// Auto-generated file. Do not edit! // Template: src/qs8-gemm/scalar.c.in // Generator: tools/xngen // // 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. #include #include "xnnpack/gemm.h" #include "xnnpack/math.h" #include "xnnpack/unaligned.h" void xnn_qu8_gemm_minmax_rndnu_ukernel_4x2__scalar( size_t mr, size_t nc, size_t kc, const uint8_t* restrict a, size_t a_stride, const void* restrict w, uint8_t* restrict c, size_t cm_stride, size_t cn_stride, const union xnn_qu8_conv_minmax_params params[restrict XNN_MIN_ELEMENTS(1)]) { assert(mr != 0); assert(mr <= 4); assert(nc != 0); assert(kc != 0); const uint8_t* a0 = a; uint8_t* c0 = c; const uint8_t* a1 = (const uint8_t*) ((uintptr_t) a0 + a_stride); uint8_t* c1 = (uint8_t*) ((uintptr_t) c0 + cm_stride); if XNN_UNPREDICTABLE(mr < 2) { a1 = a0; c1 = c0; } const uint8_t* a2 = (const uint8_t*) ((uintptr_t) a1 + a_stride); uint8_t* c2 = (uint8_t*) ((uintptr_t) c1 + cm_stride); if XNN_UNPREDICTABLE(mr <= 2) { a2 = a1; c2 = c1; } const uint8_t* a3 = (const uint8_t*) ((uintptr_t) a2 + a_stride); uint8_t* c3 = (uint8_t*) ((uintptr_t) c2 + cm_stride); if XNN_UNPREDICTABLE(mr != 4) { a3 = a2; c3 = c2; } const int32_t vmultiplier = params->rndnu_scalar.multiplier; const int64_t vrounding = params->rndnu_scalar.rounding; const uint32_t vshift = params->rndnu_scalar.shift; const int32_t voutput_min_less_zero_point = (int32_t) params->rndnu_scalar.output_min - (int32_t) params->rndnu_scalar.output_zero_point; const int32_t voutput_max_less_zero_point = (int32_t) params->rndnu_scalar.output_max - (int32_t) params->rndnu_scalar.output_zero_point; const int32_t voutput_zero_point = params->rndnu_scalar.output_zero_point; const int32_t vb_zero_point = params->rndnu_scalar.kernel_zero_point; do { int32_t vacc0x0 = unaligned_indexed_load_s32(w, 0); int32_t vacc0x1 = unaligned_indexed_load_s32(w, 1); int32_t vacc1x0 = vacc0x0; int32_t vacc1x1 = vacc0x1; int32_t vacc2x0 = vacc0x0; int32_t vacc2x1 = vacc0x1; int32_t vacc3x0 = vacc0x0; int32_t vacc3x1 = vacc0x1; w = (const int32_t*) w + 2; size_t k = kc; do { const int32_t va0 = (int32_t) (uint32_t) *a0++; const int32_t va1 = (int32_t) (uint32_t) *a1++; const int32_t va2 = (int32_t) (uint32_t) *a2++; const int32_t va3 = (int32_t) (uint32_t) *a3++; const int32_t vb0 = (int32_t) (uint32_t) ((const uint8_t*) w)[0] - vb_zero_point; const int32_t vb1 = (int32_t) (uint32_t) ((const uint8_t*) w)[1] - vb_zero_point; w = (const uint8_t*) w + 2; vacc0x0 += va0 * vb0; vacc0x1 += va0 * vb1; vacc1x0 += va1 * vb0; vacc1x1 += va1 * vb1; vacc2x0 += va2 * vb0; vacc2x1 += va2 * vb1; vacc3x0 += va3 * vb0; vacc3x1 += va3 * vb1; k -= sizeof(uint8_t); } while (k != 0); const int64_t vextacc0x0 = math_mulext_s32(vacc0x0, vmultiplier) + vrounding; const int64_t vextacc0x1 = math_mulext_s32(vacc0x1, vmultiplier) + vrounding; const int64_t vextacc1x0 = math_mulext_s32(vacc1x0, vmultiplier) + vrounding; const