D tensors),\n and width.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param alg_kind Convolution algorithm. Possible values are\n #dnnl_convolution_direct, #dnnl_convolution_winograd,\n #dnnl_convolution_auto.\n @param src_desc Source memory descriptor.\n @param weights_desc Weights memory descriptor.\n @param bias_desc Bias memory descriptor. Passing NULL, a zero memory\n descriptor, or a memory descriptor with format_kind set to\n #dnnl_format_kind_undef disables the bias term.\n @param dst_desc Destination memory descriptor.\n @param strides Array of strides for spatial dimension.\n @param dilates Array of dilations for spatial dimension. A zero value\n means no dilation in the corresponding dimension.\n @param padding_l Array of padding values for low indices for each spatial\n dimension `([[front,] top,] left)`.\n @param padding_r Array of padding values for high indices for each spatial\n dimension `([[back,] bottom,] right)`. Can be NULL in which case\n padding is considered to be symmetrical.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4769 pub fn dnnl_convolution_forward_primitive_desc_create(
4770 primitive_desc: *mut dnnl_primitive_desc_t,
4771 engine: dnnl_engine_t,
4772 prop_kind: dnnl_prop_kind_t::Type,
4773 alg_kind: dnnl_alg_kind_t::Type,
4774 src_desc: const_dnnl_memory_desc_t,
4775 weights_desc: const_dnnl_memory_desc_t,
4776 bias_desc: const_dnnl_memory_desc_t,
4777 dst_desc: const_dnnl_memory_desc_t,
4778 strides: *const dnnl_dim_t,
4779 dilates: *const dnnl_dim_t,
4780 padding_l: *const dnnl_dim_t,
4781 padding_r: *const dnnl_dim_t,
4782 attr: const_dnnl_primitive_attr_t,
4783 ) -> dnnl_status_t::Type;
4784}
4785unsafe extern "C" {
4786 #[doc = " Creates a primitive descriptor for a convolution backward propagation\n primitive.\n\n @note\n Memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n Arrays @p strides, @p dilates, @p padding_l, and @p padding_r contain\n values for spatial dimensions only and hence must have the same number of\n elements as there are spatial dimensions. The order of values is the same\n as in the tensor: depth (for 3D tensors), height (for 3D and 2D tensors),\n and width.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param alg_kind Convolution algorithm. Possible values are\n #dnnl_convolution_direct, #dnnl_convolution_winograd,\n #dnnl_convolution_auto.\n @param diff_src_desc Diff source memory descriptor.\n @param weights_desc Weights memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param strides Array of strides for spatial dimension.\n @param dilates Array of dilations for spatial dimension. A zero value\n means no dilation in the corresponding dimension.\n @param padding_l Array of padding values for low indices for each spatial\n dimension `([[front,] top,] left)`.\n @param padding_r Array of padding values for high indices for each spatial\n dimension `([[back,] bottom,] right)`. Can be NULL in which case\n padding is considered to be symmetrical.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4787 pub fn dnnl_convolution_backward_data_primitive_desc_create(
4788 primitive_desc: *mut dnnl_primitive_desc_t,
4789 engine: dnnl_engine_t,
4790 alg_kind: dnnl_alg_kind_t::Type,
4791 diff_src_desc: const_dnnl_memory_desc_t,
4792 weights_desc: const_dnnl_memory_desc_t,
4793 diff_dst_desc: const_dnnl_memory_desc_t,
4794 strides: *const dnnl_dim_t,
4795 dilates: *const dnnl_dim_t,
4796 padding_l: *const dnnl_dim_t,
4797 padding_r: *const dnnl_dim_t,
4798 hint_fwd_pd: const_dnnl_primitive_desc_t,
4799 attr: const_dnnl_primitive_attr_t,
4800 ) -> dnnl_status_t::Type;
4801}
4802unsafe extern "C" {
4803 #[doc = " Creates a primitive descriptor for a convolution weights gradient primitive.\n\n @note\n Memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n Arrays @p strides, @p dilates, @p padding_l, and @p padding_r contain\n values for spatial dimensions only and hence must have the same number of\n elements as there are spatial dimensions. The order of values is the same\n as in the tensor: depth (for 3D tensors), height (for 3D and 2D tensors),\n and width.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param alg_kind Convolution algorithm. Possible values are\n #dnnl_convolution_direct, #dnnl_convolution_winograd,\n #dnnl_convolution_auto.\n @param src_desc Source memory descriptor.\n @param diff_weights_desc Diff weights memory descriptor.\n @param diff_bias_desc Diff bias memory descriptor. Passing NULL, a zero\n memory descriptor, or a memory descriptor with format_kind set to\n #dnnl_format_kind_undef disables the bias term.\n @param diff_dst_desc Diff destination memory descriptor.\n @param strides Array of strides for spatial dimension.\n @param dilates Array of dilations for spatial dimension. A zero value\n means no dilation in the corresponding dimension.\n @param padding_l Array of padding values for low indices for each spatial\n dimension `([[front,] top,] left)`.\n @param padding_r Array of padding values for high indices for each spatial\n dimension `([[back,] bottom,] right)`. Can be NULL in which case\n padding is considered to be symmetrical.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4804 pub fn dnnl_convolution_backward_weights_primitive_desc_create(
4805 primitive_desc: *mut dnnl_primitive_desc_t,
4806 engine: dnnl_engine_t,
4807 alg_kind: dnnl_alg_kind_t::Type,
4808 src_desc: const_dnnl_memory_desc_t,
4809 diff_weights_desc: const_dnnl_memory_desc_t,
4810 diff_bias_desc: const_dnnl_memory_desc_t,
4811 diff_dst_desc: const_dnnl_memory_desc_t,
4812 strides: *const dnnl_dim_t,
4813 dilates: *const dnnl_dim_t,
4814 padding_l: *const dnnl_dim_t,
4815 padding_r: *const dnnl_dim_t,
4816 hint_fwd_pd: const_dnnl_primitive_desc_t,
4817 attr: const_dnnl_primitive_attr_t,
4818 ) -> dnnl_status_t::Type;
4819}
4820unsafe extern "C" {
4821 #[doc = " Creates a primitive descriptor for a deconvolution forward propagation\n primitive.\n\n @note\n Memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n Arrays @p strides, @p dilates, @p padding_l, and @p padding_r contain\n values for spatial dimensions only and hence must have the same number of\n elements as there are spatial dimensions. The order of values is the same\n as in the tensor: depth (for 3D tensors), height (for 3D and 2D tensors),\n and width.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param alg_kind Deconvolution algorithm. Possible values are\n #dnnl_deconvolution_direct, #dnnl_deconvolution_winograd.\n @param src_desc Source memory descriptor.\n @param weights_desc Weights memory descriptor.\n @param bias_desc Bias memory descriptor. Passing NULL, a zero memory\n descriptor, or a memory descriptor with format_kind set to\n #dnnl_format_kind_undef disables the bias term.\n @param dst_desc Destination memory descriptor.\n @param strides Array of strides for spatial dimension.\n @param dilates Array of dilations for spatial dimension. A zero value\n means no dilation in the corresponding dimension.\n @param padding_l Array of padding values for low indices for each spatial\n dimension `([[front,] top,] left)`.\n @param padding_r Array of padding values for high indices for each spatial\n dimension `([[back,] bottom,] right)`. Can be NULL in which case\n padding is considered to be symmetrical.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4822 pub fn dnnl_deconvolution_forward_primitive_desc_create(
4823 primitive_desc: *mut dnnl_primitive_desc_t,
4824 engine: dnnl_engine_t,
4825 prop_kind: dnnl_prop_kind_t::Type,
4826 alg_kind: dnnl_alg_kind_t::Type,
4827 src_desc: const_dnnl_memory_desc_t,
4828 weights_desc: const_dnnl_memory_desc_t,
4829 bias_desc: const_dnnl_memory_desc_t,
4830 dst_desc: const_dnnl_memory_desc_t,
4831 strides: *const dnnl_dim_t,
4832 dilates: *const dnnl_dim_t,
4833 padding_l: *const dnnl_dim_t,
4834 padding_r: *const dnnl_dim_t,
4835 attr: const_dnnl_primitive_attr_t,
4836 ) -> dnnl_status_t::Type;
4837}
4838unsafe extern "C" {
4839 #[doc = " Creates a primitive descriptor for a deconvolution backward propagation\n primitive.\n\n @note\n Memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n Arrays @p strides, @p dilates, @p padding_l, and @p padding_r contain\n values for spatial dimensions only and hence must have the same number of\n elements as there are spatial dimensions. The order of values is the same\n as in the tensor: depth (for 3D tensors), height (for 3D and 2D tensors),\n and width.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param alg_kind Deconvolution algorithm. Possible values are\n #dnnl_deconvolution_direct, #dnnl_deconvolution_winograd.\n @param diff_src_desc Diff source memory descriptor.\n @param weights_desc Weights memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param strides Array of strides for spatial dimension.\n @param dilates Array of dilations for spatial dimension. A zero value\n means no dilation in the corresponding dimension.\n @param padding_l Array of padding values for low indices for each spatial\n dimension `([[front,] top,] left)`.\n @param padding_r Array of padding values for high indices for each spatial\n dimension `([[back,] bottom,] right)`. Can be NULL in which case\n padding is considered to be symmetrical.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4840 pub fn dnnl_deconvolution_backward_data_primitive_desc_create(
4841 primitive_desc: *mut dnnl_primitive_desc_t,
4842 engine: dnnl_engine_t,
4843 alg_kind: dnnl_alg_kind_t::Type,
4844 diff_src_desc: const_dnnl_memory_desc_t,
4845 weights_desc: const_dnnl_memory_desc_t,
4846 diff_dst_desc: const_dnnl_memory_desc_t,
4847 strides: *const dnnl_dim_t,
4848 dilates: *const dnnl_dim_t,
4849 padding_l: *const dnnl_dim_t,
4850 padding_r: *const dnnl_dim_t,
4851 hint_fwd_pd: const_dnnl_primitive_desc_t,
4852 attr: const_dnnl_primitive_attr_t,
4853 ) -> dnnl_status_t::Type;
4854}
4855unsafe extern "C" {
4856 #[doc = " Creates a primitive descriptor for a deconvolution weights gradient\n primitive.\n\n @note\n Memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n Arrays @p strides, @p dilates, @p padding_l, and @p padding_r contain\n values for spatial dimensions only and hence must have the same number of\n elements as there are spatial dimensions. The order of values is the same\n as in the tensor: depth (for 3D tensors), height (for 3D and 2D tensors),\n and width.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param alg_kind Deconvolution algorithm. Possible values are\n #dnnl_deconvolution_direct, #dnnl_deconvolution_winograd.\n @param src_desc Source memory descriptor.\n @param diff_weights_desc Diff weights memory descriptor.\n @param diff_bias_desc Diff bias memory descriptor. Passing NULL, a zero\n memory descriptor, or a memory descriptor with format_kind set to\n #dnnl_format_kind_undef disables the bias term.\n @param diff_dst_desc Diff destination memory descriptor.\n @param strides Array of strides for spatial dimension.\n @param dilates Array of dilations for spatial dimension. A zero value\n means no dilation in the corresponding dimension.\n @param padding_l Array of padding values for low indices for each spatial\n dimension `([[front,] top,] left)`.\n @param padding_r Array of padding values for high indices for each spatial\n dimension `([[back,] bottom,] right)`. Can be NULL in which case\n padding is considered to be symmetrical.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4857 pub fn dnnl_deconvolution_backward_weights_primitive_desc_create(
4858 primitive_desc: *mut dnnl_primitive_desc_t,
4859 engine: dnnl_engine_t,
4860 alg_kind: dnnl_alg_kind_t::Type,
4861 src_desc: const_dnnl_memory_desc_t,
4862 diff_weights_desc: const_dnnl_memory_desc_t,
4863 diff_bias_desc: const_dnnl_memory_desc_t,
4864 diff_dst_desc: const_dnnl_memory_desc_t,
4865 strides: *const dnnl_dim_t,
4866 dilates: *const dnnl_dim_t,
4867 padding_l: *const dnnl_dim_t,
4868 padding_r: *const dnnl_dim_t,
4869 hint_fwd_pd: const_dnnl_primitive_desc_t,
4870 attr: const_dnnl_primitive_attr_t,
4871 ) -> dnnl_status_t::Type;
4872}
4873unsafe extern "C" {
4874 #[doc = " Creates a primitive descriptor for a shuffle forward propagation primitive\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param src_desc Source memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param axis The axis along which the data is shuffled.\n @param group_size Shuffle group size.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4875 pub fn dnnl_shuffle_forward_primitive_desc_create(
4876 primitive_desc: *mut dnnl_primitive_desc_t,
4877 engine: dnnl_engine_t,
4878 prop_kind: dnnl_prop_kind_t::Type,
4879 src_desc: const_dnnl_memory_desc_t,
4880 dst_desc: const_dnnl_memory_desc_t,
4881 axis: ::std::os::raw::c_int,
4882 group_size: dnnl_dim_t,
4883 attr: const_dnnl_primitive_attr_t,
4884 ) -> dnnl_status_t::Type;
4885}
4886unsafe extern "C" {
4887 #[doc = " Creates a primitive descriptor for a shuffle backward propagation primitive\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param diff_src_desc Diff source memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param axis The axis along which the data is shuffled.\n @param group_size Shuffle group size.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4888 pub fn dnnl_shuffle_backward_primitive_desc_create(
4889 primitive_desc: *mut dnnl_primitive_desc_t,
4890 engine: dnnl_engine_t,
4891 diff_src_desc: const_dnnl_memory_desc_t,
4892 diff_dst_desc: const_dnnl_memory_desc_t,
4893 axis: ::std::os::raw::c_int,
4894 group_size: dnnl_dim_t,
4895 hint_fwd_pd: const_dnnl_primitive_desc_t,
4896 attr: const_dnnl_primitive_attr_t,
4897 ) -> dnnl_status_t::Type;
4898}
4899unsafe extern "C" {
4900 #[doc = " Creates a primitive descriptor for an eltwise forward propagation primitive.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param alg_kind Elementwise algorithm kind.\n @param src_desc Source memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param alpha The alpha parameter for the elementwise operation. Specific\n meaning depends on the algorithm.\n @param beta The beta parameter for the elementwise operation. Specific\n meaning depends on the algorithm.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4901 pub fn dnnl_eltwise_forward_primitive_desc_create(
4902 primitive_desc: *mut dnnl_primitive_desc_t,
4903 engine: dnnl_engine_t,
4904 prop_kind: dnnl_prop_kind_t::Type,
4905 alg_kind: dnnl_alg_kind_t::Type,
4906 src_desc: const_dnnl_memory_desc_t,
4907 dst_desc: const_dnnl_memory_desc_t,
4908 alpha: f32,
4909 beta: f32,
4910 attr: const_dnnl_primitive_attr_t,
4911 ) -> dnnl_status_t::Type;
4912}
4913unsafe extern "C" {
4914 #[doc = " Creates a primitive descriptor for an eltwise backward propagation\n primitive.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param alg_kind Elementwise algorithm kind.\n @param diff_src_desc Diff source memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param data_desc Destination memory descriptor if one of the\n \"use_dst_for_bwd\" algorithms are used (such as\n #dnnl_eltwise_relu_use_dst_for_bwd), source memory descriptor otherwise.\n @param alpha The alpha parameter for the elementwise operation. Specific\n meaning depends on the algorithm.\n @param beta The beta parameter for the elementwise operation. Specific\n meaning depends on the algorithm.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4915 pub fn dnnl_eltwise_backward_primitive_desc_create(
4916 primitive_desc: *mut dnnl_primitive_desc_t,
4917 engine: dnnl_engine_t,
4918 alg_kind: dnnl_alg_kind_t::Type,
4919 diff_src_desc: const_dnnl_memory_desc_t,
4920 diff_dst_desc: const_dnnl_memory_desc_t,
4921 data_desc: const_dnnl_memory_desc_t,
4922 alpha: f32,
4923 beta: f32,
4924 hint_fwd_pd: const_dnnl_primitive_desc_t,
4925 attr: const_dnnl_primitive_attr_t,
4926 ) -> dnnl_status_t::Type;
4927}
4928unsafe extern "C" {
4929 #[doc = " Creates a primitive descriptor for a softmax forward propagation primitive.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param alg_kind Softmax algorithm kind: either #dnnl_softmax_accurate, or\n #dnnl_softmax_log.\n @param src_desc Source memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param softmax_axis Axis over which softmax is computed.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4930 pub fn dnnl_softmax_forward_primitive_desc_create(
4931 primitive_desc: *mut dnnl_primitive_desc_t,
4932 engine: dnnl_engine_t,
4933 prop_kind: dnnl_prop_kind_t::Type,
4934 alg_kind: dnnl_alg_kind_t::Type,
4935 src_desc: const_dnnl_memory_desc_t,
4936 dst_desc: const_dnnl_memory_desc_t,
4937 softmax_axis: ::std::os::raw::c_int,
4938 attr: const_dnnl_primitive_attr_t,
4939 ) -> dnnl_status_t::Type;
4940}
4941unsafe extern "C" {
4942 #[doc = " Creates a primitive descriptor for a softmax backward propagation primitive.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param alg_kind Softmax algorithm kind: either #dnnl_softmax_accurate, or\n #dnnl_softmax_log.\n @param diff_src_desc Diff source memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param softmax_axis Axis over which softmax is computed.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4943 pub fn dnnl_softmax_backward_primitive_desc_create(
4944 primitive_desc: *mut dnnl_primitive_desc_t,
4945 engine: dnnl_engine_t,
4946 alg_kind: dnnl_alg_kind_t::Type,
4947 diff_src_desc: const_dnnl_memory_desc_t,
4948 diff_dst_desc: const_dnnl_memory_desc_t,
4949 dst_desc: const_dnnl_memory_desc_t,
4950 softmax_axis: ::std::os::raw::c_int,
4951 hint_fwd_pd: const_dnnl_primitive_desc_t,
4952 attr: const_dnnl_primitive_attr_t,
4953 ) -> dnnl_status_t::Type;
4954}
4955unsafe extern "C" {
4956 #[doc = " Creates a primitive descriptor for a pooling forward propagation\n primitive.\n\n Arrays @p strides, @p kernel, @p dilation, @p padding_l and @p padding_r\n contain values for spatial dimensions only and hence must have the same\n number of elements as there are spatial dimensions. The order of values\n is the same as in the tensor: depth (for 3D tensors),\n height (for 3D and 2D tensors), and width.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param alg_kind Pooling algorithm kind: either #dnnl_pooling_max,\n #dnnl_pooling_avg_include_padding, or #dnnl_pooling_avg_exclude_padding.\n @param src_desc Source memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param strides Array of strides for spatial dimension.\n @param kernel Array of kernel spatial dimensions.\n @param dilation Array of dilations for spatial dimension.\n @param padding_l Array of padding values for low indices for each spatial\n dimension `([[front,] top,] left)`.\n @param padding_r Array of padding values for high indices for each spatial\n dimension `([[back,] bottom,] right)`. Can be NULL in which case\n padding is considered to be symmetrical.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4957 pub fn dnnl_pooling_forward_primitive_desc_create(
4958 primitive_desc: *mut dnnl_primitive_desc_t,
4959 engine: dnnl_engine_t,
4960 prop_kind: dnnl_prop_kind_t::Type,
4961 alg_kind: dnnl_alg_kind_t::Type,
