| /external/XNNPACK/src/qs8-gemm/gen/ |
| D | 1x16c8-minmax-neon-mull-padal.c | 59 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mull_padal() local
114 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mull_padal()
127 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mull_padal()
157 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mull_padal()
|
| D | 1x16c8-minmax-neon-mlal-padal.c | 59 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal() local
147 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal()
203 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal()
216 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal()
246 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_1x16c8__neon_mlal_padal()
|
| D | 1x16c16-minmax-neon-mlal-padal.c | 59 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_1x16c16__neon_mlal_padal() local
130 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_1x16c16__neon_mlal_padal()
143 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_1x16c16__neon_mlal_padal()
173 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_1x16c16__neon_mlal_padal()
|
| D | 2x16c8-minmax-neon-mull-padal.c | 65 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal() local
81 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal()
168 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal()
182 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal()
224 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mull_padal()
|
| D | 2x16c8-minmax-neon-mlal-padal.c | 65 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal() local
81 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal()
218 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal()
307 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal()
321 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal()
363 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_2x16c8__neon_mlal_padal()
|
| D | 2x16c16-minmax-neon-mlal-padal.c | 65 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal() local
81 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal()
200 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal()
214 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal()
256 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_2x16c16__neon_mlal_padal()
|
| D | 3x16c8-minmax-neon-mull-padal.c | 71 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mull_padal() local
87 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mull_padal()
103 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mull_padal()
222 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mull_padal()
237 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mull_padal()
291 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mull_padal()
|
| D | 3x16c8-minmax-neon-mlal-padal.c | 71 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal() local
87 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal()
103 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal()
289 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal()
411 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal()
426 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal()
480 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_3x16c8__neon_mlal_padal()
|
| D | 4x16c8-minmax-neon-mull-padal.c | 77 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mull_padal() local
93 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mull_padal()
109 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mull_padal()
125 int32x4_t vacc3x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mull_padal()
276 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mull_padal()
292 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mull_padal()
358 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mull_padal()
|
| D | 3x16c16-minmax-neon-mlal-padal.c | 71 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_3x16c16__neon_mlal_padal() local
87 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_3x16c16__neon_mlal_padal()
103 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_3x16c16__neon_mlal_padal()
270 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_3x16c16__neon_mlal_padal()
285 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_3x16c16__neon_mlal_padal()
339 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_3x16c16__neon_mlal_padal()
|
| D | 4x16c16-minmax-neon-mlal-padal.c | 77 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_4x16c16__neon_mlal_padal() local
93 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c16__neon_mlal_padal()
109 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c16__neon_mlal_padal()
125 int32x4_t vacc3x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c16__neon_mlal_padal()
340 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_4x16c16__neon_mlal_padal()
356 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_4x16c16__neon_mlal_padal()
422 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_4x16c16__neon_mlal_padal()
|
| D | 4x16c8-minmax-neon-mlal-padal.c | 77 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal() local
93 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal()
109 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal()
125 int32x4_t vacc3x15 = vacc0x15; in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal()
360 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal()
515 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal()
531 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal()
597 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_gemm_minmax_ukernel_4x16c8__neon_mlal_padal()
|
| /external/XNNPACK/src/qs8-igemm/gen/ |
| D | 1x16c8-minmax-neon-mull-padal.c | 62 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mull_padal() local
125 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mull_padal()
141 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mull_padal()
171 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mull_padal()
|
| D | 1x16c8-minmax-neon-mlal-padal.c | 62 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal() local
158 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal()
214 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal()
230 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal()
260 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_1x16c8__neon_mlal_padal()
|
| D | 1x16c16-minmax-neon-mlal-padal.c | 62 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_1x16c16__neon_mlal_padal() local
141 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_1x16c16__neon_mlal_padal()
157 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_1x16c16__neon_mlal_padal()
187 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_1x16c16__neon_mlal_padal()
|
| D | 2x16c8-minmax-neon-mull-padal.c | 66 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal() local
82 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal()
181 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal()
198 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal()
240 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mull_padal()
|
| D | 2x16c8-minmax-neon-mlal-padal.c | 66 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal() local
82 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal()
231 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal()
320 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal()
337 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal()
379 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_2x16c8__neon_mlal_padal()
|
| D | 3x16c8-minmax-neon-mull-padal.c | 70 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal() local
86 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal()
102 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal()
237 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal()
255 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal()
309 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mull_padal()
|
| D | 2x16c16-minmax-neon-mlal-padal.c | 66 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal() local
82 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal()
213 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal()
230 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal()
272 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_2x16c16__neon_mlal_padal()
|
| D | 4x16c8-minmax-neon-mull-padal.c | 74 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mull_padal() local
90 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mull_padal()
106 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mull_padal()
122 int32x4_t vacc3x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mull_padal()
293 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mull_padal()
312 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mull_padal()
378 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mull_padal()
|
| D | 3x16c16-minmax-neon-mlal-padal.c | 70 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_3x16c16__neon_mlal_padal() local
86 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_3x16c16__neon_mlal_padal()
102 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_3x16c16__neon_mlal_padal()
285 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_3x16c16__neon_mlal_padal()
303 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_3x16c16__neon_mlal_padal()
357 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_3x16c16__neon_mlal_padal()
|
| D | 3x16c8-minmax-neon-mlal-padal.c | 70 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal() local
86 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
102 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
304 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
426 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
444 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
498 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_3x16c8__neon_mlal_padal()
|
| D | 4x16c16-minmax-neon-mlal-padal.c | 74 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_4x16c16__neon_mlal_padal() local
90 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c16__neon_mlal_padal()
106 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c16__neon_mlal_padal()
122 int32x4_t vacc3x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c16__neon_mlal_padal()
357 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_4x16c16__neon_mlal_padal()
376 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_4x16c16__neon_mlal_padal()
442 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_4x16c16__neon_mlal_padal()
|
| D | 4x16c8-minmax-neon-mlal-padal.c | 74 …int32x4_t vacc0x15 = vld1q_lane_s32(w, vmovq_n_s32(0), 0); w = (const void*) ((uintptr_t) w + size… in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal() local
90 int32x4_t vacc1x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal()
106 int32x4_t vacc2x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal()
122 int32x4_t vacc3x15 = vacc0x15; in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal()
377 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal()
532 vacc0x15 = vpadalq_s16(vacc0x15, vprod0x15); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal()
551 const int32x4_t vsum0xEF = vpaddq_s32(vacc0x14, vacc0x15); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal()
617 const int32x2_t vpsum0xF = vadd_s32(vget_low_s32(vacc0x15), vget_high_s32(vacc0x15)); in xnn_qs8_igemm_minmax_ukernel_4x16c8__neon_mlal_padal()
|