// 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" void xnn_qd8_f32_qc4w_gemm_minmax_ukernel_2x8__wasm( size_t mr, size_t nc, size_t kc, const int8_t* restrict a, size_t a_stride, const void* restrict w, float* restrict c, size_t cm_stride, size_t cn_stride, const struct xnn_f32_qc4w_minmax_params params[restrict XNN_MIN_ELEMENTS(1)], const struct xnn_qd8_quantization_params quantization_params[restrict XNN_MIN_ELEMENTS(1)]) { assert(mr != 0); assert(mr <= 2); assert(nc != 0); assert(kc != 0); const int8_t* a0 = a; float* c0 = c; const int8_t* a1 = (const int8_t*) ((uintptr_t) a0 + a_stride); float* c1 = (float*) ((uintptr_t) c0 + cm_stride); if XNN_UNPREDICTABLE(mr != 2) { a1 = a0; c1 = c0; } kc = round_up_po2(kc, 2); do { const int32_t vksum0 = ((const int32_t*) w)[0]; const int32_t vksum1 = ((const int32_t*) w)[1]; const int32_t vksum2 = ((const int32_t*) w)[2]; const int32_t vksum3 = ((const int32_t*) w)[3]; const int32_t vksum4 = ((const int32_t*) w)[4]; const int32_t vksum5 = ((const int32_t*) w)[5]; const int32_t vksum6 = ((const int32_t*) w)[6]; const int32_t vksum7 = ((const int32_t*) w)[7]; const int32_t vinput_zero_point0 = quantization_params[0].zero_point; int32_t vacc0x0 = vksum0 * vinput_zero_point0; int32_t vacc0x1 = vksum1 * vinput_zero_point0; int32_t vacc0x2 = vksum2 * vinput_zero_point0; int32_t vacc0x3 = vksum3 * vinput_zero_point0; int32_t vacc0x4 = vksum4 * vinput_zero_point0; int32_t vacc0x5 = vksum5 * vinput_zero_point0; int32_t vacc0x6 = vksum6 * vinput_zero_point0; int32_t vacc0x7 = vksum7 * vinput_zero_point0; const int32_t vinput_zero_point1 = quantization_params[1].zero_point; int32_t vacc1x0 = vksum0 * vinput_zero_point1; int32_t vacc1x1 = vksum1 * vinput_zero_point1; int32_t vacc1x2 = vksum2 * vinput_zero_point1; int32_t vacc1x3 = vksum3 * vinput_zero_point1; int32_t vacc1x4 = vksum4 * vinput_zero_point1; int32_t vacc1x5 = vksum5 * vinput_zero_point1; int32_t vacc1x6 = vksum6 * vinput_zero_point1; int32_t vacc1x7 = vksum7 * vinput_zero_point1; w = (const int32_t*) w + 8; size_t k = kc; for (; k >= 2 * sizeof(uint8_t); k -= 2 * sizeof(uint8_t)) { const int32_t va0c0 = (int32_t) a0[0]; const int32_t va0c1 = (int32_t) a0[1]; a0 += 2; const int32_t va1c0 = (int32_t) a1[0]; const int32_t va1c1 = (int32_t) a1[1]; a1 += 2; const uint8_t vbi0 = ((const uint8_t*) w)[0]; const uint8_t vbi1 = ((const uint8_t*) w)[1]; const uint8_t vbi2 = ((const uint8_t*) w)[2]; const uint8_t vbi3 = ((const uint8_t*) w)[3]; const uint8_t vbi4 = ((const uint8_t*) w)[4]; const uint8_t vbi5 = ((const uint8_t*) w)[5]; const uint8_t vbi6 = ((const uint8_t*) w)[6]; const