chromium/third_party/xnnpack/src/src/amalgam/gen/avx512fp16.c

// Copyright 2021 Google LLC
//
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree.
//
// Auto-generated file. Do not edit!
//   Generator: tools/update-microkernels.py -a

#include <assert.h>

#include <immintrin.h>

#include "xnnpack/common.h"
#include "xnnpack/gemm.h"
#include "xnnpack/igemm.h"
#include "xnnpack/intrinsics-polyfill.h"
#include "xnnpack/reduce.h"
#include "xnnpack/vbinary.h"


void xnn_f16_gemm_minmax_ukernel_1x64__avx512fp16_broadcast(
    size_t mr,
    size_t nc,
    size_t kc,
    const void* restrict a,
    size_t a_stride,
    const void* restrict w,
    void* restrict c,
    size_t cm_stride,
    size_t cn_stride,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_gemm_minmax_ukernel_7x64__avx512fp16_broadcast(
    size_t mr,
    size_t nc,
    size_t kc,
    const void* restrict a,
    size_t a_stride,
    const void* restrict w,
    void* restrict c,
    size_t cm_stride,
    size_t cn_stride,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_igemm_minmax_ukernel_1x64__avx512fp16_broadcast(
    size_t mr,
    size_t nc,
    size_t kc,
    size_t ks,
    const void** restrict a,
    const void* restrict w,
    void* restrict c,
    size_t cm_stride,
    size_t cn_stride,
    size_t a_offset,
    const void* zero,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_igemm_minmax_ukernel_7x64__avx512fp16_broadcast(
    size_t mr,
    size_t nc,
    size_t kc,
    size_t ks,
    const void** restrict a,
    const void* restrict w,
    void* restrict c,
    size_t cm_stride,
    size_t cn_stride,
    size_t a_offset,
    const void* zero,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_rmax_ukernel__avx512fp16_u128_acc4(
    size_t batch,
    const void* input,
    void* output,
    const union xnn_f16_default_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_rminmax_ukernel__avx512fp16_u128_acc4(
    size_t batch,
    const void* input,
    void* output,
    const union xnn_f16_default_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vadd_minmax_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vaddc_minmax_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vdiv_minmax_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vdivc_minmax_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vmax_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_default_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vmaxc_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_default_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vmin_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_default_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vminc_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_default_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vmul_minmax_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vmulc_minmax_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vrdivc_minmax_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vrsubc_minmax_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vsqrdiff_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_default_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vsqrdiffc_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_default_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vsub_minmax_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }

void xnn_f16_vsubc_minmax_ukernel__avx512fp16_u64(
    size_t batch,
    const void* restrict input_a,
    const void* restrict input_b,
    void* restrict output,
    const union xnn_f16_minmax_params params[restrict XNN_MIN_ELEMENTS(1)])
{ … }