/external/XNNPACK/src/qs8-gemm/gen/ |
D | 1x8c8-minmax-neon-mlal-padal.c | 51 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal() local 99 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal() 131 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal() 140 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal() 154 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mlal_padal()
|
D | 1x8c8-minmax-neon-mull-padal.c | 51 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mull_padal() local 82 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mull_padal() 91 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mull_padal() 105 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_1x8c8__neon_mull_padal()
|
D | 1x8c16-minmax-neon-mlal-padal.c | 51 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_1x8c16__neon_mlal_padal() local 90 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_1x8c16__neon_mlal_padal() 99 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_1x8c16__neon_mlal_padal() 113 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_1x8c16__neon_mlal_padal()
|
D | 2x8c8-minmax-neon-mull-padal.c | 57 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal() local 65 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal() 112 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal() 122 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal() 142 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
|
D | 2x8c8-minmax-neon-mlal-padal.c | 57 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() local 65 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() 138 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() 187 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() 197 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() 217 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal()
|
D | 2x8c16-minmax-neon-mlal-padal.c | 57 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_2x8c16__neon_mlal_padal() local 65 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_2x8c16__neon_mlal_padal() 128 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_2x8c16__neon_mlal_padal() 138 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_2x8c16__neon_mlal_padal() 158 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_2x8c16__neon_mlal_padal()
|
D | 3x8c8-minmax-neon-mull-padal.c | 63 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mull_padal() local 71 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mull_padal() 79 int32x4_t vacc2x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mull_padal() 142 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mull_padal() 153 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mull_padal() 179 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mull_padal()
|
D | 3x8c8-minmax-neon-mlal-padal.c | 63 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() local 71 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 79 int32x4_t vacc2x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 177 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 243 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 254 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 280 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal()
|
D | 1x16c8-minmax-neon-mull-padal.c | 51 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mull_padal() local 90 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mull_padal() 123 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mull_padal() 143 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mull_padal()
|
D | 4x8c8-minmax-neon-mull-padal.c | 69 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mull_padal() local 77 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mull_padal() 85 int32x4_t vacc2x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mull_padal() 93 int32x4_t vacc3x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mull_padal() 172 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mull_padal() 184 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mull_padal() 216 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mull_padal()
|
D | 3x8c16-minmax-neon-mlal-padal.c | 63 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_3x8c16__neon_mlal_padal() local 71 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_3x8c16__neon_mlal_padal() 79 int32x4_t vacc2x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_3x8c16__neon_mlal_padal() 166 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_3x8c16__neon_mlal_padal() 177 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_3x8c16__neon_mlal_padal() 203 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_3x8c16__neon_mlal_padal()
|
D | 1x16c8-minmax-neon-mlal-padal.c | 51 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal() local 115 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal() 179 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal() 212 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal() 232 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal()
|
D | 4x8c8-minmax-neon-mlal-padal.c | 69 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() local 77 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 85 int32x4_t vacc2x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 93 int32x4_t vacc3x7 = vacc0x7; in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 216 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 299 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 311 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 343 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal()
|
/external/XNNPACK/src/qs8-igemm/gen/ |
D | 1x8c8-minmax-neon-mull-padal.c | 54 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mull_padal() local 93 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mull_padal() 105 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mull_padal() 119 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mull_padal()
|
D | 1x8c8-minmax-neon-mlal-padal.c | 54 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mlal_padal() local 110 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mlal_padal() 142 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mlal_padal() 154 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mlal_padal() 168 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_1x8c8__neon_mlal_padal()
|
D | 1x8c16-minmax-neon-mlal-padal.c | 54 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_1x8c16__neon_mlal_padal() local 101 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_1x8c16__neon_mlal_padal() 113 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_1x8c16__neon_mlal_padal() 127 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_1x8c16__neon_mlal_padal()
|
D | 2x8c8-minmax-neon-mull-padal.c | 58 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mull_padal() local 66 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mull_padal() 125 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mull_padal() 138 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mull_padal() 158 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mull_padal()
|
D | 2x8c8-minmax-neon-mlal-padal.c | 58 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() local 66 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() 151 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() 200 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() 213 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() 233 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal()
|
D | 2x8c16-minmax-neon-mlal-padal.c | 58 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_2x8c16__neon_mlal_padal() local 66 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_2x8c16__neon_mlal_padal() 141 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_2x8c16__neon_mlal_padal() 154 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_2x8c16__neon_mlal_padal() 174 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_2x8c16__neon_mlal_padal()
|
D | 3x8c8-minmax-neon-mull-padal.c | 62 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mull_padal() local 70 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mull_padal() 78 int32x4_t vacc2x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mull_padal() 157 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mull_padal() 171 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mull_padal() 197 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mull_padal()
|
D | 4x8c8-minmax-neon-mull-padal.c | 66 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mull_padal() local 74 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mull_padal() 82 int32x4_t vacc2x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mull_padal() 90 int32x4_t vacc3x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mull_padal() 189 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mull_padal() 204 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mull_padal() 236 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mull_padal()
|
D | 1x16c8-minmax-neon-mull-padal.c | 54 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mull_padal() local 101 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mull_padal() 137 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mull_padal() 157 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mull_padal()
|
D | 1x16c8-minmax-neon-mlal-padal.c | 54 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal() local 126 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal() 190 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal() 226 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal() 246 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal()
|
D | 3x8c16-minmax-neon-mlal-padal.c | 62 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_3x8c16__neon_mlal_padal() local 70 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_3x8c16__neon_mlal_padal() 78 int32x4_t vacc2x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_3x8c16__neon_mlal_padal() 181 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_3x8c16__neon_mlal_padal() 195 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_3x8c16__neon_mlal_padal() 221 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_3x8c16__neon_mlal_padal()
|
D | 3x8c8-minmax-neon-mlal-padal.c | 62 …int32x4_t vacc0x7 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + sizeo… in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() local 70 int32x4_t vacc1x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 78 int32x4_t vacc2x7 = vacc0x7; in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 192 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 258 vacc0x7 = vpadalq_s16(vacc0x7, vprod0x7); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 272 const int32x4_t vsum0x67 = vpaddq_s32(vacc0x6, vacc0x7); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 298 const int32x2_t vpsum0x7 = vadd_s32(vget_low_s32(vacc0x7), vget_high_s32(vacc0x7)); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal()
|