uint8_t vbi7 = ((const uint8_t*) w)[7]; w = (const uint8_t*) w + 8; const int32_t vb0c0 = (int32_t) (int8_t) (vbi0 << 4); const int32_t vb0c1 = (int32_t) (int8_t) (vbi0 & 0xF0); const int32_t vb1c0 = (int32_t) (int8_t) (vbi1 << 4); const int32_t vb1c1 = (int32_t) (int8_t) (vbi1 & 0xF0); const int32_t vb2c0 = (int32_t) (int8_t) (vbi2 << 4); const int32_t vb2c1 = (int32_t) (int8_t) (vbi2 & 0xF0); const int32_t vb3c0 = (int32_t) (int8_t) (vbi3 << 4); const int32_t vb3c1 = (int32_t) (int8_t) (vbi3 & 0xF0); const int32_t vb4c0 = (int32_t) (int8_t) (vbi4 << 4); const int32_t vb4c1 = (int32_t) (int8_t) (vbi4 & 0xF0); const int32_t vb5c0 = (int32_t) (int8_t) (vbi5 << 4); const int32_t vb5c1 = (int32_t) (int8_t) (vbi5 & 0xF0); const int32_t vb6c0 = (int32_t) (int8_t) (vbi6 << 4); const int32_t vb6c1 = (int32_t) (int8_t) (vbi6 & 0xF0); const int32_t vb7c0 = (int32_t) (int8_t) (vbi7 << 4); const int32_t vb7c1 = (int32_t) (int8_t) (vbi7 & 0xF0); vacc0x0 += va0c0 * vb0c0; vacc0x1 += va0c0 * vb1c0; vacc0x2 += va0c0 * vb2c0; vacc0x3 += va0c0 * vb3c0; vacc0x4 += va0c0 * vb4c0; vacc0x5 += va0c0 * vb5c0; vacc0x6 += va0c0 * vb6c0; vacc0x7 += va0c0 * vb7c0; vacc1x0 += va1c0 * vb0c0; vacc1x1 += va1c0 * vb1c0; vacc1x2 += va1c0 * vb2c0; vacc1x3 += va1c0 * vb3c0; vacc1x4 += va1c0 * vb4c0; vacc1x5 += va1c0 * vb5c0; vacc1x6 += va1c0 * vb6c0; vacc1x7 += va1c0 * vb7c0; vacc0x0 += va0c1 * vb0c1; vacc0x1 += va0c1 * vb1c1; vacc0x2 += va0c1 * vb2c1; vacc0x3 += va0c1 * vb3c1; vacc0x4 += va0c1 * vb4c1; vacc0x5 += va0c1 * vb5c1; vacc0x6 += va0c1 * vb6c1; vacc0x7 += va0c1 * vb7c1; vacc1x0 += va1c1 * vb0c1; vacc1x1 += va1c1 * vb1c1; vacc1x2 += va1c1 * vb2c1; vacc1x3 += va1c1 * vb3c1; vacc1x4 += va1c1 * vb4c1; vacc1x5 += va1c1 * vb5c1; vacc1x6 += va1c1 * vb6c1; vacc1x7 += va1c1 * vb7c1; } float vout0x0 = (float) math_asr_s32(vacc0x0, 4); float vout0x1 = (float) math_asr_s32(vacc0x1, 4); float vout0x2 = (float) math_asr_s32(vacc0x2, 4); float vout0x3 = (float) math_asr_s32(vacc0x3, 4); float vout0x4 = (float) math_asr_s32(vacc0x4, 4); float vout0x5 = (float) math_asr_s32(vacc0x5, 4); float vout0x6 = (float) math_asr_s32(vacc0x6, 4); float vout0x7 = (float) math_asr_s32(vacc0x7, 4); float vout1x0 = (float) math_asr_s32(vacc1x0, 4); float vout1x1 = (float) math_asr_s32(vacc1x1, 4); float vout1x2 = (float) math_asr_s32(vacc1x2, 4); float vout1x3 = (float) math_asr_s32(vacc1x3, 4); float vout1x4 = (float) math_asr_s32(vacc1x4, 4); float vout1x5 = (float) math_asr_s32(vacc1x5, 4); float vout1x6 = (float) math_asr_s32(vacc1x6, 4); float vout1x7 = (float) math_asr_s32(vacc1x7, 4); const float vinput_scale0 = quantization_params[0].inv_scale; vout0x0 *= vinput_scale0; vout0x1 *= vinput_scale0; vout0x2 *= vinput_scale0; vout0x3 *= vinput_scale0; vout0x4 *= vinput_scale0; vout0x5 *= vinput_scale0; vout0x6 *= vinput_scale0; vout0x7 *= vinput_scale0; const float vinput_scale1 = quantization_params[1].inv_scale; vout1x0 *= vinput_scale1; vout1x1 *= vinput_scale1; vout1x2 *= vinput_scale1; vout1x3 *= vinput_scale1; vout1x4 *= vinput_scale1; vout1x5 *= vinput_scale1; vout1x6 *= vinput_scale1; vout1x7 *= vinput_scale1; const float vfilter_output_scale0 = ((const float*) w)[0]; vout0x0 *= vfilter_output_scale0; vout1x0 *= vfilter_output_scale0; const float vfilter_output_scale1 = ((const float*) w)[1]; vout0x1 *= vfilter_output_scale1; vout1x1 *= vfilter_output_scale1; const float vfilter_output_scale2 = ((const float*) w)[2]; vout0x2 *= vfilter_output_scale2; vout1x2 *= vfilter_output_scale2; const float vfilter_output_scale3 = ((const float*) w)[3]; vout0x3 *= vfilter_output_scale3; vout1x3 *= vfilter_output_scale3; const float vfilter_output_scale4 = ((const float*) w)[4]; vout0x4 *= vfilter_output_scale4; vout1x4 *= vfilter_output_scale4; const float vfilter_output_scale5 = ((const float*) w)[5]; vout0x5 *= vfilter_output_scale5; vout1x5 *= vfilter_output_scale5; const float vfilter_output_scale6 = ((const float*) w)[6]; vout0x6 *= vfilter_output_scale6; vout1x6 *= vfilter_output_scale6; const float vfilter_output_scale7 = ((const float*) w)[7]; vout0x7 *= vfilter_output_scale7; vout1x7 *= vfilter_output_scale7; const float vbias0 = ((const float*) w)[8]; vout0x0 += vbias0; vout1x0 += vbias0; const float vbias1 = ((const float*) w)[9]; vout0x1 += vbias1; vout1x1 += vbias1; const float vbias2 = ((const float*) w)[10]; vout0x2 += vbias2; vout1x2 += vbias2; const float vbias3 = ((const float*) w)[11]; vout0x3 += vbias3; vout1x3 += vbias3; const float vbias4 = ((const float*) w)[12]; vout0x4 += vbias4; vout1x4 += vbias4; const float vbias5 = ((const float*) w)[13]; vout0x5 += vbias5; vout1x5 += vbias5; const float vbias6 = ((const float*) w)[14]; vout0x6 += vbias6; vout1x6 += vbias6; const float vbias7 = ((const float*) w)[15]; vout0x7 += vbias7; vout1x7 += vbias7; w = (const float*) w + 16; const float voutput_min = params->scalar.min; vout0x0 = __builtin_wasm_max_f32(vout0x0, voutput_min); vout1x0 = __builtin_wasm_max_f32(vout1x0, voutput_min); vout0x1 = __builtin_wasm_max_f32(vout0x1, voutput_min); vout1x1 = __builtin_wasm_max_f32(vout1x1, voutput_min); vout0x2 = __builtin_wasm_max_f32(vout0x2, voutput_min); vout1x2 = __builtin_wasm_max_f32(vout1x2, voutput_min); vout0x3 = __builtin_wasm_max_f32(vout0x3, voutput_min); vout1x3 = __builtin_wasm_max_f32(vout1x3, voutput_min); vout0x4 = __builtin_wasm_max_f32(vout0x4, voutput_min); vout1x4 = __builtin_wasm_max_f32(vout1x4, voutput_min); vout0x5 = __builtin_wasm_max_f32(vout0x5, voutput_min); vout1x5 = __builtin_wasm_max_f32(vout1x5, voutput_min); vout0x6 = __builtin_wasm_max_f32(vout0x6, voutput_min); vout1x6 = __builtin_wasm_max_f32(vout1x6, voutput_min); vout0x7 = __builtin_wasm_max_f32(vout0x7, voutput_min); vout1x7 = __builtin_wasm_max_f32(vout1x7, voutput_min); const float voutput_max = params->scalar.max; vout0x0 = __builtin_wasm_min_f32(vout0x0, voutput_max); vout1x0 = __builtin_wasm_min_f32(vout1x0, voutput_max); vout0x1 = __builtin_wasm_min_f32(vout0x1, voutput_max); vout1x1 = __builtin_wasm_min_f32(vout1x1, voutput_max); vout0x2 = __builtin_wasm_min_f32(vout0x2, voutput_max); vout1x2 = __builtin_wasm_min_f32(vout1x2, voutput_max); vout0x3 = __builtin_wasm_min_f32(vout0x3, voutput_max); vout1x3 = __builtin_wasm_min_f32(vout1x3, voutput_max); vout0x4 = __builtin_wasm_min_f32(vout0x4, voutput_max); vout1x4 = __builtin_wasm_min_f32(vout1x4, voutput_max); vout0x5 = __builtin_wasm_min_f32(vout0x5, voutput_max); vout1x5 = __builtin_wasm_min_f32(vout1x5, voutput_max); vout0x6 = __builtin_wasm_min_f32(vout0x6, voutput_max); vout1x6 = __builtin_wasm_min_f32(vout1x6, voutput_max); vout0x7 = __builtin_wasm_min_f32(vout0x7, voutput_max); vout1x7 = __builtin_wasm_min_f32(vout1x7, voutput_max); if XNN_LIKELY(nc >= 8) { c0[0] = vout0x0; c0[1] = vout0x1; c0[2] = vout0x2; c0[3] = vout0x3; c0[4] = vout0x4; c0[5] = vout0x5; c0[6] = vout0x6; c0[7] = vout0x7; c1[0] = vout1x0; c1[1] = vout1x1; c1[2] = vout1x2; c1[3] = vout1x3; c1[4] = vout1x4; c1[5] = vout1x5; c1[6] = vout1x6; c1[7] = vout1x7; a0 = (const int8_t*) ((uintptr_t) a0 - kc); a1 = (const int8_t*) ((uintptr_t) a1 - kc); c0 = (float*) ((uintptr_t) c0 + cn_stride); c1 = (float*) ((uintptr_t) c1 + cn_stride); nc -= 8; } else { if (nc & 4) { c0[0] = vout0x0; c0[1] = vout0x1; c0[2] = vout0x2; c0[3] = vout0x3; vout0x0 = vout0x4; vout0x1 = vout0x5; vout0x2 = vout0x6; c0 += 4; c1[0] = vout1x0; c1[1] = vout1x1; c1[2] = vout1x2; c1[3] = vout1x3; vout1x0 = vout1x4; vout1x1 = vout1x5; vout1x2 = vout1x6; c1 += 4; } if (nc & 2) { c0[0] = vout0x0; c0[1] = vout0x1; vout0x0 = vout0x2; vout0x1 = vout0x3; vout0x2 = vout0x4; vout0x3 = vout0x5; vout0x4 = vout0x6; c0 += 2; c1[0] = vout1x0; c1[1] = vout1x1; vout1x0 = vout1x2; vout1x1 = vout1x3; vout1x2 = vout1x4; vout1x3 = vout1x5; vout1x4 = vout1x6; c1 += 2; } if (nc & 1) { c0[0] = vout0x0; c1[0] = vout1x0; } nc = 0; } } while (nc != 0); }