onednn-src 0.1.13

/*******************************************************************************
* Copyright 2019 Intel Corporation
*
* Licensed under the Apache License, Version 2.0 (the "License");
* you may not use this file except in compliance with the License.
* You may obtain a copy of the License at
*
*     http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*******************************************************************************/

#ifndef GPU_INTEL_SOFTMAX_SIMPLE_HPP
#define GPU_INTEL_SOFTMAX_SIMPLE_HPP

#include "gpu/intel/primitive.hpp"
#include "gpu/intel/primitive_conf.hpp"
#include "gpu/intel/softmax/config.hpp"

namespace dnnl {
namespace impl {
namespace gpu {
namespace intel {
namespace softmax {

struct simple_fwd_t : public primitive_t {
    using primitive_t::primitive_t;
    struct pd_t : public fwd_pd_t {
        using fwd_pd_t::fwd_pd_t;

        DECLARE_COMMON_PD_T("ocl:simple:any", simple_fwd_t);

        bool post_ops_ok() const {
            return attr()->post_ops_.has_default_values(
                    {primitive_kind::eltwise, primitive_kind::binary});
        }

        status_t dropout_ok() const {
            if (attr_.dropout_.has_default_values()) return status::success;

            assert(memory_desc_wrapper(dst_md(0)).format_kind()
                    == format_kind::blocked);

            using namespace format_tag;
            // See `ref_dropout(...)` comment which explains the requirement.
            VDISPATCH_SOFTMAX_IC(memory_desc_matches_one_of_tag(
                                         *dst_md(0), ncdhw, nchw, ncw, nc)
                            && IMPLICATION(attr_.dropout_.has_output_mask(),
                                    memory_desc_wrapper(dst_md(0)).similar_to(
                                            attr_.dropout_.dropout_desc_, true,
                                            false)),
                    VERBOSE_UNSUPPORTED_DROPOUT);

            return status::success;
        }

        status_t init(impl::engine_t *engine) {
            auto *intel_engine = utils::downcast<intel::engine_t *>(engine);

            const memory_desc_wrapper src_d(src_md());
            const memory_desc_wrapper dst_d(dst_md());
            const auto src_dt = src_d.data_type();
            const auto dst_dt = dst_d.data_type();

            using namespace data_type;
            using skip_mask_t = primitive_attr_t::skip_mask_t;
            const int num_blocks = 2;
            auto is_not_double_blk
                    = src_md()->format_desc.blocking.inner_nblks != num_blocks;

            VDISPATCH_SOFTMAX(is_fwd(), VERBOSE_BAD_PROPKIND);
            VDISPATCH_SOFTMAX(
                    utils::one_of(src_dt, f64, f32, f16, bf16, u8, s8),
                    VERBOSE_UNSUPPORTED_DT);
            VDISPATCH_SOFTMAX(utils::one_of(dst_dt, f32, f16, f64, bf16,
                                      f8_e5m2, f8_e4m3, u8, s8),
                    VERBOSE_UNSUPPORTED_DT);
            VDISPATCH_SOFTMAX(IMPLICATION(utils::one_of(f16, src_dt, dst_dt),
                                      intel_engine->mayiuse(
                                              compute::device_ext_t::khr_fp16)),
                    VERBOSE_UNSUPPORTED_DT_CFG);
            VDISPATCH_SOFTMAX(IMPLICATION(utils::one_of(data_type::f64,
                                                  dst_md()->data_type,
                                                  src_md()->data_type),
                                      intel_engine->mayiuse(
                                              compute::device_ext_t::khr_fp64)),
                    VERBOSE_UNSUPPORTED_DT_CFG);
            VDISPATCH_SOFTMAX(intel_engine->mayiuse_sub_group(subgroup_size),
                    VERBOSE_UNSUPPORTED_DEVICE_FEATURE, "subgroup_size");
            VDISPATCH_SOFTMAX(memory_desc_ndims_ok(src_md(), dst_md()),
                    VERBOSE_INCONSISTENT_NDIMS_WITH_VALS, "src", "dst",
                    src_md()->ndims, dst_md()->ndims);
            VDISPATCH_SOFTMAX(
                    attr()->has_default_values(skip_mask_t::scales
                            | skip_mask_t::post_ops | skip_mask_t::dropout),
                    VERBOSE_UNSUPPORTED_ATTR);
            VDISPATCH_SOFTMAX(is_not_double_blk, VERBOSE_UNSUPPORTED_TAG);
            VDISPATCH_SOFTMAX(attr_scales_ok(), VERBOSE_UNSUPPORTED_SCALES_CFG);
            VDISPATCH_SOFTMAX(post_ops_ok(), VERBOSE_UNSUPPORTED_POSTOP);
            VDISPATCH_SOFTMAX_SC(
                    set_default_formats(), VERBOSE_UNSUPPORTED_TAG);
            VDISPATCH_SOFTMAX_SC(attr_.set_default_formats(dst_md(0)),
                    VERBOSE_UNSUPPORTED_POSTOP);
            CHECK(dropout_ok());

