fearless_simd 0.4.0

Safer and easier SIMD
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299

// Copyright 2024 the Fearless_SIMD Authors
// SPDX-License-Identifier: Apache-2.0 OR MIT

//! Access to SSE2 intrinsics.

use crate::impl_macros::delegate;
#[cfg(target_arch = "x86")]
use core::arch::x86 as arch;
#[cfg(target_arch = "x86_64")]
use core::arch::x86_64 as arch;

use arch::*;

/// A token for SSE2 intrinsics on `x86` and `x86_64`.
#[derive(Clone, Copy, Debug)]
pub struct Sse2 {
    _private: (),
}

#[expect(
    clippy::missing_safety_doc,
    reason = "TODO: https://github.com/linebender/fearless_simd/issues/40"
)]
impl Sse2 {
    /// Create a SIMD token.
    ///
    /// # Safety
    ///
    /// The required CPU features must be available.
    #[inline]
    pub const unsafe fn new_unchecked() -> Self {
        Self { _private: () }
    }

    delegate! { arch:
        fn _mm_pause();
        unsafe fn _mm_clflush(p: *const u8);
        fn _mm_lfence();
        fn _mm_mfence();
        fn _mm_add_epi8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_add_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_add_epi32(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_add_epi64(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_adds_epi8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_adds_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_adds_epu8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_adds_epu16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_avg_epu8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_avg_epu16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_madd_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_max_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_max_epu8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_min_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_min_epu8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_mulhi_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_mulhi_epu16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_mullo_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_mul_epu32(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_sad_epu8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_sub_epi8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_sub_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_sub_epi32(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_sub_epi64(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_subs_epi8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_subs_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_subs_epu8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_subs_epu16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_slli_si128<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_bslli_si128<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_bsrli_si128<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_slli_epi16<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_sll_epi16(a: __m128i, count: __m128i) -> __m128i;
        fn _mm_slli_epi32<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_sll_epi32(a: __m128i, count: __m128i) -> __m128i;
        fn _mm_slli_epi64<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_sll_epi64(a: __m128i, count: __m128i) -> __m128i;
        fn _mm_srai_epi16<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_sra_epi16(a: __m128i, count: __m128i) -> __m128i;
        fn _mm_srai_epi32<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_sra_epi32(a: __m128i, count: __m128i) -> __m128i;
        fn _mm_srli_si128<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_srli_epi16<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_srl_epi16(a: __m128i, count: __m128i) -> __m128i;
        fn _mm_srli_epi32<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_srl_epi32(a: __m128i, count: __m128i) -> __m128i;
        fn _mm_srli_epi64<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_srl_epi64(a: __m128i, count: __m128i) -> __m128i;
        fn _mm_and_si128(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_andnot_si128(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_or_si128(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_xor_si128(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_cmpeq_epi8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_cmpeq_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_cmpeq_epi32(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_cmpgt_epi8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_cmpgt_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_cmpgt_epi32(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_cmplt_epi8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_cmplt_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_cmplt_epi32(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_cvtepi32_pd(a: __m128i) -> __m128d;
        fn _mm_cvtsi32_sd(a: __m128d, b: i32) -> __m128d;
        fn _mm_cvtepi32_ps(a: __m128i) -> __m128;
        fn _mm_cvtps_epi32(a: __m128) -> __m128i;
        fn _mm_cvtsi32_si128(a: i32) -> __m128i;
        fn _mm_cvtsi128_si32(a: __m128i) -> i32;
        fn _mm_set_epi64x(e1: i64, e0: i64) -> __m128i;
        fn _mm_set_epi32(e3: i32, e2: i32, e1: i32, e0: i32) -> __m128i;
        fn _mm_set_epi16(
            e7: i16,
            e6: i16,
            e5: i16,
            e4: i16,
            e3: i16,
            e2: i16,
            e1: i16,
            e0: i16,
        ) -> __m128i;
        fn _mm_set_epi8(
            e15: i8,
            e14: i8,
            e13: i8,
            e12: i8,
            e11: i8,
            e10: i8,
            e9: i8,
            e8: i8,
            e7: i8,
            e6: i8,
            e5: i8,
            e4: i8,
            e3: i8,
            e2: i8,
            e1: i8,
            e0: i8,
        ) -> __m128i;
        fn _mm_set1_epi64x(a: i64) -> __m128i;
        fn _mm_set1_epi32(a: i32) -> __m128i;
        fn _mm_set1_epi16(a: i16) -> __m128i;
        fn _mm_set1_epi8(a: i8) -> __m128i;
        fn _mm_setr_epi32(e3: i32, e2: i32, e1: i32, e0: i32) -> __m128i;
        fn _mm_setr_epi16(
            e7: i16,
            e6: i16,
            e5: i16,
            e4: i16,
            e3: i16,
            e2: i16,
            e1: i16,
            e0: i16,
        ) -> __m128i;
        fn _mm_setr_epi8(
            e15: i8,
            e14: i8,
            e13: i8,
            e12: i8,
            e11: i8,
            e10: i8,
            e9: i8,
            e8: i8,
            e7: i8,
            e6: i8,
            e5: i8,
            e4: i8,
            e3: i8,
            e2: i8,
            e1: i8,
            e0: i8,
        ) -> __m128i;
        fn _mm_setzero_si128() -> __m128i;
        unsafe fn _mm_loadl_epi64(mem_addr: *const __m128i) -> __m128i;
        unsafe fn _mm_load_si128(mem_addr: *const __m128i) -> __m128i;
        unsafe fn _mm_loadu_si128(mem_addr: *const __m128i) -> __m128i;
        unsafe fn _mm_maskmoveu_si128(a: __m128i, mask: __m128i, mem_addr: *mut i8);
        unsafe fn _mm_store_si128(mem_addr: *mut __m128i, a: __m128i);
        unsafe fn _mm_storeu_si128(mem_addr: *mut __m128i, a: __m128i);