4962 src_desc: const_dnnl_memory_desc_t,
4963 dst_desc: const_dnnl_memory_desc_t,
4964 strides: *const dnnl_dim_t,
4965 kernel: *const dnnl_dim_t,
4966 dilation: *const dnnl_dim_t,
4967 padding_l: *const dnnl_dim_t,
4968 padding_r: *const dnnl_dim_t,
4969 attr: const_dnnl_primitive_attr_t,
4970 ) -> dnnl_status_t::Type;
4971}
4972unsafe extern "C" {
4973 #[doc = " Creates a primitive descriptor for a pooling backward propagation\n primitive.\n\n Arrays @p strides, @p kernel, @p dilation, @p padding_l and @p padding_r\n contain values for spatial dimensions only and hence must have the same\n number of elements as there are spatial dimensions. The order of values\n is the same as in the tensor: depth (for 3D tensors),\n height (for 3D and 2D tensors), and width.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param alg_kind Pooling algorithm kind: either #dnnl_pooling_max,\n #dnnl_pooling_avg_include_padding, or #dnnl_pooling_avg_exclude_padding.\n @param diff_src_desc Diff source memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param strides Array of strides for spatial dimension.\n @param kernel Array of kernel spatial dimensions.\n @param dilation Array of dilations for spatial dimension.\n @param padding_l Array of padding values for low indices for each spatial\n dimension `([[front,] top,] left)`.\n @param padding_r Array of padding values for high indices for each spatial\n dimension `([[back,] bottom,] right)`. Can be NULL in which case\n padding is considered to be symmetrical.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4974 pub fn dnnl_pooling_backward_primitive_desc_create(
4975 primitive_desc: *mut dnnl_primitive_desc_t,
4976 engine: dnnl_engine_t,
4977 alg_kind: dnnl_alg_kind_t::Type,
4978 diff_src_desc: const_dnnl_memory_desc_t,
4979 diff_dst_desc: const_dnnl_memory_desc_t,
4980 strides: *const dnnl_dim_t,
4981 kernel: *const dnnl_dim_t,
4982 dilation: *const dnnl_dim_t,
4983 padding_l: *const dnnl_dim_t,
4984 padding_r: *const dnnl_dim_t,
4985 hint_fwd_pd: const_dnnl_primitive_desc_t,
4986 attr: const_dnnl_primitive_attr_t,
4987 ) -> dnnl_status_t::Type;
4988}
4989unsafe extern "C" {
4990 #[doc = " Creates a primitive descriptor for a PReLU (leaky ReLU with trainable\n alpha parameter) forward propagation primitive.\n\n @note\n weights descriptor is allowed to be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param src_desc Source memory descriptor.\n @param weights_desc Alpha parameters memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
4991 pub fn dnnl_prelu_forward_primitive_desc_create(
4992 primitive_desc: *mut dnnl_primitive_desc_t,
4993 engine: dnnl_engine_t,
4994 prop_kind: dnnl_prop_kind_t::Type,
4995 src_desc: const_dnnl_memory_desc_t,
4996 weights_desc: const_dnnl_memory_desc_t,
4997 dst_desc: const_dnnl_memory_desc_t,
4998 attr: const_dnnl_primitive_attr_t,
4999 ) -> dnnl_status_t::Type;
5000}
5001unsafe extern "C" {
5002 #[doc = " Creates a primitive descriptor for a PReLU (leaky ReLU with trainable\n alpha parameter) backward propagation primitive.\n\n @note\n weights descriptor and diff_weights descriptor are allowed\n to be initialized with #dnnl_format_tag_any or with format_kind\n set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param src_desc Source memory descriptor.\n @param weights_desc Alpha parameters memory descriptor.\n @param diff_src_desc Diff source memory descriptor.\n @param diff_weights_desc Diff alpha parameters memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5003 pub fn dnnl_prelu_backward_primitive_desc_create(
5004 primitive_desc: *mut dnnl_primitive_desc_t,
5005 engine: dnnl_engine_t,
5006 src_desc: const_dnnl_memory_desc_t,
5007 weights_desc: const_dnnl_memory_desc_t,
5008 diff_src_desc: const_dnnl_memory_desc_t,
5009 diff_weights_desc: const_dnnl_memory_desc_t,
5010 diff_dst_desc: const_dnnl_memory_desc_t,
5011 hint_fwd_pd: const_dnnl_primitive_desc_t,
5012 attr: const_dnnl_primitive_attr_t,
5013 ) -> dnnl_status_t::Type;
5014}
5015unsafe extern "C" {
5016 #[doc = " Creates a primitive descriptor for an LRN forward propagation primitive.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param alg_kind LRN algorithm kind: either #dnnl_lrn_across_channels or\n #dnnl_lrn_within_channel.\n @param src_desc Source memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param local_size Regularization local size.\n @param alpha The alpha regularization parameter.\n @param beta The beta regularization parameter.\n @param k The k regularization parameter.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5017 pub fn dnnl_lrn_forward_primitive_desc_create(
5018 primitive_desc: *mut dnnl_primitive_desc_t,
5019 engine: dnnl_engine_t,
5020 prop_kind: dnnl_prop_kind_t::Type,
5021 alg_kind: dnnl_alg_kind_t::Type,
5022 src_desc: const_dnnl_memory_desc_t,
5023 dst_desc: const_dnnl_memory_desc_t,
5024 local_size: dnnl_dim_t,
5025 alpha: f32,
5026 beta: f32,
5027 k: f32,
5028 attr: const_dnnl_primitive_attr_t,
5029 ) -> dnnl_status_t::Type;
5030}
5031unsafe extern "C" {
5032 #[doc = " Creates a primitive descriptor for an LRN backward propagation primitive.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param alg_kind LRN algorithm kind: either #dnnl_lrn_across_channels or\n #dnnl_lrn_within_channel.\n @param diff_src_desc Diff source memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param src_desc Source memory descriptor.\n @param local_size Regularization local size.\n @param alpha The alpha regularization parameter.\n @param beta The beta regularization parameter.\n @param k The k regularization parameter.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5033 pub fn dnnl_lrn_backward_primitive_desc_create(
5034 primitive_desc: *mut dnnl_primitive_desc_t,
5035 engine: dnnl_engine_t,
5036 alg_kind: dnnl_alg_kind_t::Type,
5037 diff_src_desc: const_dnnl_memory_desc_t,
5038 diff_dst_desc: const_dnnl_memory_desc_t,
5039 src_desc: const_dnnl_memory_desc_t,
5040 local_size: dnnl_dim_t,
5041 alpha: f32,
5042 beta: f32,
5043 k: f32,
5044 hint_fwd_pd: const_dnnl_primitive_desc_t,
5045 attr: const_dnnl_primitive_attr_t,
5046 ) -> dnnl_status_t::Type;
5047}
5048unsafe extern "C" {
5049 #[doc = " Creates a primitive descriptor for a batch normalization forward propagation\n primitive.\n\n @note\n In-place operation is supported: the dst can refer to the same memory\n as the src.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param src_desc Source memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param epsilon Batch normalization epsilon parameter.\n @param flags Batch normalization flags (@ref dnnl_normalization_flags_t).\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5050 pub fn dnnl_batch_normalization_forward_primitive_desc_create(
5051 primitive_desc: *mut dnnl_primitive_desc_t,
5052 engine: dnnl_engine_t,
5053 prop_kind: dnnl_prop_kind_t::Type,
5054 src_desc: const_dnnl_memory_desc_t,
5055 dst_desc: const_dnnl_memory_desc_t,
5056 epsilon: f32,
5057 flags: ::std::os::raw::c_uint,
5058 attr: const_dnnl_primitive_attr_t,
5059 ) -> dnnl_status_t::Type;
5060}
5061unsafe extern "C" {
5062 #[doc = " Creates a primitive descriptor for a batch normalization backward\n propagation primitive.\n\n @note\n In-place operation is supported: the diff_dst can refer to the same\n memory as the diff_src.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_backward_data and #dnnl_backward (diffs for all parameters are\n computed in this case).\n @param diff_src_desc Diff source memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param src_desc Source memory descriptor.\n @param epsilon Batch normalization epsilon parameter.\n @param flags Batch normalization flags (@ref dnnl_normalization_flags_t).\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5063 pub fn dnnl_batch_normalization_backward_primitive_desc_create(
5064 primitive_desc: *mut dnnl_primitive_desc_t,
5065 engine: dnnl_engine_t,
5066 prop_kind: dnnl_prop_kind_t::Type,
5067 diff_src_desc: const_dnnl_memory_desc_t,
5068 diff_dst_desc: const_dnnl_memory_desc_t,
5069 src_desc: const_dnnl_memory_desc_t,
5070 epsilon: f32,
5071 flags: ::std::os::raw::c_uint,
5072 hint_fwd_pd: const_dnnl_primitive_desc_t,
5073 attr: const_dnnl_primitive_attr_t,
5074 ) -> dnnl_status_t::Type;
5075}
5076unsafe extern "C" {
5077 #[doc = " Creates a primitive descriptor for a group normalization forward propagation\n primitive.\n\n @note\n In-place operation is supported: the dst can refer to the same memory\n as the src.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param src_desc Source memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param groups Group normalization groups parameter.\n @param epsilon Group normalization epsilon parameter.\n @param flags Group normalization flags (@ref dnnl_normalization_flags_t).\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5078 pub fn dnnl_group_normalization_forward_primitive_desc_create(
5079 primitive_desc: *mut dnnl_primitive_desc_t,
5080 engine: dnnl_engine_t,
5081 prop_kind: dnnl_prop_kind_t::Type,
5082 src_desc: const_dnnl_memory_desc_t,
5083 dst_desc: const_dnnl_memory_desc_t,
5084 groups: dnnl_dim_t,
5085 epsilon: f32,
5086 flags: ::std::os::raw::c_uint,
5087 attr: const_dnnl_primitive_attr_t,
5088 ) -> dnnl_status_t::Type;
5089}
5090unsafe extern "C" {
5091 #[doc = " Creates a primitive descriptor for a group normalization backward\n propagation primitive.\n\n @note\n In-place operation is supported: the diff_dst can refer to the same\n memory as the diff_src.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_backward_data and #dnnl_backward (diffs for all parameters are\n computed in this case).\n @param diff_src_desc Diff source memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param src_desc Source memory descriptor.\n @param groups Group normalization groups parameter.\n @param epsilon Group normalization epsilon parameter.\n @param flags Group normalization flags (@ref dnnl_normalization_flags_t).\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5092 pub fn dnnl_group_normalization_backward_primitive_desc_create(
5093 primitive_desc: *mut dnnl_primitive_desc_t,
5094 engine: dnnl_engine_t,
5095 prop_kind: dnnl_prop_kind_t::Type,
5096 diff_src_desc: const_dnnl_memory_desc_t,
5097 diff_dst_desc: const_dnnl_memory_desc_t,
5098 src_desc: const_dnnl_memory_desc_t,
5099 groups: dnnl_dim_t,
5100 epsilon: f32,
5101 flags: ::std::os::raw::c_uint,
5102 hint_fwd_pd: const_dnnl_primitive_desc_t,
5103 attr: const_dnnl_primitive_attr_t,
5104 ) -> dnnl_status_t::Type;
5105}
5106unsafe extern "C" {
5107 #[doc = " Creates a primitive descriptor for a layer normalization forward propagation\n primitive.\n\n @note\n In-place operation is supported: the dst can refer to the same memory\n as the src.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param src_desc Source memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param stat_desc Memory descriptor for mean and variance. If this\n parameter is NULL, a zero memory descriptor, or a memory descriptor\n with format_kind set to #dnnl_format_kind_undef, then the memory\n descriptor for stats is derived from @p src_desc by removing the last\n dimension.\n @param epsilon Layer normalization epsilon parameter.\n @param flags Layer normalization flags (@ref dnnl_normalization_flags_t).\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5108 pub fn dnnl_layer_normalization_forward_primitive_desc_create(
5109 primitive_desc: *mut dnnl_primitive_desc_t,
5110 engine: dnnl_engine_t,
5111 prop_kind: dnnl_prop_kind_t::Type,
5112 src_desc: const_dnnl_memory_desc_t,
5113 dst_desc: const_dnnl_memory_desc_t,
5114 stat_desc: const_dnnl_memory_desc_t,
5115 epsilon: f32,
5116 flags: ::std::os::raw::c_uint,
5117 attr: const_dnnl_primitive_attr_t,
5118 ) -> dnnl_status_t::Type;
5119}
5120unsafe extern "C" {
5121 #[doc = " Creates a primitive descriptor for a layer normalization backward\n propagation primitive.\n\n @note\n In-place operation is supported: the diff_dst can refer to the same\n memory as the diff_src.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_backward_data and #dnnl_backward (diffs for all parameters are\n computed in this case).\n @param diff_src_desc Diff source memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param src_desc Source memory descriptor.\n @param stat_desc Memory descriptor for mean and variance. If this\n parameter is NULL, a zero memory descriptor, or a memory descriptor\n with format_kind set to #dnnl_format_kind_undef, then the memory\n descriptor for stats is derived from @p src_desc by removing the last\n dimension.\n @param epsilon Layer normalization epsilon parameter.\n @param flags Layer normalization flags (@ref dnnl_normalization_flags_t).\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5122 pub fn dnnl_layer_normalization_backward_primitive_desc_create(
5123 primitive_desc: *mut dnnl_primitive_desc_t,
5124 engine: dnnl_engine_t,
5125 prop_kind: dnnl_prop_kind_t::Type,
5126 diff_src_desc: const_dnnl_memory_desc_t,
5127 diff_dst_desc: const_dnnl_memory_desc_t,
5128 src_desc: const_dnnl_memory_desc_t,
5129 stat_desc: const_dnnl_memory_desc_t,
5130 epsilon: f32,
5131 flags: ::std::os::raw::c_uint,
5132 hint_fwd_pd: const_dnnl_primitive_desc_t,
5133 attr: const_dnnl_primitive_attr_t,
5134 ) -> dnnl_status_t::Type;
5135}
5136unsafe extern "C" {
5137 #[doc = " Creates a primitive descriptor for a layer normalization forward propagation\n primitive with a user-provided data type for the scale and shift\n memory objects.\n\n @note\n In-place operation is supported: the dst can refer to the same memory\n as the src.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param src_desc Source memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param stat_desc Memory descriptor for mean and variance. If this\n parameter is NULL, a zero memory descriptor, or a memory descriptor\n with format_kind set to #dnnl_format_kind_undef, then the memory\n descriptor for stats is derived from @p src_desc by removing the last\n dimension.\n @param scale_shift_data_type Data type of scale and shift memory. If neither scale\n nor shift flag are specified the parameter is ignored.\n @param epsilon Layer normalization epsilon parameter.\n @param flags Layer normalization flags (@ref dnnl_normalization_flags_t).\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5138 pub fn dnnl_layer_normalization_forward_primitive_desc_create_v2(
5139 primitive_desc: *mut dnnl_primitive_desc_t,
5140 engine: dnnl_engine_t,
5141 prop_kind: dnnl_prop_kind_t::Type,
5142 src_desc: const_dnnl_memory_desc_t,
5143 dst_desc: const_dnnl_memory_desc_t,
5144 stat_desc: const_dnnl_memory_desc_t,
5145 scale_shift_data_type: dnnl_data_type_t::Type,
5146 epsilon: f32,
5147 flags: ::std::os::raw::c_uint,
5148 attr: const_dnnl_primitive_attr_t,
5149 ) -> dnnl_status_t::Type;
5150}
5151unsafe extern "C" {
5152 #[doc = " Creates a primitive descriptor for a layer normalization backward\n propagation primitive with a user-provided data type for the\n scale and shift memory objects.\n\n @note\n In-place operation is supported: the diff_dst can refer to the same\n memory as the diff_src.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_backward_data and #dnnl_backward (diffs for all parameters are\n computed in this case).\n @param diff_src_desc Diff source memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param src_desc Source memory descriptor.\n @param stat_desc Memory descriptor for mean and variance. If this\n parameter is NULL, a zero memory descriptor, or a memory descriptor\n with format_kind set to #dnnl_format_kind_undef, then the memory\n descriptor for stats is derived from @p src_desc by removing the last\n dimension.\n @param diff_scale_shift_data_type Data type of diff scale and shift memory. If neither scale\n nor shift flag are specified the parameter is ignored.\n @param scale_shift_data_type Data type of scale and shift memory. If neither scale\n nor shift flag are specified the parameter is ignored.\n @param epsilon Layer normalization epsilon parameter.\n @param flags Layer normalization flags (@ref dnnl_normalization_flags_t).\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5153 pub fn dnnl_layer_normalization_backward_primitive_desc_create_v2(
5154 primitive_desc: *mut dnnl_primitive_desc_t,
5155 engine: dnnl_engine_t,
5156 prop_kind: dnnl_prop_kind_t::Type,
5157 diff_src_desc: const_dnnl_memory_desc_t,
5158 diff_dst_desc: const_dnnl_memory_desc_t,
5159 src_desc: const_dnnl_memory_desc_t,
5160 stat_desc: const_dnnl_memory_desc_t,
5161 diff_scale_shift_data_type: dnnl_data_type_t::Type,
5162 scale_shift_data_type: dnnl_data_type_t::Type,
5163 epsilon: f32,
5164 flags: ::std::os::raw::c_uint,
5165 hint_fwd_pd: const_dnnl_primitive_desc_t,
5166 attr: const_dnnl_primitive_attr_t,
5167 ) -> dnnl_status_t::Type;
5168}
5169unsafe extern "C" {
5170 #[doc = " Creates a primitive descriptor for an inner product forward propagation\n primitive.\n\n @note\n Memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param src_desc Source memory descriptor.\n @param weights_desc Weights memory descriptor.\n @param bias_desc Bias memory descriptor. Passing NULL, a zero memory\n descriptor, or a memory descriptor with format_kind set to\n #dnnl_format_kind_undef disables the bias term.\n @param dst_desc Destination memory descriptor.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5171 pub fn dnnl_inner_product_forward_primitive_desc_create(
5172 primitive_desc: *mut dnnl_primitive_desc_t,
5173 engine: dnnl_engine_t,
5174 prop_kind: dnnl_prop_kind_t::Type,
5175 src_desc: const_dnnl_memory_desc_t,
5176 weights_desc: const_dnnl_memory_desc_t,
5177 bias_desc: const_dnnl_memory_desc_t,
5178 dst_desc: const_dnnl_memory_desc_t,
5179 attr: const_dnnl_primitive_attr_t,
5180 ) -> dnnl_status_t::Type;
5181}
5182unsafe extern "C" {
5183 #[doc = " Creates a primitive descriptor for an inner product backward propagation\n primitive.\n\n @note\n Memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param diff_src_desc Diff source memory descriptor.\n @param weights_desc Weights memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5184 pub fn dnnl_inner_product_backward_data_primitive_desc_create(
5185 primitive_desc: *mut dnnl_primitive_desc_t,
5186 engine: dnnl_engine_t,
5187 diff_src_desc: const_dnnl_memory_desc_t,
5188 weights_desc: const_dnnl_memory_desc_t,
5189 diff_dst_desc: const_dnnl_memory_desc_t,
5190 hint_fwd_pd: const_dnnl_primitive_desc_t,
5191 attr: const_dnnl_primitive_attr_t,
5192 ) -> dnnl_status_t::Type;
5193}
5194unsafe extern "C" {
5195 #[doc = " Creates a primitive descriptor for an inner product weights gradient\n primitive.\n\n @note\n Memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive_descriptor.\n @param engine Engine to use.\n @param src_desc Source memory descriptor.\n @param diff_weights_desc Diff weights memory descriptor.\n @param diff_bias_desc Diff bias memory descriptor. Passing NULL, a zero\n memory descriptor, or a memory descriptor with format_kind set to\n #dnnl_format_kind_undef disables the bias term.\n @param diff_dst_desc Diff destination memory descriptor.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5196 pub fn dnnl_inner_product_backward_weights_primitive_desc_create(