            dim_t nelems = axis_size(true);

            if (nelems < subgroup_size) {
                group_size = subgroup_size = 1;
            } else if (nelems <= 100) {
                group_size = subgroup_size * 1;
            } else if (nelems <= 1000) {
                group_size = subgroup_size * 2;
            } else if (nelems <= 2000) {
                group_size = subgroup_size * 4;
            } else if (nelems <= 5000) {
                group_size = subgroup_size * 8;
            } else {
                group_size = subgroup_size * 16;
            }

            for (int i = 0, j = 0; i < src_md()->ndims; ++i) {
                if (i != desc()->softmax_axis) {
                    auto dim = src_md()->padded_dims[i];
                    gws[j % 3] *= dim;
                    if (j < 3) block[j % 3] = dim;
                    j++;
                }
            }

            if (group_size != 1) {
                lws[0] = group_size;
                gws[0] *= group_size;
            }

            return status::success;
        }

        compute::range_t gws = compute::range_t::one();
        compute::range_t lws = compute::range_t::one();
        compute::range_t block = compute::range_t::one();
        size_t group_size = 0;
        int subgroup_size = 16;
    };

    status_t init(impl::engine_t *engine) override {
        if (pd()->has_zero_dim_memory()) return status::success;

        using namespace dnnl::impl::format_tag;
        compute::kernel_ctx_t kernel_ctx;

        const auto *desc = pd()->desc();
        kernel_ctx.define_int("SOFTMAX_AXIS_IDX", desc->softmax_axis);
        kernel_ctx.define_int("SOFTMAX_AXIS", pd()->axis_size(true));
        kernel_ctx.define_int("GROUP_SIZE", pd()->group_size);
        kernel_ctx.define_int("SUB_GROUP_SIZE", pd()->subgroup_size);
        kernel_ctx.define_int("IS_FWD", 1);
        kernel_ctx.define_int(
                "WITH_DROPOUT", !pd()->attr()->dropout_.has_default_values());
        kernel_ctx.define_int(
                "USE_HOST_SCALARS", pd()->attr()->dropout_.use_host_scalars_);
        kernel_ctx.define_int("USE_OFFSET", pd()->attr()->dropout_.use_offset_);
        kernel_ctx.define_int(
                "HAS_OUTPUT_MASK", pd()->attr()->dropout_.has_output_mask());
        kernel_ctx.add_option("-cl-std=CL2.0");
        kernel_ctx.define_int("SOFTMAX_INF_AS_ZERO",
                pd()->alg_kind() == alg_kind::softmax_accurate_inf_as_zero);
        kernel_ctx.define_int("LOGSOFTMAX", pd()->is_logsoftmax());
        kernel_ctx.define_int("WITH_SRC_SCALES",
                !pd()->attr()->scales_.has_default_values(DNNL_ARG_SRC));
        kernel_ctx.define_int("WITH_DST_SCALES",
                !pd()->attr()->scales_.has_default_values(DNNL_ARG_DST));

