// Auto-generated file. Do not edit! // Template: src/qs8-igemm/MRx16c8-avx512vnni.c.in // Generator: tools/xngen // // Copyright 2023 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" #include "xnnpack/prefetch.h" void xnn_qs8_qc8w_igemm_minmax_fp32_ukernel_12x16c8__avx512vnni_prfm( 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 <= 12); 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; } int8_t* c5 = (int8_t*) ((uintptr_t) c4 + cm_stride); if XNN_UNPREDICTABLE(mr < 6) { c5 = c4; } int8_t* c6 = (int8_t*) ((uintptr_t) c5 + cm_stride); if XNN_UNPREDICTABLE(mr <= 6) { c6 = c5; } int8_t* c7 = (int8_t*) ((uintptr_t) c6 + cm_stride); if XNN_UNPREDICTABLE(mr < 8) { c7 = c6; } int8_t* c8 = (int8_t*) ((uintptr_t) c7 + cm_stride); if XNN_UNPREDICTABLE(mr <= 8) { c8 = c7; } int8_t* c9 = (int8_t*) ((uintptr_t) c8 + cm_stride); if XNN_UNPREDICTABLE(mr < 10) { c9 = c8; } int8_t* c10 = (int8_t*) ((uintptr_t) c9 + cm_stride); if XNN_UNPREDICTABLE(mr <= 10) { c10 = c9; } int8_t* c11 = (int8_t*) ((uintptr_t) c10 + cm_stride); if XNN_UNPREDICTABLE(mr != 12) { c11 = c10; } const __m512i vsign_mask = _mm512_set1_epi8(0x80); XNN_FORCE_REALIZATION(vsign_mask); const __m512 voutput_max_less_zero_point = _mm512_set1_ps((int32_t) params->fp32_scalar.output_max - (int32_t) params->fp32_scalar.output_zero_point); const __m512i voutput_zero_point = _mm512_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 { __m512i vacc0x01234567 = _mm512_cvtepu32_epi64(_mm256_load_si256((const __m256i*) w)); __m512i vacc0x89ABCDEF = _mm512_cvtepu32_epi64(_mm256_load_si256((const __m256i*) ((const int32_t*) w + 8))); __m512i vacc1x01234567 = vacc0x01234567; __m512i vacc1x89ABCDEF = vacc0x89ABCDEF; __m512i vacc2x01234567 = vacc0x01234567; __m512i vacc2x89ABCDEF = vacc0x89ABCDEF; __m512i vacc3x01234567 = vacc0x01234567; __m512i vacc3x89ABCDEF = vacc0x89ABCDEF; __m512i vacc4x01234567 = vacc0x01234567; __m512i vacc4x89ABCDEF = vacc0x89ABCDEF; __m512i vacc5x01234567 = vacc0x01234567; __m512i vacc5x89ABCDEF = vacc0x89ABCDEF; __m512i vacc6x01234567 = vacc0x01234567; __m512i vacc6x89ABCDEF = vacc0x89ABCDEF; __m512i vacc7x01234567 = vacc0x01234567; __m512i vacc7x89ABCDEF = vacc0x89ABCDEF; __m512i vacc8x01234567 = vacc0x01234567; __m512i vacc8x89ABCDEF = vacc0x89ABCDEF; __m512i vacc9x01234567 = vacc0x01234567; __m512i vacc9x89ABCDEF = vacc0x89ABCDEF; __m512i vacc10x01234567 = vacc0x01234567; __m512i vacc10x89ABCDEF = vacc0x89ABCDEF; __m512i vacc11x01234567 = vacc0x01234567; __m512i vacc11x89ABCDEF = vacc0x89ABCDEF; w = (const int32_t*) w + 16; 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); } const int8_t* restrict a5 = a[5]; if XNN_UNPREDICTABLE(a5 != zero) { a5 = (const int8_t*) ((uintptr_t) a5 + a_offset); } const int8_t* restrict a6 = a[6]; if XNN_UNPREDICTABLE(a6 != zero) { a6 = (const int8_t*) ((uintptr_t) a6 + a_offset); } const int8_t* restrict a7 = a[7]; if XNN_UNPREDICTABLE(a7 != zero) { a7 = (const int8_t*) ((uintptr_t) a7 + a_offset); } const int8_t* restrict a8 = a[8]; if XNN_UNPREDICTABLE(a8 != zero) { a8 = (const int8_t*) ((uintptr_t) a8 + a_offset); } const int8_t* restrict a9 = a[9]; if XNN_UNPREDICTABLE(a9 != zero) { a9 = (const int8_t*) ((uintptr_t) a9 + a_offset); } const int8_t* restrict a10 = a[10]; if XNN_UNPREDICTABLE(a10 != zero) { a10 = (const int8_t*) ((uintptr_t) a10 + a_offset); } const int8_t* restrict a11 = a[11]; if XNN_UNPREDICTABLE(a11 != zero) { a11 = (const int8_t*) ((uintptr_t) a11 + a_offset); } a += 12; size_t