5197 primitive_desc: *mut dnnl_primitive_desc_t,
5198 engine: dnnl_engine_t,
5199 src_desc: const_dnnl_memory_desc_t,
5200 diff_weights_desc: const_dnnl_memory_desc_t,
5201 diff_bias_desc: const_dnnl_memory_desc_t,
5202 diff_dst_desc: const_dnnl_memory_desc_t,
5203 hint_fwd_pd: const_dnnl_primitive_desc_t,
5204 attr: const_dnnl_primitive_attr_t,
5205 ) -> dnnl_status_t::Type;
5206}
5207unsafe extern "C" {
5208 #[doc = " Set quantization scale and shift parameters for RNN data tensors.\n\n For performance reasons, the low-precision configuration of the RNN\n primitives expects input activations to have the unsigned 8-bit integer\n data type. The scale and shift parameters are used to quantize\n floating-point data to unsigned integer and must be passed to the RNN\n primitive using attributes.\n\n The quantization formula is `scale * data + shift`.\n\n @note\n Quantization scale and shift are common for src_layer, src_iter,\n dst_iter, and dst_layer.\n\n Example usage:\n @code\n // RNN parameters\n int l = 2, t = 2, mb = 32, sic = 32, slc = 32, dic = 32, dlc = 32;\n // Activations quantization parameters\n float scale = 63.f, shift = 64.f;\n\n dnnl_primitive_attr_t rnn_attr;\n // Create default attributes\n dnnl_primitive_attr_create(&rnn_attr);\n\n // Set scale and shift for int8 quantization of activation\n dnnl_primitive_attr_set_rnn_data_qparams(rnn_attr, scale, shift);\n\n // Create an RNN primitive descriptor.\n dnnl_primitive_desc_t rnn_pd;\n dnnl_vanilla_rnn_forward_primitive_desc_create(&rnn_pd,\n engine, /* arguments */, attr);\n @endcode\n\n @param attr Primitive attributes.\n @param scale The value to scale the data by.\n @param shift The value to shift the data by.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5209 pub fn dnnl_primitive_attr_set_rnn_data_qparams(
5210 attr: dnnl_primitive_attr_t,
5211 scale: f32,
5212 shift: f32,
5213 ) -> dnnl_status_t::Type;
5214}
5215unsafe extern "C" {
5216 #[doc = " Returns the quantization scale and shift parameters for RNN data tensors.\n\n @note\n Quantization scale and shift are common for src_layer, src_iter,\n dst_iter, and dst_layer.\n\n @param attr Primitive attributes.\n @param scale The value to scale the data by.\n @param shift The value to shift the data by.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5217 pub fn dnnl_primitive_attr_get_rnn_data_qparams(
5218 attr: const_dnnl_primitive_attr_t,
5219 scale: *mut f32,
5220 shift: *mut f32,
5221 ) -> dnnl_status_t::Type;
5222}
5223unsafe extern "C" {
5224 #[doc = " Sets quantization scaling factors for RNN weights tensors. The\n low-precision configuration of the RNN primitives expects input weights to\n use the signed 8-bit integer data type. The scaling factors are used to\n quantize floating-point data to signed integer and must be passed to RNN\n primitives using attributes.\n\n @note\n The dimension order is always native and does not depend on the actual\n layout used. For example, five-dimensional weights always have (l, d,\n i, g, o) logical dimension ordering.\n\n @note\n Quantization scales are common for weights_layer and weights_iteration\n\n @param attr Primitive attributes.\n @param count Number of elements in the @p scales array.\n @param mask Scaling factors correspondence mask that defines the\n correspondence between the output tensor dimensions and the @p\n scales vector. The set i-th bit indicates that a dedicated scaling\n factor should be used for each index along that dimension. Set the\n mask to 0 to use a common scaling factor for the whole output\n tensor.\n @param scales Array of output scaling factors that must contain @p count\n values and the following equality must hold:\n \\f[count = \\prod\\limits_{d \\in mask} weights.dims[d].\\f]\n Violations can only be detected when the attributes are used to create\n a primitive descriptor.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5225 pub fn dnnl_primitive_attr_set_rnn_weights_qparams(
5226 attr: dnnl_primitive_attr_t,
5227 count: dnnl_dim_t,
5228 mask: ::std::os::raw::c_int,
5229 scales: *const f32,
5230 ) -> dnnl_status_t::Type;
5231}
5232unsafe extern "C" {
5233 #[doc = " Returns the quantization scaling factors for RNN weights tensors.\n\n @param attr Primitive attributes.\n @param count Number of elements in the @p scales array.\n @param mask Scaling factors correspondence mask that defines the\n correspondence between the output tensor dimensions and the @p\n scales vector. The set i-th bit indicates that a dedicated scaling\n factor should be used for each index along that dimension. Set the\n mask to 0 to use a common scaling factor for the whole output\n tensor.\n @param scales Array of output scaling factors that contain @p count\n values and the following equality must hold:\n \\f[count = \\prod\\limits_{d \\in mask} weights.dims[d].\\f]\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5234 pub fn dnnl_primitive_attr_get_rnn_weights_qparams(
5235 attr: const_dnnl_primitive_attr_t,
5236 count: *mut dnnl_dim_t,
5237 mask: *mut ::std::os::raw::c_int,
5238 scales: *mut *const f32,
5239 ) -> dnnl_status_t::Type;
5240}
5241unsafe extern "C" {
5242 #[doc = " Sets quantization scaling factors for RNN projection weights tensors. The\n low-precision configuration of the RNN primitives expects input weights to\n use the signed 8-bit integer data type. The scaling factors are used to\n quantize floating-point data to signed integer and must be passed to RNN\n primitives using attributes.\n\n @note\n The dimension order is always native and does not depend on the actual\n layout used. For example, five-dimensional weights always have (l, d,\n i, g, o) logical dimension ordering.\n\n @param attr Primitive attributes.\n @param count Number of elements in the @p scales array.\n @param mask Scaling factors correspondence mask that defines the\n correspondence between the output tensor dimensions and the @p\n scales vector. The set i-th bit indicates that a dedicated scaling\n factor should be used for each index along that dimension. Set the\n mask to 0 to use a common scaling factor for the whole output\n tensor.\n @param scales Array of output scaling factors that must contain @p count\n values and the following equality must hold:\n \\f[count = \\prod\\limits_{d \\in mask} weights.dims[d].\\f]\n Violations can only be detected when the attributes are used to create\n a primitive descriptor.\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5243 pub fn dnnl_primitive_attr_set_rnn_weights_projection_qparams(
5244 attr: dnnl_primitive_attr_t,
5245 count: dnnl_dim_t,
5246 mask: ::std::os::raw::c_int,
5247 scales: *const f32,
5248 ) -> dnnl_status_t::Type;
5249}
5250unsafe extern "C" {
5251 #[doc = " Returns the quantization scaling factors for RNN projection weights tensors.\n\n @param attr Primitive attributes.\n @param count Number of elements in the @p scales array.\n @param mask Scaling factors correspondence mask that defines the\n correspondence between the output tensor dimensions and the @p\n scales vector. The set i-th bit indicates that a dedicated scaling\n factor should be used for each index along that dimension. Set the\n mask to 0 to use a common scaling factor for the whole output\n tensor.\n @param scales Array of output scaling factors that contain @p count\n values and the following equality must hold:\n \\f[count = \\prod\\limits_{d \\in mask} weights.dims[d].\\f]\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5252 pub fn dnnl_primitive_attr_get_rnn_weights_projection_qparams(
5253 attr: const_dnnl_primitive_attr_t,
5254 count: *mut dnnl_dim_t,
5255 mask: *mut ::std::os::raw::c_int,
5256 scales: *mut *const f32,
5257 ) -> dnnl_status_t::Type;
5258}
5259unsafe extern "C" {
5260 #[doc = " Creates a primitive descriptor for vanilla RNN forward propagation\n primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc,\n - @p bias_desc,\n - @p dst_iter_desc.\n\n This would then indicate that the RNN forward propagation primitive should\n not use them and should default to zero values instead.\n\n @note\n All memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param activation Activation kind. Possible values are #dnnl_eltwise_relu,\n #dnnl_eltwise_tanh or #dnnl_eltwise_logistic.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param flags Unused.\n @param alpha Negative slope if activation is #dnnl_eltwise_relu.\n @param beta Unused.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5261 pub fn dnnl_vanilla_rnn_forward_primitive_desc_create(
5262 primitive_desc: *mut dnnl_primitive_desc_t,
5263 engine: dnnl_engine_t,
5264 prop_kind: dnnl_prop_kind_t::Type,
5265 activation: dnnl_alg_kind_t::Type,
5266 direction: dnnl_rnn_direction_t::Type,
5267 src_layer_desc: const_dnnl_memory_desc_t,
5268 src_iter_desc: const_dnnl_memory_desc_t,
5269 weights_layer_desc: const_dnnl_memory_desc_t,
5270 weights_iter_desc: const_dnnl_memory_desc_t,
5271 bias_desc: const_dnnl_memory_desc_t,
5272 dst_layer_desc: const_dnnl_memory_desc_t,
5273 dst_iter_desc: const_dnnl_memory_desc_t,
5274 flags: ::std::os::raw::c_uint,
5275 alpha: f32,
5276 beta: f32,
5277 attr: const_dnnl_primitive_attr_t,
5278 ) -> dnnl_status_t::Type;
5279}
5280unsafe extern "C" {
5281 #[doc = " Creates a primitive descriptor for vanilla RNN backward propagation\n primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc together with @p diff_src_iter_desc,\n - @p bias_desc together with @p diff_bias_desc,\n - @p dst_iter_desc together with @p diff_dst_iter_desc.\n\n This would then indicate that the RNN backward propagation primitive should\n not use the respective data and should use zero values instead.\n\n @note\n All memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Must be #dnnl_backward.\n @param activation Activation kind. Possible values are #dnnl_eltwise_relu,\n #dnnl_eltwise_tanh or #dnnl_eltwise_logistic.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param diff_src_layer_desc Memory descriptor for the diff of input vector.\n @param diff_src_iter_desc Memory descriptor for the diff of input recurrent\n hidden state vector.\n @param diff_weights_layer_desc Memory descriptor for the diff of weights\n applied to the layer input.\n @param diff_weights_iter_desc Memory descriptor for the diff of weights\n applied to the recurrent input.\n @param diff_bias_desc Diff bias memory descriptor.\n @param diff_dst_layer_desc Memory descriptor for the diff of output\n vector.\n @param diff_dst_iter_desc Memory descriptor for the diff of output\n recurrent hidden state vector.\n @param flags Unused.\n @param alpha Negative slope if activation is #dnnl_eltwise_relu.\n @param beta Unused.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5282 pub fn dnnl_vanilla_rnn_backward_primitive_desc_create(
5283 primitive_desc: *mut dnnl_primitive_desc_t,
5284 engine: dnnl_engine_t,
5285 prop_kind: dnnl_prop_kind_t::Type,
5286 activation: dnnl_alg_kind_t::Type,
5287 direction: dnnl_rnn_direction_t::Type,
5288 src_layer_desc: const_dnnl_memory_desc_t,
5289 src_iter_desc: const_dnnl_memory_desc_t,
5290 weights_layer_desc: const_dnnl_memory_desc_t,
5291 weights_iter_desc: const_dnnl_memory_desc_t,
5292 bias_desc: const_dnnl_memory_desc_t,
5293 dst_layer_desc: const_dnnl_memory_desc_t,
5294 dst_iter_desc: const_dnnl_memory_desc_t,
5295 diff_src_layer_desc: const_dnnl_memory_desc_t,
5296 diff_src_iter_desc: const_dnnl_memory_desc_t,
5297 diff_weights_layer_desc: const_dnnl_memory_desc_t,
5298 diff_weights_iter_desc: const_dnnl_memory_desc_t,
5299 diff_bias_desc: const_dnnl_memory_desc_t,
5300 diff_dst_layer_desc: const_dnnl_memory_desc_t,
5301 diff_dst_iter_desc: const_dnnl_memory_desc_t,
5302 flags: ::std::os::raw::c_uint,
5303 alpha: f32,
5304 beta: f32,
5305 hint_fwd_pd: const_dnnl_primitive_desc_t,
5306 attr: const_dnnl_primitive_attr_t,
5307 ) -> dnnl_status_t::Type;
5308}
5309unsafe extern "C" {
5310 #[doc = " Creates a primitive descriptor for an LSTM forward propagation primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc together with @p src_iter_c_desc,\n - @p weights_peephole_desc,\n - @p bias_desc,\n - @p dst_iter_desc together with @p dst_iter_c_desc.\n\n This would then indicate that the LSTM forward propagation primitive should\n not use them and should default to zero values instead.\n\n The @p weights_projection_desc could either be @c NULL or point to a zero\n memory descriptor. This would then indicate that the LSTM doesn't have\n recurrent projection layer.\n\n @note\n All memory descriptors can be initialized with #dnnl_format_tag_any or\n with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param src_iter_c_desc Memory descriptor for the input recurrent cell\n state vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param weights_peephole_desc Memory descriptor for the weights applied to\n the cell states (according to the Peephole LSTM formula).\n @param weights_projection_desc Memory descriptor for the weights applied to\n the hidden states to get the recurrent projection (according to the\n Projection LSTM formula).\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param dst_iter_c_desc Memory descriptor for the output recurrent cell\n state vector.\n @param flags Unused.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5311 pub fn dnnl_lstm_forward_primitive_desc_create(
5312 primitive_desc: *mut dnnl_primitive_desc_t,
5313 engine: dnnl_engine_t,
5314 prop_kind: dnnl_prop_kind_t::Type,
5315 direction: dnnl_rnn_direction_t::Type,
5316 src_layer_desc: const_dnnl_memory_desc_t,
5317 src_iter_desc: const_dnnl_memory_desc_t,
5318 src_iter_c_desc: const_dnnl_memory_desc_t,
5319 weights_layer_desc: const_dnnl_memory_desc_t,
5320 weights_iter_desc: const_dnnl_memory_desc_t,
5321 weights_peephole_desc: const_dnnl_memory_desc_t,
5322 weights_projection_desc: const_dnnl_memory_desc_t,
5323 bias_desc: const_dnnl_memory_desc_t,
5324 dst_layer_desc: const_dnnl_memory_desc_t,
5325 dst_iter_desc: const_dnnl_memory_desc_t,
5326 dst_iter_c_desc: const_dnnl_memory_desc_t,
5327 flags: ::std::os::raw::c_uint,
5328 attr: const_dnnl_primitive_attr_t,
5329 ) -> dnnl_status_t::Type;
5330}
5331unsafe extern "C" {
5332 #[doc = " Creates a primitive descriptor for an LSTM backward propagation primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc together with @p src_iter_c_desc, @p diff_src_iter_desc,\n and @p diff_src_iter_c_desc,\n - @p weights_peephole_desc together with @p diff_weights_peephole_desc,\n - @p bias_desc together with @p diff_bias_desc,\n - @p dst_iter_desc together with @p dst_iter_c_desc, @p diff_dst_iter_desc,\n and @p diff_dst_iter_c_desc.\n\n This would then indicate that the LSTM backward propagation primitive\n should not use them and should default to zero values instead.\n\n The @p weights_projection_desc together with @p\n diff_weights_projection_desc could either be @c NULL or point to a zero\n memory descriptor. This would then indicate that the LSTM doesn't have\n recurrent projection layer.\n\n @note\n All memory descriptors can be initialized with #dnnl_format_tag_any or\n with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Must be #dnnl_backward.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param src_iter_c_desc Memory descriptor for the input recurrent cell\n state vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param weights_peephole_desc Memory descriptor for the weights applied to\n the cell states (according to the Peephole LSTM formula).\n @param weights_projection_desc Memory descriptor for the weights applied to\n the hidden states to get the recurrent projection (according to the\n Projection LSTM formula).\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param dst_iter_c_desc Memory descriptor for the output recurrent cell\n state vector.\n @param diff_src_layer_desc Memory descriptor for the diff of input vector.\n @param diff_src_iter_desc Memory descriptor for the diff of input recurrent\n hidden state vector.\n @param diff_src_iter_c_desc Memory descriptor for the diff of input\n recurrent cell state vector.\n @param diff_weights_layer_desc Memory descriptor for the diff of weights\n applied to the layer input.\n @param diff_weights_iter_desc Memory descriptor for the diff of weights\n applied to the recurrent input.\n @param diff_weights_peephole_desc Memory descriptor for the diff of weights\n applied to the cell states (according to the Peephole LSTM formula).\n @param diff_weights_projection_desc Memory descriptor for the diff of\n weights applied to the hidden states to get the recurrent projection\n (according to the Projection LSTM formula).\n @param diff_bias_desc Diff bias memory descriptor.\n @param diff_dst_layer_desc Memory descriptor for the diff of output\n vector.\n @param diff_dst_iter_desc Memory descriptor for the diff of output\n recurrent hidden state vector.\n @param diff_dst_iter_c_desc Memory descriptor for the diff of output\n recurrent cell state vector.\n @param flags Unused.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5333 pub fn dnnl_lstm_backward_primitive_desc_create(
5334 primitive_desc: *mut dnnl_primitive_desc_t,
5335 engine: dnnl_engine_t,
5336 prop_kind: dnnl_prop_kind_t::Type,
5337 direction: dnnl_rnn_direction_t::Type,
5338 src_layer_desc: const_dnnl_memory_desc_t,
5339 src_iter_desc: const_dnnl_memory_desc_t,
5340 src_iter_c_desc: const_dnnl_memory_desc_t,
5341 weights_layer_desc: const_dnnl_memory_desc_t,
5342 weights_iter_desc: const_dnnl_memory_desc_t,
5343 weights_peephole_desc: const_dnnl_memory_desc_t,
5344 weights_projection_desc: const_dnnl_memory_desc_t,
5345 bias_desc: const_dnnl_memory_desc_t,
5346 dst_layer_desc: const_dnnl_memory_desc_t,
5347 dst_iter_desc: const_dnnl_memory_desc_t,
5348 dst_iter_c_desc: const_dnnl_memory_desc_t,
5349 diff_src_layer_desc: const_dnnl_memory_desc_t,
5350 diff_src_iter_desc: const_dnnl_memory_desc_t,
5351 diff_src_iter_c_desc: const_dnnl_memory_desc_t,
5352 diff_weights_layer_desc: const_dnnl_memory_desc_t,
5353 diff_weights_iter_desc: const_dnnl_memory_desc_t,
5354 diff_weights_peephole_desc: const_dnnl_memory_desc_t,
5355 diff_weights_projection_desc: const_dnnl_memory_desc_t,
5356 diff_bias_desc: const_dnnl_memory_desc_t,
5357 diff_dst_layer_desc: const_dnnl_memory_desc_t,
5358 diff_dst_iter_desc: const_dnnl_memory_desc_t,
5359 diff_dst_iter_c_desc: const_dnnl_memory_desc_t,
5360 flags: ::std::os::raw::c_uint,
5361 hint_fwd_pd: const_dnnl_primitive_desc_t,
5362 attr: const_dnnl_primitive_attr_t,
5363 ) -> dnnl_status_t::Type;
5364}
5365unsafe extern "C" {
5366 #[doc = " Creates a primitive descriptor for GRU forward propagation primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc,\n - @p bias_desc,\n - @p dst_iter_desc.\n\n This would then indicate that the GRU forward propagation primitive should\n not use them and should default to zero values instead.\n\n @note\n All memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param flags Unused.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5367 pub fn dnnl_gru_forward_primitive_desc_create(
