// Auto-generated file. Do not edit! // Template: src/qs8-igemm/MRx8c8-avxvnni.c.in // Generator: tools/xngen // // Copyright 2024 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. #include #include #include "xnnpack/common.h" #include "xnnpack/gemm.h" #include "xnnpack/intrinsics-polyfill.h" #include "xnnpack/math.h" #include "xnnpack/unaligned.h" void xnn_qs8_qc8w_igemm_minmax_fp32_ukernel_5x8c8__avx256vnni( size_t mr, size_t nc, size_t kc, size_t ks, const int8_t** restrict a, const void* restrict w, int8_t* restrict c, size_t cm_stride, size_t cn_stride, size_t a_offset, const int8_t* zero, const union xnn_qs8_qc8w_conv_minmax_params params[restrict XNN_MIN_ELEMENTS(1)]) XNN_OOB_READS { assert(mr != 0); assert(mr <= 5); assert(nc != 0); assert(kc != 0); assert(kc % sizeof(int8_t) == 0); assert(a != NULL); assert(w != NULL); assert(c != NULL); kc = round_up_po2(kc, 8 * sizeof(int8_t)); int8_t* c0 = c; int8_t* c1 = (int8_t*) ((uintptr_t) c0 + cm_stride); if XNN_UNPREDICTABLE(mr < 2) { c1 = c0; } int8_t* c2 = (int8_t*) ((uintptr_t) c1 + cm_stride); if XNN_UNPREDICTABLE(mr <= 2) { c2 = c1; } int8_t* c3 = (int8_t*) ((uintptr_t) c2 + cm_stride); if XNN_UNPREDICTABLE(mr < 4) { c3 = c2; } int8_t* c4 = (int8_t*) ((uintptr_t) c3 + cm_stride); if XNN_UNPREDICTABLE(mr <= 4) { c4 = c3; } const __m256i vsign_mask = _mm256_set1_epi8(0x80); XNN_FORCE_REALIZATION(vsign_mask); const __m256 voutput_max_less_zero_point = _mm256_set1_ps((int32_t) params->fp32_scalar.output_max - (int32_t) params->fp32_scalar.output_zero_point); const __m256i voutput_zero_point = _mm256_set1_epi32(params->fp32_scalar.output_zero_point); const __m128i voutput_min = _mm_set1_epi8(params->fp32_scalar.output_min); // XNN_FORCE_REALIZATION(voutput_max_less_zero_point); // XNN_FORCE_REALIZATION(voutput_zero_point); // XNN_FORCE_REALIZATION(voutput_min); do { __m256i vacc0x0123 = _mm256_cvtepu32_epi64(_mm_load_si128((const __m128i*) w)); __m256i vacc0x4567 = _mm256_cvtepu32_epi64(_mm_load_si128((const __m128i*) ((const int32_t*) w + 4))); __m256i vacc1x0123 = vacc0x0123; __m256i vacc1x4567 = vacc0x4567; __m256i vacc2x0123 = vacc0x0123; __m256i vacc2x4567 = vacc0x4567; __m256i vacc3x0123 = vacc0x0123; __m256i vacc3x4567 = vacc0x4567; __m256i vacc4x0123 = vacc0x0123; __m256i vacc4x4567 = vacc0x4567; w = (const int32_t*) w + 8; size_t p = ks; do { const int8_t* restrict a0 = a[0]; if XNN_UNPREDICTABLE(a0 != zero) { a0 = (const int8_t*) ((uintptr_t) a0 + a_offset); } const int8_t* restrict a1 = a[1]; if XNN_UNPREDICTABLE(a1 != zero) { a1 = (const int8_t*) ((uintptr_t) a1 + a_offset); } const int8_t* restrict a2 = a[2]; if XNN_UNPREDICTABLE(a2 != zero) { a2 = (const int8_t*) ((uintptr_t) a2 + a_offset); } const int8_t* restrict a3 = a[3]; if XNN_UNPREDICTABLE(a3 != zero) { a3 = (const int8_t*) ((uintptr_t) a3 + a_offset); } const int8_t* restrict a4 = a[4]; if XNN_UNPREDICTABLE(a4 != zero) { a4 = (const int8_t*) ((uintptr_t) a4 + a_offset); } a += 5; size_t k = kc; while (k >= 16 * sizeof(int8_t)) { const __m256i va0x01234567 = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a0)), vsign_mask); const __m256i va0x89ABCDEF = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a0 + 8)), vsign_mask); a0 += 16; const __m256i va1x01234567 = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a1)), vsign_mask); const __m256i va1x89ABCDEF = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a1 + 8)), vsign_mask); a1 += 16; const __m256i va2x01234567 = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a2)), vsign_mask); const __m256i va2x89ABCDEF = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a2 + 8)), vsign_mask); a2 += 16; const __m256i va3x01234567 = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a3)), vsign_mask); const __m256i va3x89ABCDEF = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a3 + 8)), vsign_mask); a3 += 16; const __m256i va4x01234567 = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a4)), vsign_mask); const __m256i va4x89ABCDEF = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a4 + 8)), vsign_mask); a4 += 16; const __m256i vb01234567x0123 = _mm256_load_si256(w); const __m256i vb89ABCDEFx0123 = _mm256_load_si256((const __m256i*) ((const int8_t*) w + 32)); const __m256i vb01234567x4567 = _mm256_load_si256((const __m256i*) ((const int8_t*) w + 64)); const __m256i vb89ABCDEFx4567 = _mm256_load_si256((const __m256i*) ((const int8_t*) w + 96)); vacc0x0123 = _mm256_dpbusd_epi32(vacc0x0123, va0x01234567, vb01234567x0123); vacc0x4567 = _mm256_dpbusd_epi32(vacc0x4567, va0x01234567, vb89ABCDEFx0123); vacc1x0123 = _mm256_dpbusd_epi32(vacc1x0123, va1x01234567, vb01234567x0123); vacc1x4567 = _mm256_dpbusd_epi32(vacc1x4567, va1x01234567, vb89ABCDEFx0123); vacc2x0123 = _mm256_dpbusd_epi32(vacc2x0123, va2x01234567, vb01234567x0123); vacc2x4567 = _mm256_dpbusd_epi32(vacc2x4567, va2x01234567, vb89ABCDEFx0123); vacc3x0123 = _mm256_dpbusd_epi32(vacc3x0123, va3x01234567, vb01234567x0123); vacc3x4567 = _mm256_dpbusd_epi32(vacc3x4567, va3x01234567, vb89ABCDEFx0123); vacc4x0123 = _mm256_dpbusd_epi32(vacc4x0123, va4x01234567, vb01234567x0123); vacc4x4567 = _mm256_dpbusd_epi32(vacc4x4567, va4x01234567, vb89ABCDEFx0123); vacc0x0123 = _mm256_dpbusd_epi32(vacc0x0123, va0x89ABCDEF, vb01234567x4567); vacc0x4567 = _mm256_dpbusd_epi32(vacc0x4567, va0x89ABCDEF, vb89ABCDEFx4567); vacc1x0123 = _mm256_dpbusd_epi32(vacc1x0123, va1x89ABCDEF, vb01234567x4567); vacc1x4567 = _mm256_dpbusd_epi32(vacc1x4567, va1x89ABCDEF, vb89ABCDEFx4567); vacc2x0123 = _mm256_dpbusd_epi32(vacc2x0123, va2x89ABCDEF, vb01234567x4567); vacc2x4567 = _mm256_dpbusd_epi32(vacc2x4567, va2x89ABCDEF, vb89ABCDEFx4567); vacc3x0123 = _mm256_dpbusd_epi32(vacc3x0123, va3x89ABCDEF, vb01234567x4567); vacc3x4567 = _mm256_dpbusd_epi32(vacc3x4567, va3x89ABCDEF, vb89ABCDEFx4567); vacc4x0123 = _mm256_dpbusd_epi32(vacc4x0123, va4x89ABCDEF, vb01234567x4567); vacc4x4567 = _mm256_dpbusd_epi32(vacc4x4567, va4x89ABCDEF, vb89ABCDEFx4567); w = (const int8_t*) w + 128; k -= 16 * sizeof(int8_t); } if (k != 0) { const __m256i va0x01234567 = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a0)), vsign_mask); a0 += 8; const __m256i va1x01234567 = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a1)), vsign_mask); a1 += 8; const __m256i va2x01234567 = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a2)), vsign_mask); a2 += 8; const __m256i va3x01234567 = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a3)), vsign_mask); a3 += 8; const __m256i va4x01234567 = _mm256_xor_si256(_mm256_set1_epi64x((int64_t) unaligned_load_u64(a4)), vsign_mask); a4 += 8; const __m256i vb01234567x0123 = _mm256_load_si256(w); const __m256i vb89ABCDEFx0123 = _mm256_load_si256((const __m256i*) ((const int8_t*) w + 32)); vacc0x0123 = _mm256_dpbusd_epi32(vacc0x0123, va0x01234567, vb01234567x0123); vacc0x4567 = _mm256_dpbusd_epi32(vacc0x4567, va0x01234567, vb89ABCDEFx0123); vacc1x0123 = _mm256_dpbusd_epi32(vacc1x0123, va1x01234567, vb01234567x0123); vacc1x4567 = _mm256_dpbusd_epi32(vacc1x4567, va1x01234567, vb89ABCDEFx0123); vacc2x0123 = _mm256_dpbusd_epi32(vacc2x0123, va2x01234567, vb01234567x0123); vacc2x4567 = _mm256_dpbusd_epi32(vacc2x4567, va2x01234567, vb89ABCDEFx0123); vacc3x0123 = _mm256_dpbusd_epi32(vacc3x0123, va3x01234567, vb01234567x0123); vacc3x4567 = _mm256_dpbusd_epi32(vacc3x4567, va3x01234567, vb89ABCDEFx0123); vacc4x0123 = _mm256_dpbusd_epi32(vacc4x0123, va4x01234567, vb01234567x0123); vacc4x4567 = _mm256_dpbusd_epi32(vacc4x4567, va4x01234567, vb89ABCDEFx0123); w = (const int8_t*) w + 64; k -= 8 * sizeof(int8_t); } p -= 5 * sizeof(void*); } while (p != 0); // Add adjacent pairs const __m256i vsum0x02134657 = _mm256_hadd_epi32(vacc0x0123, vacc0x4567); __m256i vacc0x01234567 = _mm256_permute4x64_epi64(vsum0x02134657, _MM_SHUFFLE(3, 1, 2, 0)); const __m256i vsum1x02134657 = _mm256_hadd_epi32(vacc1x0123, vacc1x4567); __m256i vacc1x01234567 = _mm256_permute4x64_epi64(vsum1x02134657, _MM_SHUFFLE(3, 1, 2, 0)); const __m256i vsum2x02134657 = _mm256_hadd_epi32(vacc2x0123, vacc2x4567); __m256i vacc2x01234567 = _mm256_permute4x64_epi64(vsum2x02134657, _MM_SHUFFLE(3, 1, 2, 0)); const __m256i vsum3x02134657 = _mm256_hadd_epi32(vacc3x0123, vacc3x4567); __m256i vacc3x01234567 = _mm256_permute4x64_epi64(vsum3x02134657, _MM_SHUFFLE(3, 1, 2, 0)); const __m256i vsum4x02134657 = _mm256_hadd_epi32(vacc4x0123, vacc4x4567); __m256i vacc4x01234567 = _mm256_permute4x64_epi64(vsum4x02134657, _MM_SHUFFLE(3, 1, 2, 0)); __m256 vout0x01234567 = _mm256_cvtepi32_ps(vacc0x01234567); __m256 vout1x01234567 = _mm256_cvtepi32_ps(vacc1x01234567); __m256 vout2x01234567 = _mm256_cvtepi32_ps(vacc2x01234567); __m256 vout3x01234567 = _mm256_cvtepi32_ps(vacc3x01234567); __m256 vout4x01234567 = _mm256_cvtepi32_ps(vacc4x01234567); const __m256 vscale01234567 = _mm256_load_ps(w); w = (const float*) w + 8; vout0x01234567 = _mm256_mul_ps(vout0x01234567, vscale01234567); vout1x01234567 = _mm256_mul_ps(vout1x01234567, vscale01234567); vout2x01234567 = _mm256_mul_ps(vout2x01234567, vscale01234567); vout3x01234567 = _mm256_mul_ps(vout3x01234567, vscale01234567); vout4x01234567 = _mm256_mul_ps(vout4x01234567, vscale01234567); vout0x01234567 = _mm256_min_ps(vout0x01234567, voutput_max_less_zero_point); vout1x01234567 = _mm256_min_ps(vout1x01234567, voutput_max_less_zero_point); vout2x01234567 = _mm256_min_ps(vout2x01234567, voutput_max_less_zero_point); vout3x01234567 = _mm256_min_ps(vout3x01234567, voutput_max_less_zero_point); vout4x01234567 = _mm256_min_ps(vout4x01234567, voutput_max_less_zero_point); vacc0x01234567 = _mm256_cvtps_epi32(vout0x01234567); vacc1x01234567 = _mm256_cvtps_epi32(vout1x01234567); vacc2x01234567 = _mm256_cvtps_epi32(vout2x01234567); vacc3x01234567 = _mm256_cvtps_epi32(vout3x01234567); vacc4x01234567 = _mm256_cvtps_epi32(vout4x01234567); vacc0x01234567 = _mm256_add_epi32(vacc0x01234567, voutput_zero_point); vacc1x01234567 = _mm256_add_epi32(vacc1x01234567, voutput_zero_point); vacc2x01234567 = _mm256_add_epi32(vacc2x01234567, voutput_zero_point); vacc3x01234567 = _mm256_add_epi32(vacc3x01234567, voutput_zero_point); vacc4x01234567 = _mm256_add_epi32(vacc4x01234567, voutput_zero_point); __m128i voutb0x01234567 = _mm256_cvtsepi32_epi8(vacc0x01234567); __m128i voutb1x01234567 = _mm256_cvtsepi32_epi8(vacc1x01234567); __m128i voutb2x01234567 = _mm256_cvtsepi32_epi8(vacc2x01234567); __m128i voutb3x01234567 = _mm256_cvtsepi32_epi8(vacc3x01234567); __m128i voutb4x01234567 = _mm256_cvtsepi32_epi8(vacc4x01234567); voutb0x01234567 = _mm_max_epi8(voutb0x01234567, voutput_min); voutb1x01234567 = _mm_max_epi8(voutb1x01234567, voutput_min); voutb2x01234567 = _mm_max_epi8(voutb2x01234567, voutput_min); voutb3x01234567 = _mm_max_epi8(voutb3x01234567, voutput_min); voutb4x01234567 = _mm_max_epi8(voutb4x01234567, voutput_min); if (nc >= 8) { _mm_storel_epi64((__m128i*) c4, voutb4x01234567); c4 = (int8_t*) ((uintptr_t) c4 + cn_stride); _mm_storel_epi64((__m128i*) c3, voutb3x01234567); c3 = (int8_t*) ((uintptr_t) c3 + cn_stride); _mm_storel_epi64((__m128i*) c2, voutb2x01234567); c2 = (int8_t*) ((uintptr_t) c2 + cn_stride); _mm_storel_epi64((__m128i*) c1, voutb1x01234567); c1 = (int8_t*) ((uintptr_t) c1 + cn_stride); _mm_storel_epi64((__m128i*) c0, voutb0x01234567); c0 = (int8_t*) ((uintptr_t) c0 + cn_stride); a = (const int8_t**restrict) ((uintptr_t) a - ks); nc -= 8; } else { // Prepare mask for valid 8-bit elements (depends on nc). const __mmask16 vmask = _cvtu32_mask16((UINT32_C(1) << nc) - UINT32_C(1)); _mm_mask_storeu_epi8(c4, vmask, voutb4x01234567); _mm_mask_storeu_epi8(c3, vmask, voutb3x01234567); _mm_mask_storeu_epi8(c2, vmask, voutb2x01234567); _mm_mask_storeu_epi8(c1, vmask, voutb1x01234567); _mm_mask_storeu_epi8(c0, vmask, voutb0x01234567); nc = 0; } } while (nc != 0); }