k = kc; while (k >= 16 * sizeof(int8_t)) { const __m512i va0x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a0)), vsign_mask); const __m512i va0x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a0 + 8)), vsign_mask); a0 += 16; const __m512i va1x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a1)), vsign_mask); const __m512i va1x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a1 + 8)), vsign_mask); a1 += 16; const __m512i va2x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a2)), vsign_mask); const __m512i va2x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a2 + 8)), vsign_mask); a2 += 16; const __m512i va3x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a3)), vsign_mask); const __m512i va3x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a3 + 8)), vsign_mask); a3 += 16; const __m512i va4x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a4)), vsign_mask); const __m512i va4x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a4 + 8)), vsign_mask); a4 += 16; const __m512i va5x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a5)), vsign_mask); const __m512i va5x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a5 + 8)), vsign_mask); a5 += 16; const __m512i va6x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a6)), vsign_mask); const __m512i va6x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a6 + 8)), vsign_mask); a6 += 16; const __m512i va7x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a7)), vsign_mask); const __m512i va7x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a7 + 8)), vsign_mask); a7 += 16; const __m512i va8x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a8)), vsign_mask); const __m512i va8x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a8 + 8)), vsign_mask); a8 += 16; const __m512i va9x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a9)), vsign_mask); const __m512i va9x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a9 + 8)), vsign_mask); a9 += 16; const __m512i va10x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a10)), vsign_mask); const __m512i va10x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a10 + 8)), vsign_mask); a10 += 16; const __m512i va11x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a11)), vsign_mask); const __m512i va11x89ABCDEF = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a11 + 8)), vsign_mask); a11 += 16; const __m512i vb01234567x01234567 = _mm512_load_si512(w); const __m512i vb89ABCDEFx01234567 = _mm512_load_si512((const int8_t*) w + 64); const __m512i vb01234567x89ABCDEF = _mm512_load_si512((const int8_t*) w + 128); const __m512i vb89ABCDEFx89ABCDEF = _mm512_load_si512((const int8_t*) w + 192); xnn_prefetch_to_l1((const int8_t*) w + 768); xnn_prefetch_to_l1((const int8_t*) w + 832); vacc0x01234567 = _mm512_dpbusd_epi32(vacc0x01234567, va0x01234567, vb01234567x01234567); vacc0x89ABCDEF = _mm512_dpbusd_epi32(vacc0x89ABCDEF, va0x01234567, vb89ABCDEFx01234567); vacc1x01234567 = _mm512_dpbusd_epi32(vacc1x01234567, va1x01234567, vb01234567x01234567); vacc1x89ABCDEF = _mm512_dpbusd_epi32(vacc1x89ABCDEF, va1x01234567, vb89ABCDEFx01234567); vacc2x01234567 = _mm512_dpbusd_epi32(vacc2x01234567, va2x01234567, vb01234567x01234567); vacc2x89ABCDEF = _mm512_dpbusd_epi32(vacc2x89ABCDEF, va2x01234567, vb89ABCDEFx01234567); vacc3x01234567 = _mm512_dpbusd_epi32(vacc3x01234567, va3x01234567, vb01234567x01234567); vacc3x89ABCDEF = _mm512_dpbusd_epi32(vacc3x89ABCDEF, va3x01234567, vb89ABCDEFx01234567); vacc4x01234567 = _mm512_dpbusd_epi32(vacc4x01234567, va4x01234567, vb01234567x01234567); vacc4x89ABCDEF = _mm512_dpbusd_epi32(vacc4x89ABCDEF, va4x01234567, vb89ABCDEFx01234567); vacc5x01234567 = _mm512_dpbusd_epi32(vacc5x01234567, va5x01234567, vb01234567x01234567); vacc5x89ABCDEF = _mm512_dpbusd_epi32(vacc5x89ABCDEF, va5x01234567, vb89ABCDEFx01234567); vacc6x01234567 = _mm512_dpbusd_epi32(vacc6x01234567, va6x01234567, vb01234567x01234567); vacc6x89ABCDEF = _mm512_dpbusd_epi32(vacc6x89ABCDEF, va6x01234567, vb89ABCDEFx01234567); vacc7x01234567 = _mm512_dpbusd_epi32(vacc7x01234567, va7x01234567, vb01234567x01234567); vacc7x89ABCDEF = _mm512_dpbusd_epi32(vacc7x89ABCDEF, va7x01234567, vb89ABCDEFx01234567); vacc8x01234567 = _mm512_dpbusd_epi32(vacc8x01234567, va8x01234567, vb01234567x01234567); vacc8x89ABCDEF = _mm512_dpbusd_epi32(vacc8x89ABCDEF, va8x01234567, vb89ABCDEFx01234567); vacc9x01234567 = _mm512_dpbusd_epi32(vacc9x01234567, va9x01234567, vb01234567x01234567); vacc9x89ABCDEF = _mm512_dpbusd_epi32(vacc9x89ABCDEF, va9x01234567, vb89ABCDEFx01234567); vacc10x01234567 = _mm512_dpbusd_epi32(vacc10x01234567, va10x01234567, vb01234567x01234567); vacc10x89ABCDEF = _mm512_dpbusd_epi32(vacc10x89ABCDEF, va10x01234567, vb89ABCDEFx01234567); vacc11x01234567 = _mm512_dpbusd_epi32(vacc11x01234567, va11x01234567, vb01234567x01234567); vacc11x89ABCDEF = _mm512_dpbusd_epi32(vacc11x89ABCDEF, va11x01234567, vb89ABCDEFx01234567); xnn_prefetch_to_l1((const int8_t*) w + 896); xnn_prefetch_to_l1((const int8_t*) w + 960); vacc0x01234567 = _mm512_dpbusd_epi32(vacc0x01234567, va0x89ABCDEF, vb01234567x89ABCDEF); vacc0x89ABCDEF = _mm512_dpbusd_epi32(vacc0x89ABCDEF, va0x89ABCDEF, vb89ABCDEFx89ABCDEF); vacc1x01234567 = _mm512_dpbusd_epi32(vacc1x01234567, va1x89ABCDEF, vb01234567x89ABCDEF); vacc1x89ABCDEF = _mm512_dpbusd_epi32(vacc1x89ABCDEF, va1x89ABCDEF, vb89ABCDEFx89ABCDEF); vacc2x01234567 = _mm512_dpbusd_epi32(vacc2x01234567, va2x89ABCDEF, vb01234567x89ABCDEF); vacc2x89ABCDEF = _mm512_dpbusd_epi32(vacc2x89ABCDEF, va2x89ABCDEF, vb89ABCDEFx89ABCDEF); vacc3x01234567 = _mm512_dpbusd_epi32(vacc3x01234567, va3x89ABCDEF, vb01234567x89ABCDEF); vacc3x89ABCDEF = _mm512_dpbusd_epi32(vacc3x89ABCDEF, va3x89ABCDEF, vb89ABCDEFx89ABCDEF); vacc4x01234567 = _mm512_dpbusd_epi32(vacc4x01234567, va4x89ABCDEF, vb01234567x89ABCDEF); vacc4x89ABCDEF = _mm512_dpbusd_epi32(vacc4x89ABCDEF, va4x89ABCDEF, vb89ABCDEFx89ABCDEF); vacc5x01234567 = _mm512_dpbusd_epi32(vacc5x01234567, va5x89ABCDEF, vb01234567x89ABCDEF); vacc5x89ABCDEF = _mm512_dpbusd_epi32(vacc5x89ABCDEF, va5x89ABCDEF, vb89ABCDEFx89ABCDEF); vacc6x01234567 = _mm512_dpbusd_epi32(vacc6x01234567, va6x89ABCDEF, vb01234567x89ABCDEF); vacc6x89ABCDEF = _mm512_dpbusd_epi32(vacc6x89ABCDEF, va6x89ABCDEF, vb89ABCDEFx89ABCDEF); vacc7x01234567 = _mm512_dpbusd_epi32(vacc7x01234567, va7x89ABCDEF, vb01234567x89ABCDEF); vacc7x89ABCDEF = _mm512_dpbusd_epi32(vacc7x89ABCDEF, va7x89ABCDEF, vb89ABCDEFx89ABCDEF); vacc8x01234567 = _mm512_dpbusd_epi32(vacc8x01234567, va8x89ABCDEF, vb01234567x89ABCDEF); vacc8x89ABCDEF = _mm512_dpbusd_epi32(vacc8x89ABCDEF, va8x89ABCDEF, vb89ABCDEFx89ABCDEF); vacc9x01234567 = _mm512_dpbusd_epi32(vacc9x01234567, va9x89ABCDEF, vb01234567x89ABCDEF); vacc9x89ABCDEF = _mm512_dpbusd_epi32(vacc9x89ABCDEF, va9x89ABCDEF, vb89ABCDEFx89ABCDEF); vacc10x01234567 = _mm512_dpbusd_epi32(vacc10x01234567, va10x89ABCDEF, vb01234567x89ABCDEF); vacc10x89ABCDEF = _mm512_dpbusd_epi32(vacc10x89ABCDEF, va10x89ABCDEF, vb89ABCDEFx89ABCDEF); vacc11x01234567 = _mm512_dpbusd_epi32(vacc11x01234567, va11x89ABCDEF, vb01234567x89ABCDEF); vacc11x89ABCDEF = _mm512_dpbusd_epi32(vacc11x89ABCDEF, va11x89ABCDEF, vb89ABCDEFx89ABCDEF); w = (const int8_t*) w + 256; k -= 16 * sizeof(int8_t); } if (k != 0) { const __m512i va0x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a0)), vsign_mask); a0 += 8; const __m512i va1x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a1)), vsign_mask); a1 += 8; const __m512i va2x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a2)), vsign_mask); a2 += 8; const __m512i va3x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a3)), vsign_mask); a3 += 8; const __m512i va4x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a4)), vsign_mask); a4 += 8; const __m512i va5x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a5)), vsign_mask); a5 += 8; const __m512i va6x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a6)), vsign_mask); a6 += 8; const __m512i va7x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a7)), vsign_mask); a7 += 8; const __m512i va8x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a8)), vsign_mask); a8 += 8; const __m512i va9x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a9)), vsign_mask); a9 += 8; const __m512i va10x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a10)), vsign_mask); a10 += 8; const __m512i va11x01234567 = _mm512_xor_epi64(_mm512_set1_epi64((int64_t) unaligned_load_u64(a11)), vsign_mask); a11 += 8; const __m512i vb01234567x01234567 = _mm512_load_si512(w); const __m512i vb89ABCDEFx01234567 = _mm512_load_si512((const int8_t*) w + 64); vacc0x01234567 = _mm512_dpbusd_epi32(vacc0x01234567, va0x01234567, vb01234567x01234567); vacc0x89ABCDEF = _mm512_dpbusd_epi32(vacc0x89ABCDEF, va0x01234567, vb89ABCDEFx01234567); vacc1x01234567 = _mm512_dpbusd_epi32(vacc1x01234567, va1x01234567, vb01234567x01234567); vacc1x89ABCDEF = _mm512_dpbusd_epi32(vacc1x89ABCDEF, va1x01234567, vb89ABCDEFx01234567); vacc2x01234567 = _mm512_dpbusd_epi32(vacc2x01234567, va2x01234567, vb01234567x01234567); vacc2x89ABCDEF = _mm512_dpbusd_epi32(vacc2x89ABCDEF, va2x01234567, vb89ABCDEFx01234567); vacc3x01234567 = _mm512_dpbusd_epi32(vacc3x01234567, va3x01234567, vb01234567x01234567); vacc3x89ABCDEF = _mm512_dpbusd_epi32(vacc3x89ABCDEF, va3x01234567, vb89ABCDEFx01234567); vacc4x01234567 = _mm512_dpbusd_epi32(vacc4x01234567, va4x01234567, vb01234567x01234567); vacc4x89ABCDEF = _mm512_dpbusd_epi32(vacc4x89ABCDEF, va4x01234567, vb89ABCDEFx01234567); vacc5x01234567 = _mm512_dpbusd_epi32(vacc5x01234567, va5x01234567, vb01234567x01234567); vacc5x89ABCDEF = _mm512_dpbusd_epi32(vacc5x89ABCDEF, va5x01234567, vb89ABCDEFx01234567); vacc6x01234567 = _mm512_dpbusd_epi32(vacc6x01234567, va6x01234567, vb01234567x01234567); vacc6x89ABCDEF = _mm512_dpbusd_epi32(vacc6x89ABCDEF, va6x01234567, vb89ABCDEFx01234567); vacc7x01234567 = _mm512_dpbusd_epi32(vacc7x01234567, va7x01234567, vb01234567x01234567); vacc7x89ABCDEF = _mm512_dpbusd_epi32(vacc7x89ABCDEF, va7x01234567, vb89ABCDEFx01234567); vacc8x01234567 = _mm512_dpbusd_epi32(vacc8x01234567, va8x01234567, vb01234567x01234567); vacc8x89ABCDEF = _mm512_dpbusd_epi32(vacc8x89ABCDEF, va8x01234567, vb89ABCDEFx01234567); vacc9x01234567 = _mm512_dpbusd_epi32(vacc9x01234567, va9x01234567, vb01234567x01234567); vacc9x89ABCDEF = _mm512_dpbusd_epi32(vacc9x89ABCDEF, va9x01234567, vb89ABCDEFx01234567); vacc10x01234567 = _mm512_dpbusd_epi32(vacc10x01234567, va10x01234567, vb01234567x01234567); vacc10x89ABCDEF = _mm512_dpbusd_epi32(vacc10x89ABCDEF, va10x01234567, vb89ABCDEFx01234567); vacc11x01234567 = _mm512_dpbusd_epi32(vacc11x01234567, va11x01234567, vb01234567x01234567); vacc11x89ABCDEF = _mm512_dpbusd_epi32(vacc11x89ABCDEF, va11x01234567, vb89ABCDEFx01234567); xnn_prefetch_to_l1((const int8_t*) w + 896); xnn_prefetch_to_l1((const int8_t*) w + 960); w = (const int8_t*) w + 128; k -= 8 * sizeof(int8_t); } p -= 12 * sizeof(void*); } while (p != 0); // Add adjacent pairs const __m512i vidx = _mm512_set_epi32(30, 28, 26, 24, 22, 20, 18, 16, 14, 12, 10, 8, 6, 4, 2, 0); const __m512i vsum0x01234567 = _mm512_add_epi32(vacc0x01234567, _mm512_srli_epi64(vacc0x01234567, 32)); const __m512i vsum0x89ABCDEF = _mm512_add_epi32(vacc0x89ABCDEF, _mm512_srli_epi64(vacc0x89ABCDEF, 32)); __m512i vacc0x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum0x01234567, vidx, vsum0x89ABCDEF); const __m512i vsum1x01234567 = _mm512_add_epi32(vacc1x01234567, _mm512_srli_epi64(vacc1x01234567, 32)); const __m512i vsum1x89ABCDEF = _mm512_add_epi32(vacc1x89ABCDEF, _mm512_srli_epi64(vacc1x89ABCDEF, 32)); __m512i vacc1x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum1x01234567, vidx, vsum1x89ABCDEF); const __m512i vsum2x01234567 = _mm512_add_epi32(vacc2x01234567, _mm512_srli_epi64(vacc2x01234567, 32)); const __m512i vsum2x89ABCDEF = _mm512_add_epi32(vacc2x89ABCDEF, _mm512_srli_epi64(vacc2x89ABCDEF, 32)); __m512i vacc2x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum2x01234567, vidx, vsum2x89ABCDEF); const __m512i vsum3x01234567 = _mm512_add_epi32(vacc3x01234567, _mm512_srli_epi64(vacc3x01234567, 32)); const __m512i vsum3x89ABCDEF = _mm512_add_epi32(vacc3x89ABCDEF, _mm512_srli_epi64(vacc3x89ABCDEF, 32)); __m512i vacc3x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum3x01234567, vidx, vsum3x89ABCDEF); const __m512i vsum4x01234567 = _mm512_add_epi32(vacc4x01234567, _mm512_srli_epi64(vacc4x01234567, 32)); const __m512i vsum4x89ABCDEF = _mm512_add_epi32(vacc4x89ABCDEF, _mm512_srli_epi64(vacc4x89ABCDEF, 32)); __m512i vacc4x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum4x01234567, vidx, vsum4x89ABCDEF); const __m512i vsum5x01234567 = _mm512_add_epi32(vacc5x01234567, _mm512_srli_epi64(vacc5x01234567, 32)); const __m512i vsum5x89ABCDEF = _mm512_add_epi32(vacc5x89ABCDEF, _mm512_srli_epi64(vacc5x89ABCDEF, 32)); __m512i vacc5x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum5x01234567, vidx, vsum5x89ABCDEF); const __m512i vsum6x01234567 = _mm512_add_epi32(vacc6x01234567, _mm512_srli_epi64(vacc6x01234567, 32)); const __m512i vsum6x89ABCDEF = _mm512_add_epi32(vacc6x89ABCDEF, _mm512_srli_epi64(vacc6x89ABCDEF, 32)); __m512i vacc6x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum6x01234567, vidx, vsum6x89ABCDEF); const __m512i vsum7x01234567 = _mm512_add_epi32(vacc7x01234567, _mm512_srli_epi64(vacc7x01234567, 32)); const __m512i vsum7x89ABCDEF = _mm512_add_epi32(vacc7x89ABCDEF, _mm512_srli_epi64(vacc7x89ABCDEF, 32)); __m512i vacc7x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum7x01234567, vidx, vsum7x89ABCDEF); const __m512i vsum8x01234567 = _mm512_add_epi32(vacc8x01234567, _mm512_srli_epi64(vacc8x01234567, 32)); const __m512i vsum8x89ABCDEF = _mm512_add_epi32(vacc8x89ABCDEF, _mm512_srli_epi64(vacc8x89ABCDEF, 32)); __m512i vacc8x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum8x01234567, vidx, vsum8x89ABCDEF); const __m512i vsum9x01234567 = _mm512_add_epi32(vacc9x01234567, _mm512_srli_epi64(vacc9x01234567, 