5368 primitive_desc: *mut dnnl_primitive_desc_t,
5369 engine: dnnl_engine_t,
5370 prop_kind: dnnl_prop_kind_t::Type,
5371 direction: dnnl_rnn_direction_t::Type,
5372 src_layer_desc: const_dnnl_memory_desc_t,
5373 src_iter_desc: const_dnnl_memory_desc_t,
5374 weights_layer_desc: const_dnnl_memory_desc_t,
5375 weights_iter_desc: const_dnnl_memory_desc_t,
5376 bias_desc: const_dnnl_memory_desc_t,
5377 dst_layer_desc: const_dnnl_memory_desc_t,
5378 dst_iter_desc: const_dnnl_memory_desc_t,
5379 flags: ::std::os::raw::c_uint,
5380 attr: const_dnnl_primitive_attr_t,
5381 ) -> dnnl_status_t::Type;
5382}
5383unsafe extern "C" {
5384 #[doc = " Creates a primitive descriptor for GRU backward propagation primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc together with @p diff_src_iter_desc,\n - @p bias_desc together with @p diff_bias_desc,\n - @p dst_iter_desc together with @p diff_dst_iter_desc.\n\n This would then indicate that the GRU backward propagation primitive\n should not use them and should default to zero values instead.\n\n @note\n All memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Must be #dnnl_backward.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param diff_src_layer_desc Memory descriptor for the diff of input vector.\n @param diff_src_iter_desc Memory descriptor for the diff of input recurrent\n hidden state vector.\n @param diff_weights_layer_desc Memory descriptor for the diff of weights\n applied to the layer input.\n @param diff_weights_iter_desc Memory descriptor for the diff of weights\n applied to the recurrent input.\n @param diff_bias_desc Diff bias memory descriptor.\n @param diff_dst_layer_desc Memory descriptor for the diff of output\n vector.\n @param diff_dst_iter_desc Memory descriptor for the diff of output\n recurrent hidden state vector.\n @param flags Unused.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5385 pub fn dnnl_gru_backward_primitive_desc_create(
5386 primitive_desc: *mut dnnl_primitive_desc_t,
5387 engine: dnnl_engine_t,
5388 prop_kind: dnnl_prop_kind_t::Type,
5389 direction: dnnl_rnn_direction_t::Type,
5390 src_layer_desc: const_dnnl_memory_desc_t,
5391 src_iter_desc: const_dnnl_memory_desc_t,
5392 weights_layer_desc: const_dnnl_memory_desc_t,
5393 weights_iter_desc: const_dnnl_memory_desc_t,
5394 bias_desc: const_dnnl_memory_desc_t,
5395 dst_layer_desc: const_dnnl_memory_desc_t,
5396 dst_iter_desc: const_dnnl_memory_desc_t,
5397 diff_src_layer_desc: const_dnnl_memory_desc_t,
5398 diff_src_iter_desc: const_dnnl_memory_desc_t,
5399 diff_weights_layer_desc: const_dnnl_memory_desc_t,
5400 diff_weights_iter_desc: const_dnnl_memory_desc_t,
5401 diff_bias_desc: const_dnnl_memory_desc_t,
5402 diff_dst_layer_desc: const_dnnl_memory_desc_t,
5403 diff_dst_iter_desc: const_dnnl_memory_desc_t,
5404 flags: ::std::os::raw::c_uint,
5405 hint_fwd_pd: const_dnnl_primitive_desc_t,
5406 attr: const_dnnl_primitive_attr_t,
5407 ) -> dnnl_status_t::Type;
5408}
5409unsafe extern "C" {
5410 #[doc = " Creates a descriptor for LBR GRU forward propagation primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc,\n - @p bias_desc,\n - @p dst_iter_desc.\n\n This would then indicate that the LBR GRU forward propagation primitive\n should not use them and should default to zero values instead.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param flags Unused.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5411 pub fn dnnl_lbr_gru_forward_primitive_desc_create(
5412 primitive_desc: *mut dnnl_primitive_desc_t,
5413 engine: dnnl_engine_t,
5414 prop_kind: dnnl_prop_kind_t::Type,
5415 direction: dnnl_rnn_direction_t::Type,
5416 src_layer_desc: const_dnnl_memory_desc_t,
5417 src_iter_desc: const_dnnl_memory_desc_t,
5418 weights_layer_desc: const_dnnl_memory_desc_t,
5419 weights_iter_desc: const_dnnl_memory_desc_t,
5420 bias_desc: const_dnnl_memory_desc_t,
5421 dst_layer_desc: const_dnnl_memory_desc_t,
5422 dst_iter_desc: const_dnnl_memory_desc_t,
5423 flags: ::std::os::raw::c_uint,
5424 attr: const_dnnl_primitive_attr_t,
5425 ) -> dnnl_status_t::Type;
5426}
5427unsafe extern "C" {
5428 #[doc = " Creates a primitive descriptor for LBR GRU backward propagation primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc together with @p diff_src_iter_desc,\n - @p bias_desc together with @p diff_bias_desc,\n - @p dst_iter_desc together with @p diff_dst_iter_desc.\n\n This would then indicate that the LBR GRU backward propagation primitive\n should not use them and should default to zero values instead.\n\n @note\n All memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Must be #dnnl_backward.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param diff_src_layer_desc Memory descriptor for the diff of input vector.\n @param diff_src_iter_desc Memory descriptor for the diff of input recurrent\n hidden state vector.\n @param diff_weights_layer_desc Memory descriptor for the diff of weights\n applied to the layer input.\n @param diff_weights_iter_desc Memory descriptor for the diff of weights\n applied to the recurrent input.\n @param diff_bias_desc Diff bias memory descriptor.\n @param diff_dst_layer_desc Memory descriptor for the diff of output\n vector.\n @param diff_dst_iter_desc Memory descriptor for the diff of output\n recurrent hidden state vector.\n @param flags Unused.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5429 pub fn dnnl_lbr_gru_backward_primitive_desc_create(
5430 primitive_desc: *mut dnnl_primitive_desc_t,
5431 engine: dnnl_engine_t,
5432 prop_kind: dnnl_prop_kind_t::Type,
5433 direction: dnnl_rnn_direction_t::Type,
5434 src_layer_desc: const_dnnl_memory_desc_t,
5435 src_iter_desc: const_dnnl_memory_desc_t,
5436 weights_layer_desc: const_dnnl_memory_desc_t,
5437 weights_iter_desc: const_dnnl_memory_desc_t,
5438 bias_desc: const_dnnl_memory_desc_t,
5439 dst_layer_desc: const_dnnl_memory_desc_t,
5440 dst_iter_desc: const_dnnl_memory_desc_t,
5441 diff_src_layer_desc: const_dnnl_memory_desc_t,
5442 diff_src_iter_desc: const_dnnl_memory_desc_t,
5443 diff_weights_layer_desc: const_dnnl_memory_desc_t,
5444 diff_weights_iter_desc: const_dnnl_memory_desc_t,
5445 diff_bias_desc: const_dnnl_memory_desc_t,
5446 diff_dst_layer_desc: const_dnnl_memory_desc_t,
5447 diff_dst_iter_desc: const_dnnl_memory_desc_t,
5448 flags: ::std::os::raw::c_uint,
5449 hint_fwd_pd: const_dnnl_primitive_desc_t,
5450 attr: const_dnnl_primitive_attr_t,
5451 ) -> dnnl_status_t::Type;
5452}
5453unsafe extern "C" {
5454 #[doc = " Creates a primitive descriptor for AUGRU forward propagation primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc,\n - @p bias_desc,\n - @p dst_iter_desc.\n\n This would then indicate that the AUGRU forward propagation primitive should\n not use them and should default to zero values instead.\n\n @note\n All memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param attention_desc Memory descriptor for the attention vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param flags Unused.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5455 pub fn dnnl_augru_forward_primitive_desc_create(
5456 primitive_desc: *mut dnnl_primitive_desc_t,
5457 engine: dnnl_engine_t,
5458 prop_kind: dnnl_prop_kind_t::Type,
5459 direction: dnnl_rnn_direction_t::Type,
5460 src_layer_desc: const_dnnl_memory_desc_t,
5461 src_iter_desc: const_dnnl_memory_desc_t,
5462 attention_desc: const_dnnl_memory_desc_t,
5463 weights_layer_desc: const_dnnl_memory_desc_t,
5464 weights_iter_desc: const_dnnl_memory_desc_t,
5465 bias_desc: const_dnnl_memory_desc_t,
5466 dst_layer_desc: const_dnnl_memory_desc_t,
5467 dst_iter_desc: const_dnnl_memory_desc_t,
5468 flags: ::std::os::raw::c_uint,
5469 attr: const_dnnl_primitive_attr_t,
5470 ) -> dnnl_status_t::Type;
5471}
5472unsafe extern "C" {
5473 #[doc = " Creates a primitive descriptor for AUGRU backward propagation primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc together with @p diff_src_iter_desc,\n - @p bias_desc together with @p diff_bias_desc,\n - @p dst_iter_desc together with @p diff_dst_iter_desc.\n\n This would then indicate that the AUGRU backward propagation primitive\n should not use them and should default to zero values instead.\n\n @note\n All memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Must be #dnnl_backward.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param attention_desc Memory descriptor for the attention vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param diff_src_layer_desc Memory descriptor for the diff of input vector.\n @param diff_src_iter_desc Memory descriptor for the diff of input recurrent\n hidden state vector.\n @param diff_attention_desc Memory descriptor for the diff of attention vector.\n @param diff_weights_layer_desc Memory descriptor for the diff of weights\n applied to the layer input.\n @param diff_weights_iter_desc Memory descriptor for the diff of weights\n applied to the recurrent input.\n @param diff_bias_desc Diff bias memory descriptor.\n @param diff_dst_layer_desc Memory descriptor for the diff of output\n vector.\n @param diff_dst_iter_desc Memory descriptor for the diff of output\n recurrent hidden state vector.\n @param flags Unused.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5474 pub fn dnnl_augru_backward_primitive_desc_create(
5475 primitive_desc: *mut dnnl_primitive_desc_t,
5476 engine: dnnl_engine_t,
5477 prop_kind: dnnl_prop_kind_t::Type,
5478 direction: dnnl_rnn_direction_t::Type,
5479 src_layer_desc: const_dnnl_memory_desc_t,
5480 src_iter_desc: const_dnnl_memory_desc_t,
5481 attention_desc: const_dnnl_memory_desc_t,
5482 weights_layer_desc: const_dnnl_memory_desc_t,
5483 weights_iter_desc: const_dnnl_memory_desc_t,
5484 bias_desc: const_dnnl_memory_desc_t,
5485 dst_layer_desc: const_dnnl_memory_desc_t,
5486 dst_iter_desc: const_dnnl_memory_desc_t,
5487 diff_src_layer_desc: const_dnnl_memory_desc_t,
5488 diff_src_iter_desc: const_dnnl_memory_desc_t,
5489 diff_attention_desc: const_dnnl_memory_desc_t,
5490 diff_weights_layer_desc: const_dnnl_memory_desc_t,
5491 diff_weights_iter_desc: const_dnnl_memory_desc_t,
5492 diff_bias_desc: const_dnnl_memory_desc_t,
5493 diff_dst_layer_desc: const_dnnl_memory_desc_t,
5494 diff_dst_iter_desc: const_dnnl_memory_desc_t,
5495 flags: ::std::os::raw::c_uint,
5496 hint_fwd_pd: const_dnnl_primitive_desc_t,
5497 attr: const_dnnl_primitive_attr_t,
5498 ) -> dnnl_status_t::Type;
5499}
5500unsafe extern "C" {
5501 #[doc = " Creates a primitive descriptor for LBR AUGRU forward propagation primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc,\n - @p bias_desc,\n - @p dst_iter_desc.\n\n This would then indicate that the LBR AUGRU forward propagation primitive\n should not use them and should default to zero values instead.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param attention_desc Memory descriptor for the attention vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param flags Unused.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5502 pub fn dnnl_lbr_augru_forward_primitive_desc_create(
5503 primitive_desc: *mut dnnl_primitive_desc_t,
5504 engine: dnnl_engine_t,
5505 prop_kind: dnnl_prop_kind_t::Type,
5506 direction: dnnl_rnn_direction_t::Type,
5507 src_layer_desc: const_dnnl_memory_desc_t,
5508 src_iter_desc: const_dnnl_memory_desc_t,
5509 attention_desc: const_dnnl_memory_desc_t,
5510 weights_layer_desc: const_dnnl_memory_desc_t,
5511 weights_iter_desc: const_dnnl_memory_desc_t,
5512 bias_desc: const_dnnl_memory_desc_t,
5513 dst_layer_desc: const_dnnl_memory_desc_t,
5514 dst_iter_desc: const_dnnl_memory_desc_t,
5515 flags: ::std::os::raw::c_uint,
5516 attr: const_dnnl_primitive_attr_t,
5517 ) -> dnnl_status_t::Type;
5518}
5519unsafe extern "C" {
5520 #[doc = " Creates a primitive descriptor for LBR AUGRU backward propagation primitive.\n\n The following arguments may either be @c NULL or point to a zero memory\n descriptor:\n - @p src_iter_desc together with @p diff_src_iter_desc,\n - @p bias_desc together with @p diff_bias_desc,\n - @p dst_iter_desc together with @p diff_dst_iter_desc.\n\n This would then indicate that the LBR AUGRU backward propagation primitive\n should not use them and should default to zero values instead.\n\n @note\n All memory descriptors can be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Must be #dnnl_backward.\n @param direction RNN direction. See @ref dnnl_rnn_direction_t for more\n info.\n @param src_layer_desc Memory descriptor for the input vector.\n @param src_iter_desc Memory descriptor for the input recurrent hidden\n state vector.\n @param attention_desc Memory descriptor for the attention vector.\n @param weights_layer_desc Memory descriptor for the weights applied to the\n layer input.\n @param weights_iter_desc Memory descriptor for the weights applied to the\n recurrent input.\n @param bias_desc Bias memory descriptor.\n @param dst_layer_desc Memory descriptor for the output vector.\n @param dst_iter_desc Memory descriptor for the output recurrent hidden\n state vector.\n @param diff_src_layer_desc Memory descriptor for the diff of input vector.\n @param diff_src_iter_desc Memory descriptor for the diff of input recurrent\n hidden state vector.\n @param diff_attention_desc Memory descriptor for the diff of attention vector.\n @param diff_weights_layer_desc Memory descriptor for the diff of weights\n applied to the layer input.\n @param diff_weights_iter_desc Memory descriptor for the diff of weights\n applied to the recurrent input.\n @param diff_bias_desc Diff bias memory descriptor.\n @param diff_dst_layer_desc Memory descriptor for the diff of output\n vector.\n @param diff_dst_iter_desc Memory descriptor for the diff of output\n recurrent hidden state vector.\n @param flags Unused.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5521 pub fn dnnl_lbr_augru_backward_primitive_desc_create(
5522 primitive_desc: *mut dnnl_primitive_desc_t,
5523 engine: dnnl_engine_t,
5524 prop_kind: dnnl_prop_kind_t::Type,
5525 direction: dnnl_rnn_direction_t::Type,
5526 src_layer_desc: const_dnnl_memory_desc_t,
5527 src_iter_desc: const_dnnl_memory_desc_t,
5528 attention_desc: const_dnnl_memory_desc_t,
5529 weights_layer_desc: const_dnnl_memory_desc_t,
5530 weights_iter_desc: const_dnnl_memory_desc_t,
5531 bias_desc: const_dnnl_memory_desc_t,
5532 dst_layer_desc: const_dnnl_memory_desc_t,
5533 dst_iter_desc: const_dnnl_memory_desc_t,
5534 diff_src_layer_desc: const_dnnl_memory_desc_t,
5535 diff_src_iter_desc: const_dnnl_memory_desc_t,
5536 diff_attention_desc: const_dnnl_memory_desc_t,
5537 diff_weights_layer_desc: const_dnnl_memory_desc_t,
5538 diff_weights_iter_desc: const_dnnl_memory_desc_t,
5539 diff_bias_desc: const_dnnl_memory_desc_t,
5540 diff_dst_layer_desc: const_dnnl_memory_desc_t,
5541 diff_dst_iter_desc: const_dnnl_memory_desc_t,
5542 flags: ::std::os::raw::c_uint,
5543 hint_fwd_pd: const_dnnl_primitive_desc_t,
5544 attr: const_dnnl_primitive_attr_t,
5545 ) -> dnnl_status_t::Type;
5546}
5547unsafe extern "C" {
5548 #[doc = " Creates a primitive descriptor for a matrix multiplication primitive.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param src_desc Source memory descriptor (matrix A)\n @param weights_desc Weights memory descriptor (matrix B)\n @param bias_desc Bias memory descriptor. Passing NULL, a zero memory\n descriptor, or a memory descriptor with format_kind set to\n #dnnl_format_kind_undef disables the bias term.\n @param dst_desc Destination memory descriptor (matrix C).\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5549 pub fn dnnl_matmul_primitive_desc_create(
5550 primitive_desc: *mut dnnl_primitive_desc_t,
5551 engine: dnnl_engine_t,
5552 src_desc: const_dnnl_memory_desc_t,
5553 weights_desc: const_dnnl_memory_desc_t,
5554 bias_desc: const_dnnl_memory_desc_t,
5555 dst_desc: const_dnnl_memory_desc_t,
5556 attr: const_dnnl_primitive_attr_t,
5557 ) -> dnnl_status_t::Type;
5558}
5559unsafe extern "C" {
5560 #[doc = " Creates a primitive descriptor for a resampling forward propagation\n primitive.\n\n @note\n Destination memory descriptor is allowed to be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param prop_kind Propagation kind. Possible values are\n #dnnl_forward_training and #dnnl_forward_inference.\n @param alg_kind resampling algorithm kind: either #dnnl_resampling_nearest,\n or #dnnl_resampling_linear.\n @param factors Array of scaling factors for spatial dimension.\n @param src_desc Source memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5561 pub fn dnnl_resampling_forward_primitive_desc_create(
5562 primitive_desc: *mut dnnl_primitive_desc_t,
5563 engine: dnnl_engine_t,
5564 prop_kind: dnnl_prop_kind_t::Type,
5565 alg_kind: dnnl_alg_kind_t::Type,
5566 factors: *const f32,
5567 src_desc: const_dnnl_memory_desc_t,
5568 dst_desc: const_dnnl_memory_desc_t,
5569 attr: const_dnnl_primitive_attr_t,
5570 ) -> dnnl_status_t::Type;
5571}
5572unsafe extern "C" {
5573 #[doc = " Creates a primitive descriptor for a resampling backward propagation\n primitive.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param alg_kind resamplinging algorithm kind: either\n #dnnl_resampling_nearest, or #dnnl_resampling_linear.\n @param diff_src_desc Diff source memory descriptor.\n @param diff_dst_desc Diff destination memory descriptor.\n @param factors Array of scaling factors for spatial dimension.\n @param hint_fwd_pd Primitive descriptor for a respective forward propagation\n primitive.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise.\n"]
5574 pub fn dnnl_resampling_backward_primitive_desc_create(
5575 primitive_desc: *mut dnnl_primitive_desc_t,
5576 engine: dnnl_engine_t,
5577 alg_kind: dnnl_alg_kind_t::Type,
5578 factors: *const f32,
5579 diff_src_desc: const_dnnl_memory_desc_t,
5580 diff_dst_desc: const_dnnl_memory_desc_t,
5581 hint_fwd_pd: const_dnnl_primitive_desc_t,
5582 attr: const_dnnl_primitive_attr_t,
5583 ) -> dnnl_status_t::Type;
5584}
5585unsafe extern "C" {
5586 #[doc = " Creates a primitive descriptor for a reduction primitive.\n\n @note\n Destination memory descriptor is allowed to be initialized with\n #dnnl_format_tag_any or with format_kind set to #dnnl_format_kind_any.\n\n @param primitive_desc Output primitive descriptor.\n @param engine Engine to use.\n @param alg_kind reduction algorithm kind. Possible values:\n #dnnl_reduction_max, #dnnl_reduction_min, #dnnl_reduction_sum,\n #dnnl_reduction_mul, #dnnl_reduction_mean, #dnnl_reduction_norm_lp_max,\n #dnnl_reduction_norm_lp_sum, #dnnl_reduction_norm_lp_power_p_max,\n #dnnl_reduction_norm_lp_power_p_sum.\n @param p Algorithm specific parameter.\n @param eps Algorithm specific parameter.\n @param src_desc Source memory descriptor.\n @param dst_desc Destination memory descriptor.\n @param attr Primitive attributes (can be NULL).\n @returns #dnnl_success on success and a status describing the error\n otherwise."]
