/external/XNNPACK/src/qs8-gemm/gen/ |
D | 2x8c8-minmax-neon-mull-padal.c | 62 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal() local 98 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal() 125 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal() 153 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mull_padal()
|
D | 2x8c8-minmax-neon-mlal-padal.c | 62 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() local 118 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() 173 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() 200 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal() 228 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_2x8c8__neon_mlal_padal()
|
D | 2x8c16-minmax-neon-mlal-padal.c | 62 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_2x8c16__neon_mlal_padal() local 111 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_2x8c16__neon_mlal_padal() 141 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_2x8c16__neon_mlal_padal() 169 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_2x8c16__neon_mlal_padal()
|
D | 3x8c8-minmax-neon-mlal-padal.c | 68 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() local 148 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 223 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 257 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal() 291 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mlal_padal()
|
D | 3x8c8-minmax-neon-mull-padal.c | 68 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mull_padal() local 122 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mull_padal() 156 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mull_padal() 190 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_3x8c8__neon_mull_padal()
|
D | 3x8c16-minmax-neon-mlal-padal.c | 68 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_3x8c16__neon_mlal_padal() local 140 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_3x8c16__neon_mlal_padal() 180 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_3x8c16__neon_mlal_padal() 214 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_3x8c16__neon_mlal_padal()
|
D | 4x8c8-minmax-neon-mull-padal.c | 74 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mull_padal() local 146 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mull_padal() 187 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mull_padal() 227 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mull_padal()
|
D | 2x16c8-minmax-neon-mull-padal.c | 70 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal() local 114 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal() 185 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal() 235 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal()
|
D | 4x8c8-minmax-neon-mlal-padal.c | 74 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() local 178 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 273 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 314 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal() 354 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_4x8c8__neon_mlal_padal()
|
D | 2x16c8-minmax-neon-mlal-padal.c | 70 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() local 142 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() 253 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() 324 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() 374 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal()
|
D | 2x16c16-minmax-neon-mlal-padal.c | 70 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal() local 135 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal() 217 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal() 267 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal()
|
D | 4x8c16-minmax-neon-mlal-padal.c | 74 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_gemm_minmax_ukernel_4x8c16__neon_mlal_padal() local 169 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_gemm_minmax_ukernel_4x8c16__neon_mlal_padal() 219 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_gemm_minmax_ukernel_4x8c16__neon_mlal_padal() 259 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_gemm_minmax_ukernel_4x8c16__neon_mlal_padal()
|
/external/XNNPACK/src/qs8-igemm/gen/ |
D | 2x8c8-minmax-neon-mlal-padal.c | 63 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() local 131 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() 186 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() 216 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal() 244 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mlal_padal()
|
D | 2x8c8-minmax-neon-mull-padal.c | 63 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mull_padal() local 111 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mull_padal() 141 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mull_padal() 169 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_2x8c8__neon_mull_padal()
|
D | 2x8c16-minmax-neon-mlal-padal.c | 63 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_2x8c16__neon_mlal_padal() local 124 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_2x8c16__neon_mlal_padal() 157 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_2x8c16__neon_mlal_padal() 185 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_2x8c16__neon_mlal_padal()
|
D | 3x8c8-minmax-neon-mull-padal.c | 67 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mull_padal() local 137 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mull_padal() 174 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mull_padal() 208 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mull_padal()
|
D | 3x8c8-minmax-neon-mlal-padal.c | 67 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() local 163 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 238 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 275 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal() 309 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_3x8c8__neon_mlal_padal()
|
D | 3x8c16-minmax-neon-mlal-padal.c | 67 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_3x8c16__neon_mlal_padal() local 155 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_3x8c16__neon_mlal_padal() 198 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_3x8c16__neon_mlal_padal() 232 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_3x8c16__neon_mlal_padal()
|
D | 4x8c8-minmax-neon-mull-padal.c | 71 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mull_padal() local 163 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mull_padal() 207 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mull_padal() 247 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mull_padal()
|
D | 2x16c8-minmax-neon-mull-padal.c | 71 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal() local 127 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal() 201 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal() 251 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal()
|
D | 2x16c8-minmax-neon-mlal-padal.c | 71 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() local 155 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() 266 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() 340 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() 390 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal()
|
D | 4x8c8-minmax-neon-mlal-padal.c | 71 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() local 195 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() 290 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() 334 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal() 374 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_4x8c8__neon_mlal_padal()
|
D | 4x8c16-minmax-neon-mlal-padal.c | 71 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_4x8c16__neon_mlal_padal() local 186 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_4x8c16__neon_mlal_padal() 239 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_4x8c16__neon_mlal_padal() 279 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_4x8c16__neon_mlal_padal()
|
D | 2x16c16-minmax-neon-mlal-padal.c | 71 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal() local 148 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal() 233 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal() 283 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal()
|
D | 3x16c8-minmax-neon-mull-padal.c | 75 int32x4_t vacc1x4 = vacc0x4; in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal() local 161 vacc1x4 = vpadalq_s16(vacc1x4, vprod1x4); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal() 258 const int32x4_t vsum1x45 = vpaddq_s32(vacc1x4, vacc1x5); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal() 320 const int32x2_t vpsum1x4 = vadd_s32(vget_low_s32(vacc1x4), vget_high_s32(vacc1x4)); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal()
|