32)); const __m512i vsum9x89ABCDEF = _mm512_add_epi32(vacc9x89ABCDEF, _mm512_srli_epi64(vacc9x89ABCDEF, 32)); __m512i vacc9x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum9x01234567, vidx, vsum9x89ABCDEF); const __m512i vsum10x01234567 = _mm512_add_epi32(vacc10x01234567, _mm512_srli_epi64(vacc10x01234567, 32)); const __m512i vsum10x89ABCDEF = _mm512_add_epi32(vacc10x89ABCDEF, _mm512_srli_epi64(vacc10x89ABCDEF, 32)); __m512i vacc10x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum10x01234567, vidx, vsum10x89ABCDEF); const __m512i vsum11x01234567 = _mm512_add_epi32(vacc11x01234567, _mm512_srli_epi64(vacc11x01234567, 32)); const __m512i vsum11x89ABCDEF = _mm512_add_epi32(vacc11x89ABCDEF, _mm512_srli_epi64(vacc11x89ABCDEF, 32)); __m512i vacc11x0123456789ABCDEF = _mm512_permutex2var_epi32(vsum11x01234567, vidx, vsum11x89ABCDEF); __m512 vscaled0x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc0x0123456789ABCDEF); __m512 vscaled1x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc1x0123456789ABCDEF); __m512 vscaled2x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc2x0123456789ABCDEF); __m512 vscaled3x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc3x0123456789ABCDEF); __m512 vscaled4x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc4x0123456789ABCDEF); __m512 vscaled5x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc5x0123456789ABCDEF); __m512 vscaled6x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc6x0123456789ABCDEF); __m512 vscaled7x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc7x0123456789ABCDEF); __m512 vscaled8x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc8x0123456789ABCDEF); __m512 vscaled9x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc9x0123456789ABCDEF); __m512 vscaled10x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc10x0123456789ABCDEF); __m512 vscaled11x0123456789ABCDEF = _mm512_cvtepi32_ps(vacc11x0123456789ABCDEF); const __m512 vscale012345678ABCDEF = _mm512_load_ps(w); w = (const float*) w + 16; vscaled0x0123456789ABCDEF = _mm512_mul_ps(vscaled0x0123456789ABCDEF, vscale012345678ABCDEF); vscaled1x0123456789ABCDEF = _mm512_mul_ps(vscaled1x0123456789ABCDEF, vscale012345678ABCDEF); vscaled2x0123456789ABCDEF = _mm512_mul_ps(vscaled2x0123456789ABCDEF, vscale012345678ABCDEF); vscaled3x0123456789ABCDEF = _mm512_mul_ps(vscaled3x0123456789ABCDEF, vscale012345678ABCDEF); vscaled4x0123456789ABCDEF = _mm512_mul_ps(vscaled4x0123456789ABCDEF, vscale012345678ABCDEF); vscaled5x0123456789ABCDEF = _mm512_mul_ps(vscaled5x0123456789ABCDEF, vscale012345678ABCDEF); vscaled6x0123456789ABCDEF = _mm512_mul_ps(vscaled6x0123456789ABCDEF, vscale012345678ABCDEF); vscaled7x0123456789ABCDEF = _mm512_mul_ps(vscaled7x0123456789ABCDEF, vscale012345678ABCDEF); vscaled8x0123456789ABCDEF = _mm512_mul_ps(vscaled8x0123456789ABCDEF, vscale012345678ABCDEF); vscaled9x0123456789ABCDEF = _mm512_mul_ps(vscaled9x0123456789ABCDEF, vscale012345678ABCDEF); vscaled10x0123456789ABCDEF = _mm512_mul_ps(vscaled10x0123456789ABCDEF, vscale012345678ABCDEF); vscaled11x0123456789ABCDEF = _mm512_mul_ps(vscaled11x0123456789ABCDEF, vscale012345678ABCDEF); vscaled0x0123456789ABCDEF = _mm512_min_ps(vscaled0x0123456789ABCDEF, voutput_max_less_zero_point); vscaled1x0123456789ABCDEF = _mm512_min_ps(vscaled1x0123456789ABCDEF, voutput_max_less_zero_point); vscaled2x0123456789ABCDEF = _mm512_min_ps(vscaled2x0123456789ABCDEF, voutput_max_less_zero_point); vscaled3x0123456789ABCDEF = _mm512_min_ps(vscaled3x0123456789ABCDEF, voutput_max_less_zero_point); vscaled4x0123456789ABCDEF = _mm512_min_ps(vscaled4x0123456789ABCDEF, voutput_max_less_zero_point); vscaled5x0123456789ABCDEF = _mm512_min_ps(vscaled5x0123456789ABCDEF, voutput_max_less_zero_point); vscaled6x0123456789ABCDEF = _mm512_min_ps(vscaled6x0123456789ABCDEF, voutput_max_less_zero_point); vscaled7x0123456789ABCDEF = _mm512_min_ps(vscaled7x0123456789ABCDEF, voutput_max_less_zero_point); vscaled8x0123456789ABCDEF = _mm512_min_ps(vscaled8x0123456789ABCDEF, voutput_max_less_zero_point); vscaled9x0123456789ABCDEF = _mm512_min_ps(vscaled9x0123456789ABCDEF, voutput_max_less_zero_point); vscaled10x0123456789ABCDEF = _mm512_min_ps(vscaled10x0123456789ABCDEF, voutput_max_less_zero_point); vscaled11x0123456789ABCDEF = _mm512_min_ps(vscaled11x0123456789ABCDEF, voutput_max_less_zero_point); vacc0x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled0x0123456789ABCDEF); vacc1x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled1x0123456789ABCDEF); vacc2x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled2x0123456789ABCDEF); vacc3x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled3x0123456789ABCDEF); vacc4x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled4x0123456789ABCDEF); vacc5x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled5x0123456789ABCDEF); vacc6x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled6x0123456789ABCDEF); vacc7x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled7x0123456789ABCDEF); vacc8x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled8x0123456789ABCDEF); vacc9x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled9x0123456789ABCDEF); vacc10x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled10x0123456789ABCDEF); vacc11x0123456789ABCDEF = _mm512_cvtps_epi32(vscaled11x0123456789ABCDEF); vacc0x0123456789ABCDEF = _mm512_add_epi32(vacc0x0123456789ABCDEF, voutput_zero_point); vacc1x0123456789ABCDEF = _mm512_add_epi32(vacc1x0123456789ABCDEF, voutput_zero_point); vacc2x0123456789ABCDEF = _mm512_add_epi32(vacc2x0123456789ABCDEF, voutput_zero_point); vacc3x0123456789ABCDEF = _mm512_add_epi32(vacc3x0123456789ABCDEF, voutput_zero_point); vacc4x0123456789ABCDEF = _mm512_add_epi32(vacc4x0123456789ABCDEF, voutput_zero_point); vacc5x0123456789ABCDEF = _mm512_add_epi32(vacc5x0123456789ABCDEF, voutput_zero_point); vacc6x0123456789ABCDEF = _mm512_add_epi32(vacc6x0123456789ABCDEF, voutput_zero_point); vacc7x0123456789ABCDEF = _mm512_add_epi32(vacc7x0123456789ABCDEF, voutput_zero_point); vacc8x0123456789ABCDEF = _mm512_add_epi32(vacc8x0123456789ABCDEF, voutput_zero_point); vacc9x0123456789ABCDEF = _mm512_add_epi32(vacc9x0123456789ABCDEF, voutput_zero_point); vacc10x0123456789ABCDEF = _mm512_add_epi32(vacc10x0123456789ABCDEF, voutput_zero_point); vacc11x0123456789ABCDEF = _mm512_add_epi32(vacc11x0123456789ABCDEF, voutput_zero_point); __m128i vout0x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc0x0123456789ABCDEF); __m128i vout1x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc1x0123456789ABCDEF); __m128i vout2x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc2x0123456789ABCDEF); __m128i vout3x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc3x0123456789ABCDEF); __m128i vout4x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc4x0123456789ABCDEF); __m128i vout5x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc5x0123456789ABCDEF); __m128i