5587 pub fn dnnl_reduction_primitive_desc_create(
5588 primitive_desc: *mut dnnl_primitive_desc_t,
5589 engine: dnnl_engine_t,
5590 alg_kind: dnnl_alg_kind_t::Type,
5591 src_desc: const_dnnl_memory_desc_t,
5592 dst_desc: const_dnnl_memory_desc_t,
5593 p: f32,
5594 eps: f32,
5595 attr: const_dnnl_primitive_attr_t,
5596 ) -> dnnl_status_t::Type;
5597}
5598unsafe extern "C" {
5599 #[doc = " Returns the number of primitives that can be held in the primitive cache\n at the same time.\n\n @param capacity Primitive cache capacity to query. Concurrently\n accessing @p capacity is safe.\n @returns #dnnl_invalid_arguments/#dnnl::status::invalid_arguments if the\n @p capacity value is invalid, and #dnnl_success/#dnnl::status::success on\n success."]
5600 pub fn dnnl_get_primitive_cache_capacity(
5601 capacity: *mut ::std::os::raw::c_int,
5602 ) -> dnnl_status_t::Type;
5603}
5604unsafe extern "C" {
5605 #[doc = " Sets a number of primitives that can be held in the primitive cache\n at a time.\n\n @param capacity Primitive cache capacity to set. If a new @p capacity is\n less than a number of primitives that the primitive cache already has\n then the excess entries will be evicted. Setting the @p capacity to 0\n clears the primitive cache and disables it. Concurrently modifying\n @p capacity is safe.\n @returns #dnnl_invalid_arguments/#dnnl::status::invalid_arguments if the\n @p capacity value is invalid, and #dnnl_success/#dnnl::status::success on\n success."]
5606 pub fn dnnl_set_primitive_cache_capacity(
5607 capacity: ::std::os::raw::c_int,
5608 ) -> dnnl_status_t::Type;
5609}
5610unsafe extern "C" {
5611 #[doc = " Configures dumping of JIT-generated code.\n\n @note\n This setting overrides the DNNL_JIT_DUMP environment variable.\n\n @param enable Flag value. Set to 0 to disable and set to 1 to enable.\n @returns #dnnl_invalid_arguments/#dnnl::status::invalid_arguments if the\n @p flag value is invalid, and #dnnl_success/#dnnl::status::success on\n success."]
5612 pub fn dnnl_set_jit_dump(enable: ::std::os::raw::c_int) -> dnnl_status_t::Type;
5613}
5614unsafe extern "C" {
5615 #[doc = " Sets library profiling flags. The flags define which profilers are\n supported.\n\n @note\n This setting overrides DNNL_JIT_PROFILE environment variable.\n\n @sa @ref dev_guide_profilers\n\n @param flags Profiling flags that can contain the following bits:\n - @ref DNNL_JIT_PROFILE_VTUNE -- integration with VTune Profiler\n (on by default)\n - @ref DNNL_JIT_PROFILE_LINUX_JITDUMP -- produce Linux-specific\n jit-pid.dump output (off by default). The location of the output\n is controlled via JITDUMPDIR environment variable or via\n dnnl_set_jit_profiling_jitdumpdir() function.\n - @ref DNNL_JIT_PROFILE_LINUX_PERFMAP -- produce Linux-specific\n perf-pid.map output (off by default). The output is always placed\n into /tmp.\n\n Passing @ref DNNL_JIT_PROFILE_NONE disables profiling completely.\n\n @returns #dnnl_invalid_arguments/#dnnl::status::invalid_arguments if the\n @p flags value is invalid, and #dnnl_success/#dnnl::status::success on\n success."]
5616 pub fn dnnl_set_jit_profiling_flags(flags: ::std::os::raw::c_uint) -> dnnl_status_t::Type;
5617}
5618unsafe extern "C" {
5619 #[doc = " Sets JIT dump output path. Only applicable to Linux and is only\n used when profiling flags have DNNL_JIT_PROFILE_LINUX_PERF bit set.\n\n After the first JIT kernel is generated, the jitdump output will be placed\n into temporary directory created using the mkdtemp template\n 'dir/.debug/jit/dnnl.XXXXXX'.\n\n @sa @ref dev_guide_profilers\n\n @note\n This setting overrides JITDUMPDIR environment variable. If\n JITDUMPDIR is not set, and this function is never called, the path\n defaults to HOME. Passing NULL reverts the value to default.\n\n @note\n The directory is accessed only when the first JIT kernel is being\n created. JIT profiling will be disabled in case of any errors\n accessing or creating this directory.\n\n @param dir JIT dump output path.\n @returns #dnnl_success/#dnnl::status::success if the\n output directory was set correctly and an error status otherwise.\n @returns #dnnl_unimplemented/#dnnl::status::unimplemented on Windows."]
5620 pub fn dnnl_set_jit_profiling_jitdumpdir(
5621 dir: *const ::std::os::raw::c_char,
5622 ) -> dnnl_status_t::Type;
5623}
5624unsafe extern "C" {
5625 #[doc = " Sets the maximal ISA the library can dispatch to on the CPU. See\n #dnnl_cpu_isa_t and #dnnl::cpu_isa for the list of the values accepted by\n the C and C++ API functions respectively.\n\n This function has effect only once, and returns an error on subsequent\n calls. It should also be invoked before any other oneDNN API call, otherwise\n it may return an error.\n\n This function overrides the DNNL_MAX_CPU_ISA environment variable. The\n environment variable can be set to the desired maximal ISA name in upper\n case and with dnnl_cpu_isa prefix removed. For example:\n `DNNL_MAX_CPU_ISA=AVX2`.\n\n @note\n The ISAs are only partially ordered:\n - SSE41 < AVX < AVX2 < AVX2_VNNI < AVX2_VNNI_2,\n - AVX2 < AVX512_CORE < AVX512_CORE_VNNI < AVX512_CORE_BF16\n < AVX10_1_512 < AVX10_1_512_AMX < AVX10_1_512_AMX_FP16,\n - AVX2_VNNI < AVX10_1_512.\n Aliases:\n - AVX512_CORE_FP16 = AVX10_1_512\n - AVX512_CORE_AMX = AVX10_1_512_AMX\n - AVX512_CORE_AMX_FP16 = AVX10_1_512_AMX_FP16\n\n @sa @ref dev_guide_cpu_dispatcher_control for more details\n\n @param isa Maximal ISA the library should dispatch to. Pass\n #dnnl_cpu_isa_default/#dnnl::cpu_isa::isa_default to remove ISA restrictions\n (except for ISAs with initial support in the library).\n @returns #dnnl_success/#dnnl::status::success on success and a\n #dnnl_invalid_arguments/#dnnl::status::invalid_arguments if the @p isa\n parameter is invalid or the ISA cannot be changed at this time.\n @returns #dnnl_unimplemented/#dnnl::status::unimplemented if the feature\n was disabled at build time (see @ref dev_guide_build_options for more\n details)."]
5626 pub fn dnnl_set_max_cpu_isa(isa: dnnl_cpu_isa_t::Type) -> dnnl_status_t::Type;
5627}
5628unsafe extern "C" {
5629 #[doc = " Gets the maximal ISA the library can dispatch to on the CPU. See\n #dnnl_cpu_isa_t and #dnnl::cpu_isa for the list of the values returned by\n the C and C++ API functions respectively.\n\n @sa @ref dev_guide_cpu_dispatcher_control for more details\n\n @returns #dnnl_cpu_isa_t value reflecting the maximal ISA the library may\n dispatch to."]
5630 pub fn dnnl_get_effective_cpu_isa() -> dnnl_cpu_isa_t::Type;
5631}
5632unsafe extern "C" {
5633 #[doc = " Sets the hints flag for the CPU ISA. See #dnnl_cpu_isa_hints_t and\n #dnnl::cpu_isa_hints for the list of the values accepted by the C and C++\n API functions respectively.\n\n This function has effect only once, and returns an error on subsequent\n calls. It should also be invoked before any other oneDNN API call, otherwise\n it may return an error.\n\n This function overrides the DNNL_CPU_ISA_HINTS environment variable.\n @sa @ref dev_guide_cpu_isa_hints for more details\n\n @param isa_hints CPU ISA hints to be passed over to the implementation.\n Pass #dnnl_cpu_isa_no_hints/#dnnl::cpu_isa_hints::no_hints to use\n default features i.e. no hints.\n @returns #dnnl_success/#dnnl::status::success on success and a\n #dnnl_runtime_error/#dnnl::status::runtime_error if the ISA hints cannot\n be specified at the current time.\n @returns #dnnl_unimplemented/#dnnl::status::unimplemented if the feature\n was disabled at build time (see @ref dev_guide_build_options for more\n details)."]
5634 pub fn dnnl_set_cpu_isa_hints(isa_hints: dnnl_cpu_isa_hints_t::Type) -> dnnl_status_t::Type;
5635}
5636unsafe extern "C" {
5637 #[doc = " Gets the ISA specific hints that library can follow. See\n #dnnl_cpu_isa_hints_t and #dnnl::cpu_isa_hints for the list of the values\n returned by the C and C++ API functions respectively.\n\n @sa @ref dev_guide_cpu_isa_hints for more details\n\n @returns #dnnl_cpu_isa_hints_t value reflecting the ISA specific hints the\n library can follow."]
5638 pub fn dnnl_get_cpu_isa_hints() -> dnnl_cpu_isa_hints_t::Type;
5639}
5640unsafe extern "C" {
5641 #[doc = " Performs single-precision matrix-matrix multiply.\n\n The operation is defined as:\n\n `C := alpha * op( A ) * op( B ) + beta * C`\n\n where\n - `op( X ) = X` or `op( X ) = X**T`,\n - `alpha` and `beta` are scalars, and\n - `A`, `B`, and `C` are matrices:\n - `op( A )` is an `MxK` matrix,\n - `op( B )` is an `KxN` matrix,\n - `C` is an `MxN` matrix.\n\n The matrices are assumed to be stored in row-major order (the elements in\n each of the matrix rows are contiguous in memory).\n\n @note\n This API does not support XERBLA. Instead, unlike the standard BLAS\n functions, this one returns a dnnl_status_t value to allow error\n handling.\n\n @param transa Transposition flag for matrix A: 'N' or 'n' means A is not\n transposed, and 'T' or 't' means that A is transposed.\n @param transb Transposition flag for matrix B: 'N' or 'n' means B is not\n transposed, and 'T' or 't' means that B is transposed.\n @param M The M dimension.\n @param N The N dimension.\n @param K The K dimension.\n @param alpha The alpha parameter that is used to scale the product of\n matrices A and B.\n @param A A pointer to the A matrix data.\n @param lda The leading dimension for the matrix A.\n @param B A pointer to the B matrix data.\n @param ldb The leading dimension for the matrix B.\n @param beta The beta parameter that is used to scale the matrix C.\n @param C A pointer to the C matrix data.\n @param ldc The leading dimension for the matrix C.\n @returns #dnnl_success/#dnnl::status::success on success and a status\n describing the error otherwise."]
5642 pub fn dnnl_sgemm(
5643 transa: ::std::os::raw::c_char,
5644 transb: ::std::os::raw::c_char,
5645 M: dnnl_dim_t,
5646 N: dnnl_dim_t,
5647 K: dnnl_dim_t,
5648 alpha: f32,
5649 A: *const f32,
5650 lda: dnnl_dim_t,
5651 B: *const f32,
5652 ldb: dnnl_dim_t,
5653 beta: f32,
5654 C: *mut f32,
5655 ldc: dnnl_dim_t,
5656 ) -> dnnl_status_t::Type;
5657}
5658unsafe extern "C" {
5659 #[doc = " Performs integer matrix-matrix multiply on 8-bit unsigned matrix A, 8-bit\n signed matrix B, and 32-bit signed resulting matrix C.\n\n The operation is defined as:\n\n `C := alpha * (op(A) - A_offset) * (op(B) - B_offset) + beta * C + C_offset`\n\n where\n - `op( X ) = X` or `op( X ) = X**T`,\n - `alpha` and `beta` are scalars, and\n - `A`, `B`, and `C` are matrices:\n - `op( A )` is an `MxK` matrix,\n - `op( B )` is an `KxN` matrix,\n - `C` is an `MxN` matrix.\n - `A_offset` is an `MxK` matrix with every element equal the `ao` value,\n - `B_offset` is an `KxN` matrix with every element equal the `bo` value,\n - `C_offset` is an `MxN` matrix which is defined by the `co` array of size `len`:\n - if `offsetc = F`: the `len` must be at least `1`,\n - if `offsetc = C`: the `len` must be at least `max(1, m)`,\n - if `offsetc = R`: the `len` must be at least `max(1, n)`,\n\n The matrices are assumed to be stored in row-major order (the elements in\n each of the matrix rows are contiguous in memory).\n\n @note\n This API does not support XERBLA. Instead, unlike the standard BLAS\n functions, this one returns a dnnl_status_t value to allow error\n handling.\n\n @warning\n On some architectures saturation may happen during intermediate\n computations, which would lead to unexpected results. For more\n details, refer to @ref dev_guide_int8_computations.\n\n @param transa Transposition flag for matrix A: 'N' or 'n' means A is not\n transposed, and 'T' or 't' means that A is transposed.\n @param transb Transposition flag for matrix B: 'N' or 'n' means B is not\n transposed, and 'T' or 't' means that B is transposed.\n @param offsetc Flag specifying how offsets should be applied to matrix C:\n - 'F' means that the same offset will be applied to each element of\n the matrix C,\n - 'C' means that individual offset will be applied to each element\n within each column,\n - 'R' means that individual offset will be applied to each element\n within each row.\n @param M The M dimension.\n @param N The N dimension.\n @param K The K dimension.\n @param alpha The alpha parameter that is used to scale the product of\n matrices A and B.\n @param A A pointer to the A matrix data.\n @param lda The leading dimension for the matrix A.\n @param ao The offset value for the matrix A.\n @param B A pointer to the B matrix data.\n @param ldb The leading dimension for the matrix B.\n @param bo The offset value for the matrix B.\n @param beta The beta parameter that is used to scale the matrix C.\n @param C A pointer to the C matrix data.\n @param ldc The leading dimension for the matrix C.\n @param co An array of offset values for the matrix C. The number of\n elements in the array depends on the value of @p offsetc.\n @returns #dnnl_success/#dnnl::status::success on success and a status\n describing the error otherwise."]
5660 pub fn dnnl_gemm_u8s8s32(
5661 transa: ::std::os::raw::c_char,
5662 transb: ::std::os::raw::c_char,
5663 offsetc: ::std::os::raw::c_char,
5664 M: dnnl_dim_t,
5665 N: dnnl_dim_t,
5666 K: dnnl_dim_t,
5667 alpha: f32,
5668 A: *const u8,
5669 lda: dnnl_dim_t,
5670 ao: u8,
5671 B: *const i8,
5672 ldb: dnnl_dim_t,
5673 bo: i8,
5674 beta: f32,
5675 C: *mut i32,
5676 ldc: dnnl_dim_t,
5677 co: *const i32,
5678 ) -> dnnl_status_t::Type;
5679}
5680unsafe extern "C" {
5681 #[doc = " Performs integer matrix-matrix multiply on 8-bit signed matrix A, 8-bit\n signed matrix B, and 32-bit signed resulting matrix C.\n\n The operation is defined as:\n\n `C := alpha * (op(A) - A_offset) * (op(B) - B_offset) + beta * C + C_offset`\n\n where\n - `op( X ) = X` or `op( X ) = X**T`,\n - `alpha` and `beta` are scalars, and\n - `A`, `B`, and `C` are matrices:\n - `op( A )` is an `MxK` matrix,\n - `op( B )` is an `KxN` matrix,\n - `C` is an `MxN` matrix.\n - `A_offset` is an `MxK` matrix with every element equal the `ao` value,\n - `B_offset` is an `KxN` matrix with every element equal the `bo` value,\n - `C_offset` is an `MxN` matrix which is defined by the `co` array of size `len`:\n - if `offsetc = F`: the `len` must be at least `1`,\n - if `offsetc = C`: the `len` must be at least `max(1, m)`,\n - if `offsetc = R`: the `len` must be at least `max(1, n)`,\n\n The matrices are assumed to be stored in row-major order (the elements in\n each of the matrix rows are contiguous in memory).\n\n @note\n This API does not support XERBLA. Instead, unlike the standard BLAS\n functions, this one returns a dnnl_status_t value to allow error\n handling.\n\n @warning\n On some architectures saturation may happen during intermediate\n computations, which would lead to unexpected results. For more\n details, refer to @ref dev_guide_int8_computations.\n\n @param transa Transposition flag for matrix A: 'N' or 'n' means A is not\n transposed, and 'T' or 't' means that A is transposed.\n @param transb Transposition flag for matrix B: 'N' or 'n' means B is not\n transposed, and 'T' or 't' means that B is transposed.\n @param offsetc Flag specifying how offsets should be applied to matrix C:\n - 'F' means that the same offset will be applied to each element of\n the matrix C,\n - 'C' means that individual offset will be applied to each element\n within each column,\n - 'R' means that individual offset will be applied to each element\n within each row.\n @param M The M dimension.\n @param N The N dimension.\n @param K The K dimension.\n @param alpha The alpha parameter that is used to scale the product of\n matrices A and B.\n @param A A pointer to the A matrix data.\n @param lda The leading dimension for the matrix A.\n @param ao The offset value for the matrix A.\n @param B A pointer to the B matrix data.\n @param ldb The leading dimension for the matrix B.\n @param bo The offset value for the matrix B.\n @param beta The beta parameter that is used to scale the matrix C.\n @param C A pointer to the C matrix data.\n @param ldc The leading dimension for the matrix C.\n @param co An array of offset values for the matrix C. The number of\n elements in the array depends on the value of @p offsetc.\n @returns #dnnl_success/#dnnl::status::success on success and a status\n describing the error otherwise."]