        unsafe fn _mm_storel_epi64(mem_addr: *mut __m128i, a: __m128i);
        unsafe fn _mm_stream_si128(mem_addr: *mut __m128i, a: __m128i);
        unsafe fn _mm_stream_si32(mem_addr: *mut i32, a: i32);
        fn _mm_move_epi64(a: __m128i) -> __m128i;
        fn _mm_packs_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_packs_epi32(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_packus_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_extract_epi16<const IMM8: i32>(a: __m128i) -> i32;
        fn _mm_insert_epi16<const IMM8: i32>(a: __m128i, i: i32) -> __m128i;
        fn _mm_movemask_epi8(a: __m128i) -> i32;
        fn _mm_shuffle_epi32<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_shufflehi_epi16<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_shufflelo_epi16<const IMM8: i32>(a: __m128i) -> __m128i;
        fn _mm_unpackhi_epi8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_unpackhi_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_unpackhi_epi32(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_unpackhi_epi64(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_unpacklo_epi8(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_unpacklo_epi16(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_unpacklo_epi32(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_unpacklo_epi64(a: __m128i, b: __m128i) -> __m128i;
        fn _mm_add_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_add_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_div_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_div_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_max_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_max_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_min_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_min_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_mul_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_mul_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_sqrt_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_sqrt_pd(a: __m128d) -> __m128d;
        fn _mm_sub_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_sub_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_and_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_andnot_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_or_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_xor_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpeq_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmplt_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmple_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpgt_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpge_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpord_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpunord_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpneq_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpnlt_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpnle_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpngt_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpnge_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpeq_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmplt_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmple_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpgt_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpge_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpord_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpunord_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpneq_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpnlt_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpnle_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpngt_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_cmpnge_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_comieq_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_comilt_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_comile_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_comigt_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_comige_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_comineq_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_ucomieq_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_ucomilt_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_ucomile_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_ucomigt_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_ucomige_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_ucomineq_sd(a: __m128d, b: __m128d) -> i32;
        fn _mm_cvtpd_ps(a: __m128d) -> __m128;
        fn _mm_cvtps_pd(a: __m128) -> __m128d;
        fn _mm_cvtpd_epi32(a: __m128d) -> __m128i;
        fn _mm_cvtsd_si32(a: __m128d) -> i32;
        fn _mm_cvtsd_ss(a: __m128, b: __m128d) -> __m128;
        fn _mm_cvtsd_f64(a: __m128d) -> f64;
        fn _mm_cvtss_sd(a: __m128d, b: __m128) -> __m128d;
        fn _mm_cvttpd_epi32(a: __m128d) -> __m128i;
        fn _mm_cvttsd_si32(a: __m128d) -> i32;
        fn _mm_cvttps_epi32(a: __m128) -> __m128i;
        fn _mm_set_sd(a: f64) -> __m128d;
        fn _mm_set1_pd(a: f64) -> __m128d;
        fn _mm_set_pd1(a: f64) -> __m128d;
        fn _mm_set_pd(a: f64, b: f64) -> __m128d;
        fn _mm_setr_pd(a: f64, b: f64) -> __m128d;
        fn _mm_setzero_pd() -> __m128d;
        fn _mm_movemask_pd(a: __m128d) -> i32;
        unsafe fn _mm_load_pd(mem_addr: *const f64) -> __m128d;
        unsafe fn _mm_load_sd(mem_addr: *const f64) -> __m128d;
        unsafe fn _mm_loadh_pd(a: __m128d, mem_addr: *const f64) -> __m128d;
        unsafe fn _mm_loadl_pd(a: __m128d, mem_addr: *const f64) -> __m128d;
        unsafe fn _mm_stream_pd(mem_addr: *mut f64, a: __m128d);
        unsafe fn _mm_store_sd(mem_addr: *mut f64, a: __m128d);
        unsafe fn _mm_store_pd(mem_addr: *mut f64, a: __m128d);
        unsafe fn _mm_storeu_pd(mem_addr: *mut f64, a: __m128d);
        unsafe fn _mm_store1_pd(mem_addr: *mut f64, a: __m128d);
        unsafe fn _mm_store_pd1(mem_addr: *mut f64, a: __m128d);
        unsafe fn _mm_storer_pd(mem_addr: *mut f64, a: __m128d);
        unsafe fn _mm_storeh_pd(mem_addr: *mut f64, a: __m128d);
        unsafe fn _mm_storel_pd(mem_addr: *mut f64, a: __m128d);
        unsafe fn _mm_load1_pd(mem_addr: *const f64) -> __m128d;
        unsafe fn _mm_load_pd1(mem_addr: *const f64) -> __m128d;
        unsafe fn _mm_loadr_pd(mem_addr: *const f64) -> __m128d;
        unsafe fn _mm_loadu_pd(mem_addr: *const f64) -> __m128d;
        fn _mm_shuffle_pd<const MASK: i32>(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_move_sd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_castpd_ps(a: __m128d) -> __m128;
        fn _mm_castpd_si128(a: __m128d) -> __m128i;
        fn _mm_castps_pd(a: __m128) -> __m128d;
        fn _mm_castps_si128(a: __m128) -> __m128i;
        fn _mm_castsi128_pd(a: __m128i) -> __m128d;
        fn _mm_castsi128_ps(a: __m128i) -> __m128;
        fn _mm_undefined_pd() -> __m128d;
        fn _mm_undefined_si128() -> __m128i;
        fn _mm_unpackhi_pd(a: __m128d, b: __m128d) -> __m128d;
        fn _mm_unpacklo_pd(a: __m128d, b: __m128d) -> __m128d;
    }
}