vout6x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc6x0123456789ABCDEF); __m128i vout7x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc7x0123456789ABCDEF); __m128i vout8x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc8x0123456789ABCDEF); __m128i vout9x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc9x0123456789ABCDEF); __m128i vout10x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc10x0123456789ABCDEF); __m128i vout11x0123456789ABCDEF = _mm512_cvtsepi32_epi8(vacc11x0123456789ABCDEF); vout0x0123456789ABCDEF = _mm_max_epi8(vout0x0123456789ABCDEF, voutput_min); vout1x0123456789ABCDEF = _mm_max_epi8(vout1x0123456789ABCDEF, voutput_min); vout2x0123456789ABCDEF = _mm_max_epi8(vout2x0123456789ABCDEF, voutput_min); vout3x0123456789ABCDEF = _mm_max_epi8(vout3x0123456789ABCDEF, voutput_min); vout4x0123456789ABCDEF = _mm_max_epi8(vout4x0123456789ABCDEF, voutput_min); vout5x0123456789ABCDEF = _mm_max_epi8(vout5x0123456789ABCDEF, voutput_min); vout6x0123456789ABCDEF = _mm_max_epi8(vout6x0123456789ABCDEF, voutput_min); vout7x0123456789ABCDEF = _mm_max_epi8(vout7x0123456789ABCDEF, voutput_min); vout8x0123456789ABCDEF = _mm_max_epi8(vout8x0123456789ABCDEF, voutput_min); vout9x0123456789ABCDEF = _mm_max_epi8(vout9x0123456789ABCDEF, voutput_min); vout10x0123456789ABCDEF = _mm_max_epi8(vout10x0123456789ABCDEF, voutput_min); vout11x0123456789ABCDEF = _mm_max_epi8(vout11x0123456789ABCDEF, voutput_min); if (nc >= 16) { _mm_storeu_si128((__m128i*) c11, vout11x0123456789ABCDEF); c11 = (int8_t*) ((uintptr_t) c11 + cn_stride); _mm_storeu_si128((__m128i*) c10, vout10x0123456789ABCDEF); c10 = (int8_t*) ((uintptr_t) c10 + cn_stride); _mm_storeu_si128((__m128i*) c9, vout9x0123456789ABCDEF); c9 = (int8_t*) ((uintptr_t) c9 + cn_stride); _mm_storeu_si128((__m128i*) c8, vout8x0123456789ABCDEF); c8 = (int8_t*) ((uintptr_t) c8 + cn_stride); _mm_storeu_si128((__m128i*) c7, vout7x0123456789ABCDEF); c7 = (int8_t*) ((uintptr_t) c7 + cn_stride); _mm_storeu_si128((__m128i*) c6, vout6x0123456789ABCDEF); c6 = (int8_t*) ((uintptr_t) c6 + cn_stride); _mm_storeu_si128((__m128i*) c5, vout5x0123456789ABCDEF); c5 = (int8_t*) ((uintptr_t) c5 + cn_stride); _mm_storeu_si128((__m128i*) c4, vout4x0123456789ABCDEF); c4 = (int8_t*) ((uintptr_t) c4 + cn_stride); _mm_storeu_si128((__m128i*) c3, vout3x0123456789ABCDEF); c3 = (int8_t*) ((uintptr_t) c3 + cn_stride); _mm_storeu_si128((__m128i*) c2, vout2x0123456789ABCDEF); c2 = (int8_t*) ((uintptr_t) c2 + cn_stride); _mm_storeu_si128((__m128i*) c1, vout1x0123456789ABCDEF); c1 = (int8_t*) ((uintptr_t) c1 + cn_stride); _mm_storeu_si128((__m128i*) c0, vout0x0123456789ABCDEF); c0 = (int8_t*) ((uintptr_t) c0 + cn_stride); a = (const int8_t**restrict) ((uintptr_t) a - ks); nc -= 16; } 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(c11, vmask, vout11x0123456789ABCDEF); _mm_mask_storeu_epi8(c10, vmask, vout10x0123456789ABCDEF); _mm_mask_storeu_epi8(c9, vmask, vout9x0123456789ABCDEF); _mm_mask_storeu_epi8(c8, vmask, vout8x0123456789ABCDEF); _mm_mask_storeu_epi8(c7, vmask, vout7x0123456789ABCDEF); _mm_mask_storeu_epi8(c6, vmask, vout6x0123456789ABCDEF); _mm_mask_storeu_epi8(c5, vmask, vout5x0123456789ABCDEF); _mm_mask_storeu_epi8(c4, vmask, vout4x0123456789ABCDEF); _mm_mask_storeu_epi8(c3, vmask, vout3x0123456789ABCDEF); _mm_mask_storeu_epi8(c2, vmask, vout2x0123456789ABCDEF); _mm_mask_storeu_epi8(c1, vmask, vout1x0123456789ABCDEF); _mm_mask_storeu_epi8(c0, vmask, vout0x0123456789ABCDEF); nc = 0; } } while (nc != 0); }