5682 pub fn dnnl_gemm_s8s8s32(
5683 transa: ::std::os::raw::c_char,
5684 transb: ::std::os::raw::c_char,
5685 offsetc: ::std::os::raw::c_char,
5686 M: dnnl_dim_t,
5687 N: dnnl_dim_t,
5688 K: dnnl_dim_t,
5689 alpha: f32,
5690 A: *const i8,
5691 lda: dnnl_dim_t,
5692 ao: i8,
5693 B: *const i8,
5694 ldb: dnnl_dim_t,
5695 bo: i8,
5696 beta: f32,
5697 C: *mut i32,
5698 ldc: dnnl_dim_t,
5699 co: *const i32,
5700 ) -> dnnl_status_t::Type;
5701}
5702pub mod dnnl_graph_layout_type_t {
5703 #[doc = " Layout type specification"]
5704 pub type Type = ::std::os::raw::c_uint;
5705 #[doc = " Undefined layout type"]
5706 pub const dnnl_graph_layout_type_undef: Type = 0;
5707 #[doc = " Any means to let the library to decide the layout for a tensor during\n partition compilation."]
5708 pub const dnnl_graph_layout_type_any: Type = 1;
5709 #[doc = " Strided means that the layout of a tensor is determined by the strides\n field in the logical tensor."]
5710 pub const dnnl_graph_layout_type_strided: Type = 2;
5711 #[doc = " Opaque means that the layout of a tensor is the library specific.\n Usually, an opaque layout is generated by a partition which is compiled\n with layout type any."]
5712 pub const dnnl_graph_layout_type_opaque: Type = 3;
5713}
5714pub mod dnnl_graph_tensor_property_t {
5715 #[doc = " Logical tensor property"]
5716 pub type Type = ::std::os::raw::c_uint;
5717 #[doc = " Undefined tensor property"]
5718 pub const dnnl_graph_tensor_property_undef: Type = 0;
5719 #[doc = " Variable means the tensor may be changed during computation or between\n different iterations."]
5720 pub const dnnl_graph_tensor_property_variable: Type = 1;
5721 #[doc = " Constant means the tensor will keep unchanged during computation and\n between different iterations. It's useful for the library to apply\n optimizations for constant tensors or cache constant tensors inside the\n library. For example, constant weight tensors in inference scenarios."]
5722 pub const dnnl_graph_tensor_property_constant: Type = 2;
5723}
5724#[doc = " Logical tensor. It is based on an ID, a number of dimensions, dimensions\n themselves, element data type, tensor property and tensor memory layout."]
5725#[repr(C)]
5726#[derive(Copy, Clone)]
5727pub struct dnnl_graph_logical_tensor_t {
5728 #[doc = " Unique id of each logical tensor. The library uses logical tensor IDs to\n build up the connections between operations if the output of one\n operation has the same ID as the input of another operation."]
5729 pub id: usize,
5730 #[doc = " Number of dimensions. -1 means unknown (DNNL_GRAPH_UNKNOWN_NDIMS). 0 is\n used to define scalar tensor."]
5731 pub ndims: ::std::os::raw::c_int,
5732 #[doc = " Size of each dimension. #DNNL_GRAPH_UNKNOWN_DIM means the size of that\n dimension is unknown. 0 is used to define zero-dimension tensor. The\n library supports to deduce output shapes according to input shapes\n during compilation. Unlike memory descriptor in oneDNN primitive API,\n the order of dimensions is not defined in logical tensor. It is defined\n by the operations which respect the order through the attributes\n #dnnl_graph_op_attr_data_format or #dnnl_graph_op_attr_weights_format.\n For example, for a Convolution with `data_format=NXC`, it means the\n first element of dims of activation tensor is mini-batch size, the last\n effective element of dims is channel size, and other elements between\n them are spatial dimensions."]
5733 pub dims: dnnl_dims_t,
5734 #[doc = " Data type of the tensor elements."]
5735 pub data_type: dnnl_data_type_t::Type,
5736 #[doc = " Property type of the tensor."]
5737 pub property: dnnl_graph_tensor_property_t::Type,
5738 #[doc = " Layout type of the tensor."]
5739 pub layout_type: dnnl_graph_layout_type_t::Type,
5740 pub layout: dnnl_graph_logical_tensor_t__bindgen_ty_1,
5741}
5742#[repr(C)]
5743#[derive(Copy, Clone)]
5744pub union dnnl_graph_logical_tensor_t__bindgen_ty_1 {
5745 #[doc = " The field is valid when `layout_type` is\n #dnnl_graph_layout_type_strided. #DNNL_GRAPH_UNKNOWN_DIM means the\n stride of the dimension is unknown. The library currently doesn't\n support other negative stride values."]
5746 pub strides: dnnl_dims_t,
5747 #[doc = " The field is valid when `layout_type` is\n #dnnl_graph_layout_type_opaque. An opaque layout ID is usually\n generated by a partition which is compiled with layout type any."]
5748 pub layout_id: usize,
5749}
5750#[allow(clippy::unnecessary_operation, clippy::identity_op)]
5751const _: () = {
5752 ["Size of dnnl_graph_logical_tensor_t__bindgen_ty_1"]
5753 [::std::mem::size_of::<dnnl_graph_logical_tensor_t__bindgen_ty_1>() - 96usize];
5754 ["Alignment of dnnl_graph_logical_tensor_t__bindgen_ty_1"]
5755 [::std::mem::align_of::<dnnl_graph_logical_tensor_t__bindgen_ty_1>() - 8usize];
5756 ["Offset of field: dnnl_graph_logical_tensor_t__bindgen_ty_1::strides"]
5757 [::std::mem::offset_of!(dnnl_graph_logical_tensor_t__bindgen_ty_1, strides) - 0usize];
5758 ["Offset of field: dnnl_graph_logical_tensor_t__bindgen_ty_1::layout_id"]
5759 [::std::mem::offset_of!(dnnl_graph_logical_tensor_t__bindgen_ty_1, layout_id) - 0usize];
5760};
5761#[allow(clippy::unnecessary_operation, clippy::identity_op)]
5762const _: () = {
5763 ["Size of dnnl_graph_logical_tensor_t"]
5764 [::std::mem::size_of::<dnnl_graph_logical_tensor_t>() - 224usize];
5765 ["Alignment of dnnl_graph_logical_tensor_t"]
5766 [::std::mem::align_of::<dnnl_graph_logical_tensor_t>() - 8usize];
5767 ["Offset of field: dnnl_graph_logical_tensor_t::id"]
5768 [::std::mem::offset_of!(dnnl_graph_logical_tensor_t, id) - 0usize];
5769 ["Offset of field: dnnl_graph_logical_tensor_t::ndims"]
5770 [::std::mem::offset_of!(dnnl_graph_logical_tensor_t, ndims) - 8usize];
5771 ["Offset of field: dnnl_graph_logical_tensor_t::dims"]
5772 [::std::mem::offset_of!(dnnl_graph_logical_tensor_t, dims) - 16usize];
5773 ["Offset of field: dnnl_graph_logical_tensor_t::data_type"]
5774 [::std::mem::offset_of!(dnnl_graph_logical_tensor_t, data_type) - 112usize];
5775 ["Offset of field: dnnl_graph_logical_tensor_t::property"]
5776 [::std::mem::offset_of!(dnnl_graph_logical_tensor_t, property) - 116usize];
5777 ["Offset of field: dnnl_graph_logical_tensor_t::layout_type"]
5778 [::std::mem::offset_of!(dnnl_graph_logical_tensor_t, layout_type) - 120usize];
5779 ["Offset of field: dnnl_graph_logical_tensor_t::layout"]
5780 [::std::mem::offset_of!(dnnl_graph_logical_tensor_t, layout) - 128usize];
5781};
5782pub mod dnnl_graph_partition_policy_t {
5783 #[doc = " Policy specifications for partitioning"]
5784 pub type Type = ::std::os::raw::c_uint;
5785 #[doc = " Fusion policy returns partitions with typical post-op fusions, eg.\n Convolution + ReLU or other element-wise operations or a chian of\n post-ops."]
5786 pub const dnnl_graph_partition_policy_fusion: Type = 1;
5787 #[doc = " Debug policy doesn't not apply any fusions. It returns partitions with\n single operation in each partition. The policy is useful when users\n notice any bug or correctness issue in fusion policy."]
5788 pub const dnnl_graph_partition_policy_debug: Type = 2;
5789}
5790#[doc = " An opaque structure to describe a partition."]
5791#[repr(C)]
5792#[derive(Debug, Copy, Clone)]
5793pub struct dnnl_graph_partition {
5794 _unused: [u8; 0],
5795}
5796#[doc = " A partition handle."]
5797pub type dnnl_graph_partition_t = *mut dnnl_graph_partition;
5798#[doc = " A constant partition handle."]
5799pub type const_dnnl_graph_partition_t = *const dnnl_graph_partition;
5800#[doc = " An opaque structure to describe a graph."]
5801#[repr(C)]
5802#[derive(Debug, Copy, Clone)]
5803pub struct dnnl_graph_graph {
5804 _unused: [u8; 0],
5805}
5806#[doc = " A graph handle."]
5807pub type dnnl_graph_graph_t = *mut dnnl_graph_graph;
5808#[doc = " A constant graph handle."]
5809pub type const_dnnl_graph_graph_t = *const dnnl_graph_graph;
5810pub mod dnnl_graph_op_kind_t {
5811 #[doc = " Kinds of operations"]
5812 pub type Type = ::std::os::raw::c_uint;
5813 pub const dnnl_graph_op_abs: Type = 0;
5814 pub const dnnl_graph_op_abs_backward: Type = 1;
5815 pub const dnnl_graph_op_add: Type = 2;
5816 pub const dnnl_graph_op_avg_pool: Type = 3;
5817 pub const dnnl_graph_op_avg_pool_backward: Type = 4;
5818 pub const dnnl_graph_op_batch_norm_backward: Type = 5;
5819 pub const dnnl_graph_op_batch_norm_forward_training: Type = 6;
5820 pub const dnnl_graph_op_batch_norm_inference: Type = 7;
5821 pub const dnnl_graph_op_bias_add: Type = 8;
5822 pub const dnnl_graph_op_bias_add_backward: Type = 9;
5823 pub const dnnl_graph_op_clamp: Type = 10;
5824 pub const dnnl_graph_op_clamp_backward: Type = 11;
5825 pub const dnnl_graph_op_concat: Type = 12;
5826 pub const dnnl_graph_op_convolution: Type = 13;
5827 pub const dnnl_graph_op_convolution_backward_data: Type = 14;
5828 pub const dnnl_graph_op_convolution_backward_weights: Type = 15;
5829 pub const dnnl_graph_op_conv_transpose: Type = 16;
5830 pub const dnnl_graph_op_conv_transpose_backward_data: Type = 17;
5831 pub const dnnl_graph_op_conv_transpose_backward_weights: Type = 18;
5832 pub const dnnl_graph_op_dequantize: Type = 19;
5833 pub const dnnl_graph_op_divide: Type = 20;
5834 pub const dnnl_graph_op_dynamic_dequantize: Type = 21;
5835 pub const dnnl_graph_op_dynamic_quantize: Type = 22;
5836 pub const dnnl_graph_op_elu: Type = 23;
5837 pub const dnnl_graph_op_elu_backward: Type = 24;
5838 pub const dnnl_graph_op_end: Type = 25;
5839 pub const dnnl_graph_op_exp: Type = 26;
5840 pub const dnnl_graph_op_gelu: Type = 27;
5841 pub const dnnl_graph_op_gelu_backward: Type = 28;
5842 pub const dnnl_graph_op_hard_swish: Type = 29;
5843 pub const dnnl_graph_op_hard_swish_backward: Type = 30;
5844 pub const dnnl_graph_op_interpolate: Type = 31;
5845 pub const dnnl_graph_op_interpolate_backward: Type = 32;
5846 pub const dnnl_graph_op_layer_norm: Type = 33;
5847 pub const dnnl_graph_op_layer_norm_backward: Type = 34;
5848 pub const dnnl_graph_op_leaky_relu: Type = 35;
5849 pub const dnnl_graph_op_log: Type = 36;
5850 pub const dnnl_graph_op_log_softmax: Type = 37;
5851 pub const dnnl_graph_op_log_softmax_backward: Type = 38;
5852 pub const dnnl_graph_op_matmul: Type = 39;
5853 pub const dnnl_graph_op_maximum: Type = 40;
5854 pub const dnnl_graph_op_max_pool: Type = 41;
5855 pub const dnnl_graph_op_max_pool_backward: Type = 42;
5856 pub const dnnl_graph_op_minimum: Type = 43;
5857 pub const dnnl_graph_op_mish: Type = 44;
5858 pub const dnnl_graph_op_mish_backward: Type = 45;
5859 pub const dnnl_graph_op_multiply: Type = 46;
5860 pub const dnnl_graph_op_prelu: Type = 47;
5861 pub const dnnl_graph_op_prelu_backward: Type = 48;
5862 pub const dnnl_graph_op_quantize: Type = 49;
5863 pub const dnnl_graph_op_reciprocal: Type = 50;
5864 pub const dnnl_graph_op_reduce_l1: Type = 51;
5865 pub const dnnl_graph_op_reduce_l2: Type = 52;
5866 pub const dnnl_graph_op_reduce_max: Type = 53;
5867 pub const dnnl_graph_op_reduce_mean: Type = 54;
5868 pub const dnnl_graph_op_reduce_min: Type = 55;
5869 pub const dnnl_graph_op_reduce_prod: Type = 56;
5870 pub const dnnl_graph_op_reduce_sum: Type = 57;
5871 pub const dnnl_graph_op_relu: Type = 58;
5872 pub const dnnl_graph_op_relu_backward: Type = 59;
5873 pub const dnnl_graph_op_reorder: Type = 60;
5874 pub const dnnl_graph_op_round: Type = 61;
5875 pub const dnnl_graph_op_sigmoid: Type = 62;
5876 pub const dnnl_graph_op_sigmoid_backward: Type = 63;
5877 pub const dnnl_graph_op_softmax: Type = 64;
5878 pub const dnnl_graph_op_softmax_backward: Type = 65;
5879 pub const dnnl_graph_op_softplus: Type = 66;
5880 pub const dnnl_graph_op_softplus_backward: Type = 67;
5881 pub const dnnl_graph_op_sqrt: Type = 68;
5882 pub const dnnl_graph_op_sqrt_backward: Type = 69;
5883 pub const dnnl_graph_op_square: Type = 70;
5884 pub const dnnl_graph_op_squared_difference: Type = 71;
5885 pub const dnnl_graph_op_static_reshape: Type = 72;
5886 pub const dnnl_graph_op_static_transpose: Type = 73;
5887 pub const dnnl_graph_op_subtract: Type = 74;
5888 pub const dnnl_graph_op_tanh: Type = 75;
5889 pub const dnnl_graph_op_tanh_backward: Type = 76;
5890 pub const dnnl_graph_op_type_cast: Type = 77;
5891 pub const dnnl_graph_op_wildcard: Type = 78;
5892 pub const dnnl_graph_op_hard_sigmoid: Type = 79;
5893 pub const dnnl_graph_op_hard_sigmoid_backward: Type = 80;
5894 pub const dnnl_graph_op_select: Type = 81;
5895 pub const dnnl_graph_op_pow: Type = 82;
5896 pub const dnnl_graph_op_group_norm: Type = 83;
5897 pub const dnnl_graph_op_gen_index: Type = 84;
5898 pub const dnnl_graph_op_greater_equal: Type = 85;
5899 pub const dnnl_graph_op_last_symbol: Type = 86;
5900}
5901pub mod dnnl_graph_op_attr_t {
5902 #[doc = " Attributes of operations"]
5903 pub type Type = ::std::os::raw::c_uint;
5904 #[doc = " Undefined op attribute."]
5905 pub const dnnl_graph_op_attr_undef: Type = 0;
5906 #[doc = " Specifies an alpha attribute to an op."]
5907 pub const dnnl_graph_op_attr_alpha: Type = 1;
5908 #[doc = " Specifies an beta attribute to an op."]
5909 pub const dnnl_graph_op_attr_beta: Type = 2;
5910 #[doc = " Specifies an epsilon attribute to an op."]
5911 pub const dnnl_graph_op_attr_epsilon: Type = 3;
5912 #[doc = " Specifies a max attribute to an op."]
5913 pub const dnnl_graph_op_attr_max: Type = 4;
5914 #[doc = "Specifies a min attribute to an op."]
5915 pub const dnnl_graph_op_attr_min: Type = 5;
5916 #[doc = " Specifies a momentum attribute to an op."]
5917 pub const dnnl_graph_op_attr_momentum: Type = 6;
5918 #[doc = " Specifies a scales attribute to an op."]
5919 pub const dnnl_graph_op_attr_scales: Type = 32;
5920 #[doc = " Specifies an axis attribute to an op."]
5921 pub const dnnl_graph_op_attr_axis: Type = 48;
5922 #[doc = " Specifies a begin_norm_axis attribute to an op."]
5923 pub const dnnl_graph_op_attr_begin_norm_axis: Type = 49;
5924 #[doc = " Specifies a groups attribute to an op."]
5925 pub const dnnl_graph_op_attr_groups: Type = 50;
5926 #[doc = " Specifies an axes attribute to an op."]
5927 pub const dnnl_graph_op_attr_axes: Type = 64;
5928 #[doc = " Specifies a dilations attribute to an op."]
5929 pub const dnnl_graph_op_attr_dilations: Type = 65;
5930 #[doc = " Specifies an dst_shape attribute to an op."]
5931 pub const dnnl_graph_op_attr_dst_shape: Type = 66;
5932 #[doc = " Specifies a kernel attribute to an op."]
5933 pub const dnnl_graph_op_attr_kernel: Type = 67;
5934 #[doc = " Specifies an order attribute to an op."]
5935 pub const dnnl_graph_op_attr_order: Type = 68;
5936 #[doc = " Specifies an output_padding attribute to an op."]
5937 pub const dnnl_graph_op_attr_output_padding: Type = 69;
5938 #[doc = " Specifies a pads_begin attribute to an op."]
5939 pub const dnnl_graph_op_attr_pads_begin: Type = 70;
5940 #[doc = " Specifies a pads_end attribute to an op."]
5941 pub const dnnl_graph_op_attr_pads_end: Type = 71;
5942 #[doc = " Specifies a shape attribute to an op."]
5943 pub const dnnl_graph_op_attr_shape: Type = 72;
5944 #[doc = " Specifies a sizes attribute to an op."]
5945 pub const dnnl_graph_op_attr_sizes: Type = 73;
5946 #[doc = " Specifies a input_shape attribute to an op."]
5947 pub const dnnl_graph_op_attr_src_shape: Type = 74;
5948 #[doc = " Specifies a strides attribute to an op."]
5949 pub const dnnl_graph_op_attr_strides: Type = 75;
5950 #[doc = " Specifies a weight_shape attribute to an op."]
5951 pub const dnnl_graph_op_attr_weights_shape: Type = 76;
5952 #[doc = " Specifies a zps attribute to an op."]
5953 pub const dnnl_graph_op_attr_zps: Type = 77;
5954 #[doc = " Specifies a group shape attribute to an op."]
5955 pub const dnnl_graph_op_attr_group_shape: Type = 78;
5956 #[doc = " Specifies an exclude_pad attribute to an op."]
5957 pub const dnnl_graph_op_attr_exclude_pad: Type = 96;
5958 #[doc = " Specifies a keep_dims attribute to an op."]
5959 pub const dnnl_graph_op_attr_keep_dims: Type = 97;
5960 #[doc = " Specifies a keep_stats attribute to an op."]
5961 pub const dnnl_graph_op_attr_keep_stats: Type = 98;
5962 #[doc = " Specifies a per_channel_broadcast attribute to an op."]