        const memory_desc_wrapper dst_mdw(pd()->dst_md());
        const memory_desc_wrapper src_mdw(pd()->src_md());
        const auto dst_md_info = memory_desc_info_t::create(dst_mdw);
        const auto src_md_info = memory_desc_info_t::create(src_mdw);
        def_memory_desc_info(kernel_ctx, dst_md_info, "DST");
        def_memory_desc_info(kernel_ctx, src_md_info, "SRC");
        kernel_ctx.set_data_type(dst_mdw.data_type());
        kernel_ctx.require_stateless_addressing(pd()->has_large_buffers());
        set_offsets(kernel_ctx, pd()->dst_md(), "DATA");

        const int ndims = pd()->dst_md()->ndims;
        const dim_t OC = pd()->dst_md()->dims[1];
        dim_t spatial_dims_size = 1;
        for (int i = 2; i < ndims; i++) {
            spatial_dims_size *= pd()->dst_md()->dims[i];
        }
        kernel_ctx.define_int("OC", OC);
        kernel_ctx.define_int("SPATIAL_DIMS_SIZE", spatial_dims_size);
        kernel_ctx.define_int("NDIMS", ndims);
        kernel_ctx.define_int("SPATIAL_DIM_0", pd()->dst_md()->dims[2]);
        if (ndims > 3) {
            kernel_ctx.define_int("SPATIAL_DIM_1", pd()->dst_md()->dims[3]);
        }
        if (ndims > 4) {
            kernel_ctx.define_int("SPATIAL_DIM_2", pd()->dst_md()->dims[4]);
        }
        kernel_ctx.define_int("IS_CHANNEL_LAST",
                dst_mdw.matches_one_of_tag(nwc, nhwc, ndhwc));

        CHECK(def_attr_info(kernel_ctx, attr_info_t::create(pd()->attr()),
                pd()->attr()->post_ops_, *pd()->invariant_dst_md()));

        for (int i = 0; i < 3; i++)
            kernel_ctx.define_int(utils::format("BLOCK_%d", i), pd()->block[i]);

        CHECK(create_kernel(
                engine, &kernel_, "simple_softmax_fwd_generic", kernel_ctx));
        if (!kernel_) return status::runtime_error;

        return status::success;
    }

    status_t execute(const exec_ctx_t &ctx) const override {
        return execute_generic(ctx);
    }

protected:
    status_t execute_generic(const exec_ctx_t &ctx) const;
    const pd_t *pd() const { return (const pd_t *)primitive_t::pd().get(); }
    compute::kernel_t kernel_;
};

struct simple_bwd_t : public primitive_t {
    using primitive_t::primitive_t;
    struct pd_t : public bwd_pd_t {
        using bwd_pd_t::bwd_pd_t;

        DECLARE_COMMON_PD_T("ocl:simple:any", simple_bwd_t);

        status_t init(impl::engine_t *engine) {
            auto *intel_engine = utils::downcast<intel::engine_t *>(engine);

            const memory_desc_wrapper diff_dst_d(diff_dst_md());
            const memory_desc_wrapper diff_src_d(diff_src_md());
            const memory_desc_wrapper dst_d(dst_md());

            using namespace data_type;
            VDISPATCH_SOFTMAX(!is_fwd(), VERBOSE_BAD_PROPKIND);
            VDISPATCH_SOFTMAX(
                    utils::one_of(diff_src_d.data_type(), f64, f32, bf16, f16),
                    VERBOSE_UNSUPPORTED_DT);
            VDISPATCH_SOFTMAX(
                    utils::one_of(diff_dst_d.data_type(), f64, f32, bf16, f16),
                    VERBOSE_UNSUPPORTED_DT);
            VDISPATCH_SOFTMAX(IMPLICATION(utils::one_of(data_type::f64,
                                                  diff_dst_md()->data_type,
                                                  diff_src_md()->data_type),
                                      intel_engine->mayiuse(
                                              compute::device_ext_t::khr_fp64)),
                    VERBOSE_UNSUPPORTED_DT_CFG);
            VDISPATCH_SOFTMAX(IMPLICATION(utils::one_of(data_type::f16,
                                                  diff_dst_md()->data_type,
                                                  diff_src_md()->data_type),
                                      intel_engine->mayiuse(
                                              compute::device_ext_t::khr_fp16)),
                    VERBOSE_UNSUPPORTED_DT_CFG);
            VDISPATCH_SOFTMAX(intel_engine->mayiuse_sub_group(16),
                    VERBOSE_UNSUPPORTED_DEVICE_FEATURE, "subgroup_size");
            VDISPATCH_SOFTMAX(memory_desc_ndims_ok(
                                      dst_md(), diff_src_md(), diff_dst_md()),
                    VERBOSE_INCONSISTENT_NDIMS_WITH_VALS, "dst, diff_src",
                    "diff_dst", dst_md()->ndims, diff_dst_md()->ndims);
            VDISPATCH_SOFTMAX(
                    attr()->has_default_values(), VERBOSE_UNSUPPORTED_ATTR);
            VDISPATCH_SOFTMAX_SC(
                    set_default_formats(), VERBOSE_UNSUPPORTED_TAG);
            VDISPATCH_SOFTMAX(diff_dst_d.data_type() == dst_d.data_type(),
                    VERBOSE_INCONSISTENT_DT, "diff_dst_d", "dst_d");