int64_t vextacc1x1 = math_mulext_s32(vacc1x1, vmultiplier) + vrounding; const int64_t vextacc2x0 = math_mulext_s32(vacc2x0, vmultiplier) + vrounding; const int64_t vextacc2x1 = math_mulext_s32(vacc2x1, vmultiplier) + vrounding; const int64_t vextacc3x0 = math_mulext_s32(vacc3x0, vmultiplier) + vrounding; const int64_t vextacc3x1 = math_mulext_s32(vacc3x1, vmultiplier) + vrounding; int32_t vout0x0 = (int32_t) math_asr_s64(vextacc0x0, vshift); int32_t vout0x1 = (int32_t) math_asr_s64(vextacc0x1, vshift); int32_t vout1x0 = (int32_t) math_asr_s64(vextacc1x0, vshift); int32_t vout1x1 = (int32_t) math_asr_s64(vextacc1x1, vshift); int32_t vout2x0 = (int32_t) math_asr_s64(vextacc2x0, vshift); int32_t vout2x1 = (int32_t) math_asr_s64(vextacc2x1, vshift); int32_t vout3x0 = (int32_t) math_asr_s64(vextacc3x0, vshift); int32_t vout3x1 = (int32_t) math_asr_s64(vextacc3x1, vshift); vout0x0 = math_max_s32(vout0x0, voutput_min_less_zero_point); vout0x1 = math_max_s32(vout0x1, voutput_min_less_zero_point); vout1x0 = math_max_s32(vout1x0, voutput_min_less_zero_point); vout1x1 = math_max_s32(vout1x1, voutput_min_less_zero_point); vout2x0 = math_max_s32(vout2x0, voutput_min_less_zero_point); vout2x1 = math_max_s32(vout2x1, voutput_min_less_zero_point); vout3x0 = math_max_s32(vout3x0, voutput_min_less_zero_point); vout3x1 = math_max_s32(vout3x1, voutput_min_less_zero_point); vout0x0 = math_min_s32(vout0x0, voutput_max_less_zero_point); vout0x1 = math_min_s32(vout0x1, voutput_max_less_zero_point); vout1x0 = math_min_s32(vout1x0, voutput_max_less_zero_point); vout1x1 = math_min_s32(vout1x1, voutput_max_less_zero_point); vout2x0 = math_min_s32(vout2x0, voutput_max_less_zero_point); vout2x1 = math_min_s32(vout2x1, voutput_max_less_zero_point); vout3x0 = math_min_s32(vout3x0, voutput_max_less_zero_point); vout3x1 = math_min_s32(vout3x1, voutput_max_less_zero_point); vout0x0 += voutput_zero_point; vout0x1 += voutput_zero_point; vout1x0 += voutput_zero_point; vout1x1 += voutput_zero_point; vout2x0 += voutput_zero_point; vout2x1 += voutput_zero_point; vout3x0 += voutput_zero_point; vout3x1 += voutput_zero_point; if XNN_LIKELY(nc >= 2) { c0[0] = (uint8_t) vout0x0; c0[1] = (uint8_t) vout0x1; c1[0] = (uint8_t) vout1x0; c1[1] = (uint8_t) vout1x1; c2[0] = (uint8_t) vout2x0; c2[1] = (uint8_t) vout2x1; c3[0] = (uint8_t) vout3x0; c3[1] = (uint8_t) vout3x1; a0 = (const uint8_t*) ((uintptr_t) a0 - kc); a1 = (const uint8_t*) ((uintptr_t) a1 - kc); a2 = (const uint8_t*) ((uintptr_t) a2 - kc); a3 = (const uint8_t*) ((uintptr_t) a3 - kc); c0 = (uint8_t*) ((uintptr_t) c0 + cn_stride); c1 = (uint8_t*) ((uintptr_t) c1 + cn_stride); c2 = (uint8_t*) ((uintptr_t) c2 + cn_stride); c3 = (uint8_t*) ((uintptr_t) c3 + cn_stride); nc -= 2; } else { if (nc & 1) { c0[0] = (uint8_t) vout0x0; c1[0] = (uint8_t) vout1x0; c2[0] = (uint8_t) vout2x0; c3[0] = (uint8_t) vout3x0; } nc = 0; } } while (nc != 0); }