5963 pub const dnnl_graph_op_attr_per_channel_broadcast: Type = 99;
5964 #[doc = " Specifies a special_zero attribute to an op."]
5965 pub const dnnl_graph_op_attr_special_zero: Type = 100;
5966 #[doc = " Specifies a transpose_a attribute to an op."]
5967 pub const dnnl_graph_op_attr_transpose_a: Type = 101;
5968 #[doc = " Specifies a transpose_b attribute to an op."]
5969 pub const dnnl_graph_op_attr_transpose_b: Type = 102;
5970 #[doc = " Specifies an use_affine attribute to an op."]
5971 pub const dnnl_graph_op_attr_use_affine: Type = 103;
5972 #[doc = " Specifies an use_dst attribute to an op."]
5973 pub const dnnl_graph_op_attr_use_dst: Type = 104;
5974 #[doc = " Specifies an auto_broadcast attribute to an op. The value can be \"none\"\n or \"numpy\"."]
5975 pub const dnnl_graph_op_attr_auto_broadcast: Type = 128;
5976 #[doc = " Specifies an auto_pad attribute to an op. The value can be \"none\",\n \"same_upper\", \"same_lower\", or \"valid\"."]
5977 pub const dnnl_graph_op_attr_auto_pad: Type = 129;
5978 #[doc = " Specifies an coordinate_transformation_mode attribute to an op. The\n value can be \"half_pixel\" or \"align_corners\". The attribute is defined\n for Interpolate operations."]
5979 pub const dnnl_graph_op_attr_coordinate_transformation_mode: Type = 130;
5980 #[doc = " Specifies a data_format of an op. The value can be \"NCX\" or \"NXC\"."]
5981 pub const dnnl_graph_op_attr_data_format: Type = 131;
5982 #[doc = " Specifies a mode attribute of an op.\n Interpolate: \"nearest\", \"linear\", \"bilinear\", or \"trilinear\".\n SoftMax: \"none\", \"inf_as_zero\"."]
5983 pub const dnnl_graph_op_attr_mode: Type = 132;
5984 #[doc = " Specifies a qtype attribute to an op. The value can be \"per_channel\" or\n \"per_tensor\". The attribute is defined for quantization operations."]
5985 pub const dnnl_graph_op_attr_qtype: Type = 133;
5986 #[doc = " Specifies a rounding_type attribute to an op. The value can be \"ceil\" or\n \"floor\"."]
5987 pub const dnnl_graph_op_attr_rounding_type: Type = 134;
5988 #[doc = " Specifies a weights_format of an op. The value can be \"OIX\", \"XIO\",\n \"IOX\", or \"XOI\". Different operations may support different values."]
5989 pub const dnnl_graph_op_attr_weights_format: Type = 135;
5990 #[doc = " Specifies the end of all above exteral attributes for check."]
5991 pub const dnnl_graph_op_attr_end: Type = 255;
5992}
5993#[doc = " An opaque structure to describe an operation."]
5994#[repr(C)]
5995#[derive(Debug, Copy, Clone)]
5996pub struct dnnl_graph_op {
5997 _unused: [u8; 0],
5998}
5999#[doc = " An operation handle."]
6000pub type dnnl_graph_op_t = *mut dnnl_graph_op;
6001#[doc = " A constant operation handle."]
6002pub type const_dnnl_graph_op_t = *const dnnl_graph_op;
6003#[doc = " Allocation call-back function interface for host. For SYCL allocator, see\n #dnnl_graph_sycl_allocate_f."]
6004pub type dnnl_graph_host_allocate_f = ::std::option::Option<
6005 unsafe extern "C" fn(size: usize, alignment: usize) -> *mut ::std::os::raw::c_void,
6006>;
6007#[doc = " Deallocation call-back function interface for host. For SYCL allocator, see\n #dnnl_graph_sycl_deallocate_f."]
6008pub type dnnl_graph_host_deallocate_f =
6009 ::std::option::Option<unsafe extern "C" fn(arg1: *mut ::std::os::raw::c_void)>;
6010#[doc = " An opaque structure to describe an allocator."]
6011#[repr(C)]
6012#[derive(Debug, Copy, Clone)]
6013pub struct dnnl_graph_allocator {
6014 _unused: [u8; 0],
6015}
6016#[doc = " An allocator handle."]
6017pub type dnnl_graph_allocator_t = *mut dnnl_graph_allocator;
6018#[doc = " A constant allocator handle."]
6019pub type const_dnnl_graph_allocator_t = *const dnnl_graph_allocator;
6020#[doc = " In-place pair definition. It can queried from a compiled partition\n indicating that an input and an output of the partition can share the same\n memory buffer for computation. In-place computation helps to reduce the\n memory footprint and improves cache locality. But since the library may not\n have a global view of user's application, it's possible that the tensor with\n `input_id` is used at other places in user's computation graph. In this\n case, the user should take the in-place pair as a hint and pass a different\n memory buffer for output tensor to avoid overwriting the input memory buffer\n which will probably cause unexpected incorrect results."]
6021#[repr(C)]
6022#[derive(Debug, Copy, Clone)]
6023pub struct dnnl_graph_inplace_pair_t {
6024 #[doc = " The id of input tensor"]
6025 pub input_id: usize,
6026 #[doc = " The id of output tensor"]
6027 pub output_id: usize,
6028}
6029#[allow(clippy::unnecessary_operation, clippy::identity_op)]
6030const _: () = {
6031 ["Size of dnnl_graph_inplace_pair_t"]
6032 [::std::mem::size_of::<dnnl_graph_inplace_pair_t>() - 16usize];
6033 ["Alignment of dnnl_graph_inplace_pair_t"]
6034 [::std::mem::align_of::<dnnl_graph_inplace_pair_t>() - 8usize];
6035 ["Offset of field: dnnl_graph_inplace_pair_t::input_id"]
6036 [::std::mem::offset_of!(dnnl_graph_inplace_pair_t, input_id) - 0usize];
6037 ["Offset of field: dnnl_graph_inplace_pair_t::output_id"]
6038 [::std::mem::offset_of!(dnnl_graph_inplace_pair_t, output_id) - 8usize];
6039};
6040#[doc = " An opaque structure to describe a compiled partition."]
6041#[repr(C)]
6042#[derive(Debug, Copy, Clone)]
6043pub struct dnnl_graph_compiled_partition {
6044 _unused: [u8; 0],
6045}
6046#[doc = " A compiled partition handle."]
6047pub type dnnl_graph_compiled_partition_t = *mut dnnl_graph_compiled_partition;
6048#[doc = " A constant compiled partition handle."]
6049pub type const_dnnl_graph_compiled_partition_t = *const dnnl_graph_compiled_partition;
6050#[doc = " An opaque structure to describe a tensor."]
6051#[repr(C)]
6052#[derive(Debug, Copy, Clone)]
6053pub struct dnnl_graph_tensor {
6054 _unused: [u8; 0],
6055}
6056#[doc = " A tensor handle."]
6057pub type dnnl_graph_tensor_t = *mut dnnl_graph_tensor;
6058#[doc = " A constant tensor handle."]
6059pub type const_dnnl_graph_tensor_t = *const dnnl_graph_tensor;
6060unsafe extern "C" {
6061 #[doc = " Creates a host allocator with the given allocation and deallocation\n call-back function pointers.\n\n @param allocator Output allocator.\n @param host_malloc A pointer to malloc function for host.\n @param host_free A pointer to free function for host.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6062 pub fn dnnl_graph_allocator_create(
6063 allocator: *mut dnnl_graph_allocator_t,
6064 host_malloc: dnnl_graph_host_allocate_f,
6065 host_free: dnnl_graph_host_deallocate_f,
6066 ) -> dnnl_status_t::Type;
6067}
6068unsafe extern "C" {
6069 #[doc = " Destroys an allocator.\n\n @param allocator The allocator to be destroyed.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6070 pub fn dnnl_graph_allocator_destroy(allocator: dnnl_graph_allocator_t) -> dnnl_status_t::Type;
6071}
6072unsafe extern "C" {
6073 #[doc = " This API is a supplement for existing onednn engine API."]
6074 pub fn dnnl_graph_make_engine_with_allocator(
6075 engine: *mut dnnl_engine_t,
6076 kind: dnnl_engine_kind_t::Type,
6077 index: usize,
6078 alloc: const_dnnl_graph_allocator_t,
6079 ) -> dnnl_status_t::Type;
6080}
6081unsafe extern "C" {
6082 #[doc = " Initializes a logical tensor with id, data type, number of dimensions,\n layout type, and property. The logical tensor's dims are unknown with this\n interface.\n\n @param logical_tensor Output logical tensor.\n @param tid The unique id of the output logical tensor.\n @param dtype Elements data type.\n @param ndims Number of dimensions.\n @param ltype Layout type of the underlying tensor buffer.\n @param ptype Tensor property type.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6083 pub fn dnnl_graph_logical_tensor_init(
6084 logical_tensor: *mut dnnl_graph_logical_tensor_t,
6085 tid: usize,
6086 dtype: dnnl_data_type_t::Type,
6087 ndims: i32,
6088 ltype: dnnl_graph_layout_type_t::Type,
6089 ptype: dnnl_graph_tensor_property_t::Type,
6090 ) -> dnnl_status_t::Type;
6091}
6092unsafe extern "C" {
6093 #[doc = " Initializes a logical tensor with basic information and dims. The logical\n tensor's dimensions and layout will be initialized according to the input\n arguments.\n\n @note\n If dims contains all valid values and layout type is\n #dnnl_graph_layout_type_strided. The strides field in\n #dnnl_graph_logical_tensor_t will be calculated in a row major and\n contiguous way. Otherwise, Accessing the strides field is an undefined\n behavior.\n\n Eg. dims (2, 3, 4, 5) will get strides (60, 20, 5, 1)\n\n @param logical_tensor Output logical tensor.\n @param tid The unique id of output logical tensor.\n @param dtype Elements data type.\n @param ndims Number of dimensions.\n @param dims Array of dimensions.\n @param ltype Layout type of the underlying tensor memory.\n @param ptype Tensor property type.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6094 pub fn dnnl_graph_logical_tensor_init_with_dims(
6095 logical_tensor: *mut dnnl_graph_logical_tensor_t,
6096 tid: usize,
6097 dtype: dnnl_data_type_t::Type,
6098 ndims: i32,
6099 dims: *const dnnl_dim_t,
6100 ltype: dnnl_graph_layout_type_t::Type,
6101 ptype: dnnl_graph_tensor_property_t::Type,
6102 ) -> dnnl_status_t::Type;
6103}
6104unsafe extern "C" {
6105 #[doc = " Initializes a logical tensor with dimensions and strides provided by user.\n\n @note\n Once strides are explicitly provided through the API, the `layout_type`\n in #dnnl_graph_logical_tensor_t can only be\n #dnnl_graph_layout_type_strided or #dnnl_graph_layout_type_any.\n\n @param logical_tensor Output logical tensor.\n @param tid The unique id of output logical tensor.\n @param dtype Elements data type.\n @param ndims Number of dimensions.\n @param dims Array of dimensions.\n @param strides Array of strides.\n @param ptype Tensor property type.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6106 pub fn dnnl_graph_logical_tensor_init_with_strides(
6107 logical_tensor: *mut dnnl_graph_logical_tensor_t,
6108 tid: usize,
6109 dtype: dnnl_data_type_t::Type,
6110 ndims: i32,
6111 dims: *const dnnl_dim_t,
6112 strides: *const dnnl_dim_t,
6113 ptype: dnnl_graph_tensor_property_t::Type,
6114 ) -> dnnl_status_t::Type;
6115}
6116unsafe extern "C" {
6117 #[doc = " Returns the memory size described by the logical tensor. If it's a strided\n layout, the size will be calculated by `dims` and `strides`. If it's an\n opaque layout, the size will be decided by `layout_id`.\n\n @param logical_tensor Logical tensor.\n @param size Output memory size in bytes.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6118 pub fn dnnl_graph_logical_tensor_get_mem_size(
6119 logical_tensor: *const dnnl_graph_logical_tensor_t,
6120 size: *mut usize,
6121 ) -> dnnl_status_t::Type;
6122}
6123unsafe extern "C" {
6124 #[doc = " Compares if two logical tenors are equal. Users can decide accordingly\n if layout reordering is needed for two logical tensors. The method will\n return true for below two circumstances:\n\n 1. the two logical tensors are equal regarding each field in the struct,\n eg. id, ndims, dims, layout type, property, etc.\n 2. If all other fields are equal but the layout types in two logical\n tensors are different, the method will return true when the underlying\n memory layout is the same. For example, one logical tensor has strided\n layout type while the other one has opaque layout type, but underneath,\n both layouts are NHWC, the method will still return true for this case.\n\n @param lt1 The handle of first logical tensor.\n @param lt2 The handle of second logical tensor.\n @param is_equal 1 if these two logical tensors are equal, 0 otherwise.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6125 pub fn dnnl_graph_logical_tensor_is_equal(
6126 lt1: *const dnnl_graph_logical_tensor_t,
6127 lt2: *const dnnl_graph_logical_tensor_t,
6128 is_equal: *mut u8,
6129 ) -> dnnl_status_t::Type;
6130}
6131unsafe extern "C" {
6132 #[doc = " Creates a tensor with logical tensor, engine, and data handle.\n\n @param tensor Output tensor.\n @param logical_tensor Description for this tensor.\n @param engine Engine to use.\n @param handle Handle of the memory buffer to use as an underlying storage.\n - A pointer to the user-allocated buffer. In this case the library\n doesn't own the buffer.\n - The DNNL_MEMORY_ALLOCATE special value. Instructs the library to\n allocate the buffer for the tensor. In this case the library\n owns the buffer.\n - DNNL_MEMORY_NONE to create tensor without an underlying buffer.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6133 pub fn dnnl_graph_tensor_create(
6134 tensor: *mut dnnl_graph_tensor_t,
6135 logical_tensor: *const dnnl_graph_logical_tensor_t,
6136 engine: dnnl_engine_t,
6137 handle: *mut ::std::os::raw::c_void,
6138 ) -> dnnl_status_t::Type;
6139}
6140unsafe extern "C" {
6141 #[doc = " Destroys a tensor.\n\n @param tensor The tensor to be destroyed.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6142 pub fn dnnl_graph_tensor_destroy(tensor: dnnl_graph_tensor_t) -> dnnl_status_t::Type;
6143}
6144unsafe extern "C" {
6145 #[doc = " Gets the data handle of a tensor.\n\n @param tensor The input tensor.\n @param handle Pointer to the data of input tensor.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6146 pub fn dnnl_graph_tensor_get_data_handle(
6147 tensor: const_dnnl_graph_tensor_t,
6148 handle: *mut *mut ::std::os::raw::c_void,
6149 ) -> dnnl_status_t::Type;
6150}
6151unsafe extern "C" {
6152 #[doc = " Set data handle for a tensor.\n\n @param tensor The input tensor.\n @param handle New data handle for tensor.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6153 pub fn dnnl_graph_tensor_set_data_handle(
6154 tensor: dnnl_graph_tensor_t,
6155 handle: *mut ::std::os::raw::c_void,
6156 ) -> dnnl_status_t::Type;
6157}
6158unsafe extern "C" {
6159 #[doc = " Returns the engine of a tensor object.\n\n @param tensor The input tensor.\n @param engine Output engine on which the tensor is located.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6160 pub fn dnnl_graph_tensor_get_engine(
6161 tensor: const_dnnl_graph_tensor_t,
6162 engine: *mut dnnl_engine_t,
6163 ) -> dnnl_status_t::Type;
6164}
6165unsafe extern "C" {
6166 #[doc = " Returns the logical tensor of a tensor object.\n\n @param tensor The input tensor.\n @param logical_tensor Output logical tensor of the tensor object.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6167 pub fn dnnl_graph_tensor_get_logical_tensor(
6168 tensor: const_dnnl_graph_tensor_t,
6169 logical_tensor: *mut dnnl_graph_logical_tensor_t,
6170 ) -> dnnl_status_t::Type;
6171}
6172unsafe extern "C" {
6173 #[doc = " Initializes an op with unique id, kind, and name.\n\n @param op Output op\n @param id The unique id of the output op.\n @param kind The op kind.\n @param verbose_name The string added as the op name.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6174 pub fn dnnl_graph_op_create(
6175 op: *mut dnnl_graph_op_t,
6176 id: usize,
6177 kind: dnnl_graph_op_kind_t::Type,
6178 verbose_name: *const ::std::os::raw::c_char,
6179 ) -> dnnl_status_t::Type;
6180}
6181unsafe extern "C" {
6182 #[doc = " Destroys an op.\n\n @param op The op to be destroyed.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6183 pub fn dnnl_graph_op_destroy(op: dnnl_graph_op_t) -> dnnl_status_t::Type;
6184}
6185unsafe extern "C" {
6186 #[doc = " Adds input logical tensor to the op.\n\n @param op Input op.\n @param input The input logical tensor to be added.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6187 pub fn dnnl_graph_op_add_input(
6188 op: dnnl_graph_op_t,
6189 input: *const dnnl_graph_logical_tensor_t,
6190 ) -> dnnl_status_t::Type;
6191}
6192unsafe extern "C" {
6193 #[doc = " Adds output logical tensor to the op.\n\n @param op Input op.\n @param output The output logical tensor to be added.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6194 pub fn dnnl_graph_op_add_output(
6195 op: dnnl_graph_op_t,
6196 output: *const dnnl_graph_logical_tensor_t,
6197 ) -> dnnl_status_t::Type;
6198}
6199unsafe extern "C" {
6200 #[doc = " Sets floating point attribute to an op.\n\n @param op Input op.\n @param name The attribute's name.\n @param value The attribute's value.\n @param value_len The number of value element.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6201 pub fn dnnl_graph_op_set_attr_f32(
6202 op: dnnl_graph_op_t,
6203 name: dnnl_graph_op_attr_t::Type,
6204 value: *const f32,
6205 value_len: usize,
6206 ) -> dnnl_status_t::Type;
6207}
6208unsafe extern "C" {
6209 #[doc = " Sets boolean attribute to an op.\n\n @param op Input op.\n @param name The attribute's name.\n @param value The attribute's value.\n @param value_len The number of value element.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6210 pub fn dnnl_graph_op_set_attr_bool(
6211 op: dnnl_graph_op_t,
6212 name: dnnl_graph_op_attr_t::Type,
6213 value: *const u8,
6214 value_len: usize,
6215 ) -> dnnl_status_t::Type;
6216}
6217unsafe extern "C" {
6218 #[doc = " Sets integer attribute to an op.\n\n @param op Input op.\n @param name The attribute's name.\n @param value The attribute's value.\n @param value_len The number of value element.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6219 pub fn dnnl_graph_op_set_attr_s64(
6220 op: dnnl_graph_op_t,
6221 name: dnnl_graph_op_attr_t::Type,
6222 value: *const i64,
6223 value_len: usize,
6224 ) -> dnnl_status_t::Type;
6225}
6226unsafe extern "C" {
6227 #[doc = " Sets string attribute to an op.\n\n @param op Input op.\n @param name The attribute's name.\n @param value The attribute's value.\n @param value_len The length of the string value.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6228 pub fn dnnl_graph_op_set_attr_str(
6229 op: dnnl_graph_op_t,
6230 name: dnnl_graph_op_attr_t::Type,
6231 value: *const ::std::os::raw::c_char,
6232 value_len: usize,
6233 ) -> dnnl_status_t::Type;
6234}
6235unsafe extern "C" {
6236 #[doc = " Returns the unique id of an op.\n\n @param op Input op.\n @param id Output the unique id.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6237 pub fn dnnl_graph_op_get_id(op: const_dnnl_graph_op_t, id: *mut usize) -> dnnl_status_t::Type;
6238}
6239unsafe extern "C" {
6240 #[doc = " Returns the kind of an op.\n\n @param op Input op.\n @param kind Output op kind.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6241 pub fn dnnl_graph_op_get_kind(
6242 op: const_dnnl_graph_op_t,
6243 kind: *mut dnnl_graph_op_kind_t::Type,
6244 ) -> dnnl_status_t::Type;
6245}
6246unsafe extern "C" {
6247 #[doc = " Creates a new partition with a given operator and engine kind. The API is\n used to create a partition from an operation directly without creating the\n graph and calling `get_partitions()`. The output partition contains only one\n operation specified by the parameter. The output partition instance should\n be destroyed via #dnnl_graph_partition_destroy after use.\n\n @param partition The handle of output partition.\n @param op The operation used to create partition.\n @param ekind The engine kind used to create partition.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6248 pub fn dnnl_graph_partition_create_with_op(