            for (int i = 0, j = 0; i < dst_d.ndims(); ++i) {
                if (i != axis()) {
                    auto dim = dst_d.padded_dims()[i];
                    gws[j % 3] *= dim;
                    if (j < 3) block[j % 3] = dim;
                    j++;
                }
            }

            dim_t nelems = axis_size(true);
            if (nelems <= 100) {
                group_size = 16;
            } else if (nelems <= 1000) {
                group_size = 32;
            } else if (nelems <= 2000) {
                group_size = 64;
            } else if (nelems <= 5000) {
                group_size = 128;
            } else {
                group_size = 256;
            }

            lws[0] = group_size;
            gws[0] *= group_size;

            return status::success;
        }

        compute::range_t gws = compute::range_t::one();
        compute::range_t lws = compute::range_t::one();
        compute::range_t block = compute::range_t::one();
        size_t group_size = 0;
    };

    status_t init(impl::engine_t *engine) override {
        if (pd()->has_zero_dim_memory()) return status::success;

        compute::kernel_ctx_t kernel_ctx;

        kernel_ctx.define_int("SOFTMAX_AXIS_IDX", pd()->axis());
        kernel_ctx.define_int("SOFTMAX_AXIS", pd()->axis_size(true));
        kernel_ctx.define_int("GROUP_SIZE", pd()->group_size);
        kernel_ctx.define_int("SUB_GROUP_SIZE", 16);
        kernel_ctx.define_int("IS_BWD", 1);
        kernel_ctx.add_option("-cl-std=CL2.0");
        kernel_ctx.define_int("LOGSOFTMAX", pd()->is_logsoftmax());

        const memory_desc_wrapper diff_src_mdw(pd()->diff_src_md());
        const memory_desc_wrapper diff_dst_mdw(pd()->diff_dst_md());
        const auto diff_src_md_info = memory_desc_info_t::create(diff_src_mdw);
        const auto diff_dst_md_info = memory_desc_info_t::create(diff_dst_mdw);
        def_memory_desc_info(kernel_ctx, diff_src_md_info, "SRC");
        def_memory_desc_info(kernel_ctx, diff_dst_md_info, "DST");
        kernel_ctx.set_data_type(diff_src_mdw.data_type());
        kernel_ctx.register_buffer_size(*pd()->dst_md());
        kernel_ctx.require_stateless_addressing(pd()->has_large_buffers());
        set_offsets(kernel_ctx, *pd()->diff_src_md(), "DATA");

        for (int i = 0; i < 3; i++)
            kernel_ctx.define_int(utils::format("BLOCK_%d", i), pd()->block[i]);

        CHECK(create_kernel(
                engine, &kernel_, "simple_softmax_bwd_generic", kernel_ctx));
        if (!kernel_) return status::runtime_error;

        return status::success;
    }

    status_t execute(const exec_ctx_t &ctx) const override {
        return execute_generic(ctx);
    }

protected:
    status_t execute_generic(const exec_ctx_t &ctx) const;
    const pd_t *pd() const { return (const pd_t *)primitive_t::pd().get(); }
    compute::kernel_t kernel_;
};

} // namespace softmax
} // namespace intel
} // namespace gpu
} // namespace impl
} // namespace dnnl

#endif