/external/XNNPACK/src/qs8-igemm/gen/ |
D | 2x16c8-minmax-neon-mull-padal.c | 82 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal() local 182 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal() 206 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal() 268 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal()
|
D | 2x16c8-minmax-neon-mlal-padal.c | 82 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() local 232 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() 321 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() 345 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() 407 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal()
|
D | 2x16c16-minmax-neon-mlal-padal.c | 82 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal() local 214 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal() 238 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal() 300 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal()
|
D | 3x16c8-minmax-neon-mull-padal.c | 86 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal() local 238 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal() 263 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal() 337 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal()
|
D | 3x16c8-minmax-neon-mlal-padal.c | 86 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal() local 305 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal() 427 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal() 452 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal() 526 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
|
D | 3x16c16-minmax-neon-mlal-padal.c | 86 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_3x16c16__neon_mlal_padal() local 286 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_3x16c16__neon_mlal_padal() 311 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_igemm_minmax_ukernel_3x16c16__neon_mlal_padal() 385 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_igemm_minmax_ukernel_3x16c16__neon_mlal_padal()
|
D | 4x16c8-minmax-neon-mull-padal.c | 90 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mull_padal() local 294 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mull_padal() 320 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mull_padal() 406 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mull_padal()
|
D | 4x16c8-minmax-neon-mlal-padal.c | 90 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() local 378 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() 533 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() 559 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() 645 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal()
|
D | 4x16c16-minmax-neon-mlal-padal.c | 90 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c16__neon_mlal_padal() local 358 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_igemm_minmax_ukernel_4x16c16__neon_mlal_padal() 384 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_igemm_minmax_ukernel_4x16c16__neon_mlal_padal() 470 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_igemm_minmax_ukernel_4x16c16__neon_mlal_padal()
|
/external/XNNPACK/src/qs8-gemm/gen/ |
D | 2x16c8-minmax-neon-mull-padal.c | 81 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal() local 169 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal() 190 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal() 252 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal()
|
D | 2x16c8-minmax-neon-mlal-padal.c | 81 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() local 219 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() 308 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() 329 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() 391 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal()
|
D | 2x16c16-minmax-neon-mlal-padal.c | 81 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal() local 201 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal() 222 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal() 284 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal()
|
D | 3x16c8-minmax-neon-mull-padal.c | 87 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mull_padal() local 223 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mull_padal() 245 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mull_padal() 319 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mull_padal()
|
D | 3x16c8-minmax-neon-mlal-padal.c | 87 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal() local 290 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal() 412 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal() 434 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal() 508 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal()
|
D | 3x16c16-minmax-neon-mlal-padal.c | 87 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_3x16c16__neon_mlal_padal() local 271 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_3x16c16__neon_mlal_padal() 293 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_gemm_minmax_ukernel_3x16c16__neon_mlal_padal() 367 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_gemm_minmax_ukernel_3x16c16__neon_mlal_padal()
|
D | 4x16c8-minmax-neon-mull-padal.c | 93 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mull_padal() local 277 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mull_padal() 300 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mull_padal() 386 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mull_padal()
|
D | 4x16c8-minmax-neon-mlal-padal.c | 93 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() local 361 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() 516 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() 539 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() 625 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal()
|
D | 4x16c16-minmax-neon-mlal-padal.c | 93 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c16__neon_mlal_padal() local 341 vacc1x15 = vpadalq_s16(vacc1x15, vprod1x15); in xnn_qs8_gemm_minmax_ukernel_4x16c16__neon_mlal_padal() 364 const int32x4_t vsum1xEF = vpaddq_s32(vacc1x14, vacc1x15); in xnn_qs8_gemm_minmax_ukernel_4x16c16__neon_mlal_padal() 450 const int32x2_t vpsum1xF = vadd_s32(vget_low_s32(vacc1x15), vget_high_s32(vacc1x15)); in xnn_qs8_gemm_minmax_ukernel_4x16c16__neon_mlal_padal()
|