6249 partition: *mut dnnl_graph_partition_t,
6250 op: const_dnnl_graph_op_t,
6251 ekind: dnnl_engine_kind_t::Type,
6252 ) -> dnnl_status_t::Type;
6253}
6254unsafe extern "C" {
6255 #[doc = " Destroys a partition.\n\n @param partition The partition to be destroyed.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6256 pub fn dnnl_graph_partition_destroy(partition: dnnl_graph_partition_t) -> dnnl_status_t::Type;
6257}
6258unsafe extern "C" {
6259 #[doc = " Returns the number of operations in a partition.\n\n @param partition The target partition.\n @param num Output the number of operations.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6260 pub fn dnnl_graph_partition_get_op_num(
6261 partition: const_dnnl_graph_partition_t,
6262 num: *mut usize,
6263 ) -> dnnl_status_t::Type;
6264}
6265unsafe extern "C" {
6266 #[doc = " Returns the list of op IDs of the partition.\n\n @param partition The target partition.\n @param num The number of ops.\n @param ids Output the op IDs.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6267 pub fn dnnl_graph_partition_get_ops(
6268 partition: dnnl_graph_partition_t,
6269 num: usize,
6270 ids: *mut usize,
6271 ) -> dnnl_status_t::Type;
6272}
6273unsafe extern "C" {
6274 #[doc = " Returns the ID of a partition.\n\n @param partition The target partition.\n @param id Output the ID of the partition.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6275 pub fn dnnl_graph_partition_get_id(
6276 partition: const_dnnl_graph_partition_t,
6277 id: *mut usize,
6278 ) -> dnnl_status_t::Type;
6279}
6280unsafe extern "C" {
6281 #[doc = " Compiles a partition with given input and output logical tensors. The output\n logical tensors can contain unknown dimensions. For this case, the\n compilation will deduce the output shapes according to input shapes. The\n output logical tensors can also have layout type `any`. The compilation will\n choose the optimal layout for output tensors. The optimal layout will be\n represented as an opaque layout ID saved in the output logical tensor.\n\n @param partition The target partition.\n @param compiled_partition Output compiled partition.\n @param in_num The number of input logical tensors.\n @param inputs A list of input logical tensors.\n @param out_num The number of output logical tensors.\n @param outputs A list of output logical tensors.\n @param engine The target engine of the compilation.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6282 pub fn dnnl_graph_partition_compile(
6283 partition: dnnl_graph_partition_t,
6284 compiled_partition: dnnl_graph_compiled_partition_t,
6285 in_num: usize,
6286 inputs: *mut *const dnnl_graph_logical_tensor_t,
6287 out_num: usize,
6288 outputs: *mut *const dnnl_graph_logical_tensor_t,
6289 engine: dnnl_engine_t,
6290 ) -> dnnl_status_t::Type;
6291}
6292unsafe extern "C" {
6293 #[doc = " Returns the number of input logical tensors of a partition.\n\n @param partition The target partition.\n @param num Output the number of input logical tensors.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6294 pub fn dnnl_graph_partition_get_input_ports_num(
6295 partition: const_dnnl_graph_partition_t,
6296 num: *mut usize,
6297 ) -> dnnl_status_t::Type;
6298}
6299unsafe extern "C" {
6300 #[doc = " Returns a list of input logical tensors from a partition.\n\n @param partition The target partition.\n @param num The number of input logical tensors.\n @param inputs The list of input logical tensors.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6301 pub fn dnnl_graph_partition_get_input_ports(
6302 partition: const_dnnl_graph_partition_t,
6303 num: usize,
6304 inputs: *mut dnnl_graph_logical_tensor_t,
6305 ) -> dnnl_status_t::Type;
6306}
6307unsafe extern "C" {
6308 #[doc = " Returns the number of output logical tensors of a partition.\n\n @param partition The target partition.\n @param num Output the number of output logical tensors.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6309 pub fn dnnl_graph_partition_get_output_ports_num(
6310 partition: const_dnnl_graph_partition_t,
6311 num: *mut usize,
6312 ) -> dnnl_status_t::Type;
6313}
6314unsafe extern "C" {
6315 #[doc = " Returns a list of output logical tensors from a partition.\n\n @param partition The target partition.\n @param num The number of output logical tensors.\n @param outputs The list of output logical tensors.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6316 pub fn dnnl_graph_partition_get_output_ports(
6317 partition: const_dnnl_graph_partition_t,
6318 num: usize,
6319 outputs: *mut dnnl_graph_logical_tensor_t,
6320 ) -> dnnl_status_t::Type;
6321}
6322unsafe extern "C" {
6323 #[doc = " Returns the supporting status of a partition. Some operations may not be\n supported by the library under certain circumstances. During partitioning\n stage, unsupported partitions will be returned to users with each containing\n an unsupported operation. Users should check the supporting status of a\n partition before transforming the computation graph or compiling the\n partition.\n\n @param partition The target partition.\n @param is_supported Output flag to indicate the supporting status. 0 means\n unsupported while 1 means supported.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6324 pub fn dnnl_graph_partition_is_supported(
6325 partition: const_dnnl_graph_partition_t,
6326 is_supported: *mut u8,
6327 ) -> dnnl_status_t::Type;
6328}
6329unsafe extern "C" {
6330 #[doc = " Returns the engine kind of a partition.\n\n @param partition The target partition.\n @param kind The output engine kind.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6331 pub fn dnnl_graph_partition_get_engine_kind(
6332 partition: const_dnnl_graph_partition_t,
6333 kind: *mut dnnl_engine_kind_t::Type,
6334 ) -> dnnl_status_t::Type;
6335}
6336unsafe extern "C" {
6337 #[doc = " Creates a new compiled partition handle.\n\n @param compiled_partition The handle of output compiled partition.\n @param partition The handle of input partition.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6338 pub fn dnnl_graph_compiled_partition_create(
6339 compiled_partition: *mut dnnl_graph_compiled_partition_t,
6340 partition: dnnl_graph_partition_t,
6341 ) -> dnnl_status_t::Type;
6342}
6343unsafe extern "C" {
6344 #[doc = " Executes a compiled partition.\n\n @param compiled_partition The handle of target compiled partition.\n @param stream The stream used for execution.\n @param num_inputs The number of input tensors.\n @param inputs A list of input tensors.\n @param num_outputs The number of output tensors.\n @param outputs A non-empty list of output tensors.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6345 pub fn dnnl_graph_compiled_partition_execute(
6346 compiled_partition: const_dnnl_graph_compiled_partition_t,
6347 stream: dnnl_stream_t,
6348 num_inputs: usize,
6349 inputs: *mut const_dnnl_graph_tensor_t,
6350 num_outputs: usize,
6351 outputs: *mut const_dnnl_graph_tensor_t,
6352 ) -> dnnl_status_t::Type;
6353}
6354unsafe extern "C" {
6355 #[doc = " Destroys a compiled partition.\n\n @param compiled_partition The compiled partition to be destroyed.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6356 pub fn dnnl_graph_compiled_partition_destroy(
6357 compiled_partition: dnnl_graph_compiled_partition_t,
6358 ) -> dnnl_status_t::Type;
6359}
6360unsafe extern "C" {
6361 #[doc = " Queries an input or output logical tensor according to tensor ID. If the\n tensor ID doesn't belong to any input or output of the compiled partition,\n an error status #dnnl_invalid_arguments will be returned by the API.\n\n @param compiled_partition The handle of target compiled_partition.\n @param tid The unique id of required tensor.\n @param lt The output logical tensor.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6362 pub fn dnnl_graph_compiled_partition_query_logical_tensor(
6363 compiled_partition: const_dnnl_graph_compiled_partition_t,
6364 tid: usize,
6365 lt: *mut dnnl_graph_logical_tensor_t,
6366 ) -> dnnl_status_t::Type;
6367}
6368unsafe extern "C" {
6369 #[doc = " Returns the hint of in-place pairs from a compiled partition. It indicates\n that an input and an output of the partition can share the same memory\n buffer for computation. In-place computation helps to reduce the memory\n footprint and improves cache locality. But since the library may not have a\n global view of user's application, it's possible that the tensor with\n `input_id` is used at other places in user's computation graph. In this\n case, the user should take the in-place pair as a hint and pass a different\n memory buffer for output tensor to avoid overwriting the input memory buffer\n which will probably cause unexpected incorrect results.\n\n @param compiled_partition The handle of target compiled_partition.\n @param num_inplace_pairs The number of in-place pairs.\n @param inplace_pairs The handle of in-place pairs.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6370 pub fn dnnl_graph_compiled_partition_get_inplace_ports(
6371 compiled_partition: const_dnnl_graph_compiled_partition_t,
6372 num_inplace_pairs: *mut usize,
6373 inplace_pairs: *mut *const dnnl_graph_inplace_pair_t,
6374 ) -> dnnl_status_t::Type;
6375}
6376unsafe extern "C" {
6377 #[doc = " Creates a new empty graph. A graph is associated to a specific engine kind.\n The partitions returned from the graph will inherit the engine kind of the\n graph.\n\n @param graph The handle of output graph.\n @param engine_kind The target engine kind.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6378 pub fn dnnl_graph_graph_create(
6379 graph: *mut dnnl_graph_graph_t,
6380 engine_kind: dnnl_engine_kind_t::Type,
6381 ) -> dnnl_status_t::Type;
6382}
6383unsafe extern "C" {
6384 #[doc = " Creates a new empty graph with an engine kind and a floating-point math\n mode. All partitions returned from the graph will inherit the engine kind\n and floating-point math mode.\n\n @param graph The handle of output graph.\n @param engine_kind The kind for engine.\n @param mode The floating-point math mode.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6385 pub fn dnnl_graph_graph_create_with_fpmath_mode(
6386 graph: *mut dnnl_graph_graph_t,
6387 engine_kind: dnnl_engine_kind_t::Type,
6388 mode: dnnl_fpmath_mode_t::Type,
6389 ) -> dnnl_status_t::Type;
6390}
6391unsafe extern "C" {
6392 #[doc = " Destroys a graph.\n\n @param graph The graph to be destroyed.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6393 pub fn dnnl_graph_graph_destroy(graph: dnnl_graph_graph_t) -> dnnl_status_t::Type;
6394}
6395unsafe extern "C" {
6396 #[doc = " Set the floating point math mode for a graph.\n\n @param graph The target graph.\n @param mode The floating-point math mode.\n @param apply_to_int The flag that controls whether to use floating-point\n arithmetic for integral operations.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6397 pub fn dnnl_graph_graph_set_fpmath_mode(
6398 graph: dnnl_graph_graph_t,
6399 mode: dnnl_fpmath_mode_t::Type,
6400 apply_to_int: ::std::os::raw::c_int,
6401 ) -> dnnl_status_t::Type;
6402}
6403unsafe extern "C" {
6404 #[doc = " Get the floating point math mode for a graph.\n\n @param graph The target graph.\n @param mode The floating-point math mode.\n @param apply_to_int The flag that controls whether to use floating-point\n arithmetic for integral operations.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6405 pub fn dnnl_graph_graph_get_fpmath_mode(
6406 graph: dnnl_graph_graph_t,
6407 mode: *mut dnnl_fpmath_mode_t::Type,
6408 apply_to_int: *mut ::std::os::raw::c_int,
6409 ) -> dnnl_status_t::Type;
6410}
6411unsafe extern "C" {
6412 #[doc = " Adds an operation into a graph. The API will return failure if the operator\n has already been added to the graph or the operation cannot pass the schema\n check in the library (eg. input and output numbers and data types, the\n attributes of the operation, etc.).\n\n @param graph The target graph.\n @param op The operation to be added.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6413 pub fn dnnl_graph_add_op(graph: dnnl_graph_graph_t, op: dnnl_graph_op_t)
6414 -> dnnl_status_t::Type;
6415}
6416unsafe extern "C" {
6417 #[doc = " Finalizes a graph. It means users have finished adding operations into the\n graph and the graph is ready for partitioning. Adding a new operation into a\n finalized graph will return failures. Similarly, partitioning on a\n un-finalized graph will also return failures.\n\n @param graph The target graph to be finalized.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6418 pub fn dnnl_graph_graph_finalize(graph: dnnl_graph_graph_t) -> dnnl_status_t::Type;
6419}
6420unsafe extern "C" {
6421 #[doc = " Checks if a graph is finalized.\n\n @param graph The target graph to be finalized.\n @param finalized Output the finalization status. 0 means then graph is not\n finalized. Other values means the graph is finalized.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6422 pub fn dnnl_graph_graph_is_finalized(
6423 graph: dnnl_graph_graph_t,
6424 finalized: *mut u8,
6425 ) -> dnnl_status_t::Type;
6426}
6427unsafe extern "C" {
6428 #[doc = " Filters a graph. Partitions will be claimed internally according to the\n capability of the library, the engine kind, and the policy.\n\n @param graph The target graph.\n @param policy The partition policy.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6429 pub fn dnnl_graph_graph_filter(
6430 graph: dnnl_graph_graph_t,
6431 policy: dnnl_graph_partition_policy_t::Type,
6432 ) -> dnnl_status_t::Type;
6433}
6434unsafe extern "C" {
6435 #[doc = " Returns the number of partitions of a graph. The API should be called after\n a partition is already filtered. Otherwise, the output number is zero.\n\n @param graph The graph.\n @param num Output the number of partitions.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6436 pub fn dnnl_graph_graph_get_partition_num(
6437 graph: const_dnnl_graph_graph_t,
6438 num: *mut usize,
6439 ) -> dnnl_status_t::Type;
6440}
6441unsafe extern "C" {
6442 #[doc = " Returns the partitions from a filtered graph. Output partition instances\n will be written into the parameter `partitions`. Users need to make sure\n `partitions` is valid and has enough space to accept the partition\n instances. Each output partition instance should be destroyed via\n #dnnl_graph_partition_destroy explicitly after use.\n\n @param graph The target graph.\n @param num The number of partitions.\n @param partitions Output the partitions.\n @returns #dnnl_success on success or a status describing the error\n otherwise."]
6443 pub fn dnnl_graph_graph_get_partitions(
6444 graph: dnnl_graph_graph_t,
6445 num: usize,
6446 partitions: *mut dnnl_graph_partition_t,
6447 ) -> dnnl_status_t::Type;
6448}
6449unsafe extern "C" {
6450 #[doc = " Returns the number of compiled partitions that can be held in the compiled\n partition cache at the same time.\n\n @param capacity Compiled partition cache capacity to query. Concurrently\n accessing @p capacity is safe.\n @returns #dnnl_invalid_arguments if the @p capacity value\n is invalid, and #dnnl_success on success."]
6451 pub fn dnnl_graph_get_compiled_partition_cache_capacity(
6452 capacity: *mut ::std::os::raw::c_int,
6453 ) -> dnnl_status_t::Type;
6454}
6455unsafe extern "C" {
6456 #[doc = " Sets a number of compiled partitions that can be held in the compiled\n partition cache at the same time. The default capacity of compiled partition\n cache is 1024.\n\n @param capacity Compiled partition cache capacity to set. The default cache\n capacity is 1024. If a new @p capacity is less than a number of compiled\n partition that the compiled partition cache already has, then the excess\n entries will be evicted. Setting the @p capacity to 0 clears the compiled\n partition cache and disables it. Concurrently modifying @p capacity is safe.\n @returns #dnnl_invalid_arguments if the @p capacity value\n is invalid, and #dnnl_success on success."]
6457 pub fn dnnl_graph_set_compiled_partition_cache_capacity(
6458 capacity: ::std::os::raw::c_int,
6459 ) -> dnnl_status_t::Type;
6460}
6461unsafe extern "C" {
6462 #[doc = " Control the enabling or disabling of constant tensor cache. This API must\n be called once before compilation stage. By default, constant tensor cache is\n disabled in the library.\n\n @param flag Set to positive value to enable the cache and set to 0 to\n disable the cache. Negative values are invalid.\n @returns #dnnl_invalid_arguments if the @p flag value is\n invalid, and #dnnl_success on success.\n @note This API is deprecated and will be removed in future release, please\n use the dnnl_graph_set_constant_tensor_cache_capacity API to disable\n constant tensor cache by setting it's capacity to zero."]
6463 pub fn dnnl_graph_set_constant_tensor_cache(flag: ::std::os::raw::c_int)
6464 -> dnnl_status_t::Type;
6465}
6466unsafe extern "C" {
6467 #[doc = " Return the enabling or disabling status of constant tensor cache.\n\n @param flag The constant tensor cache enabling status to query.\n @returns #dnnl_invalid_arguments if the @p flag value is\n nullptr, and #dnnl_success on success.\n @note This API is deprecated and will be removed in future release, please\n use the dnnl_graph_get_constant_tensor_cache_capacity API to check the\n enabling status by checking it's capacity."]
6468 pub fn dnnl_graph_get_constant_tensor_cache(
6469 flag: *mut ::std::os::raw::c_int,
6470 ) -> dnnl_status_t::Type;
6471}
6472unsafe extern "C" {
6473 #[doc = " Control the capacity for the constant tensor cache that used for specific\n engine kind. This API is thread safe and can be called multiple times at\n runtime. The capacity is set to zero by default which means the cache is\n disabled. When calling this API, the corresponding cache will be flushed.\n Setting capacity to 0 means to clear all cached tensors and disable cache.\n Once the capacity limit is reached, no new tensors will be cached. If there\n are multiple devices for an engine kind, the capacity set here is for each\n device.\n\n @param eng_kind The engine kind that the constant tensor cache used for.\n @param size The constant tensor cache capacity size to set.\n @returns #dnnl_invalid_arguments if the @p eng_kind value is invalid, and\n #dnnl_success on success."]
6474 pub fn dnnl_graph_set_constant_tensor_cache_capacity(
6475 eng_kind: dnnl_engine_kind_t::Type,
6476 size: usize,
6477 ) -> dnnl_status_t::Type;
6478}
6479unsafe extern "C" {
6480 #[doc = " Return the current capacity of constant tensor cache.\n\n @param eng_kind The engine kind that the constant tensor cache used for.\n @param size The constant tensor cache capacity size to query.\n @returns #dnnl_invalid_arguments if the @p eng_kind value is\n nullptr or the @p size is nullptr, and #dnnl_success on success."]
6481 pub fn dnnl_graph_get_constant_tensor_cache_capacity(
6482 eng_kind: dnnl_engine_kind_t::Type,
6483 size: *mut usize,
6484 ) -> dnnl_status_t::Type;
6485}