Lines Matching refs:ABC

6 $ABC = "0123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ"
55 …int32x4_t vacc${ABC[C:C+4]} = vld1q_s32(w); w = (const void*) ((uintptr_t) w + 4 * sizeof(int32_t)…
60           const int16x8_t vi${K}x${ABC[C:C+8]} = vmovl_s8(vld1_s8(i${K})); i${K} += 8;
61 …const int16x8_t vk${K}x${ABC[C:C+8]} = vmovl_s8(vld1_s8(w)); w = (const void*) ((uintptr_t) w + 8 …
64 …vacc${ABC[C:C+4]} = vmlal_s16(vacc${ABC[C:C+4]}, vget_low_s16(vi${K}x${ABC[C:C+8]}), vget_low_s16(…
65 …vacc${ABC[C+4:C+8]} = vmlal_s16(vacc${ABC[C+4:C+8]}, vget_high_s16(vi${K}x${ABC[C:C+8]}), vget_hig…
68         vacc${ABC[C:C+4]} = vqrdmulhq_s32(vacc${ABC[C:C+4]}, vmultiplier);
71 …vacc${ABC[C:C+4]} = vsraq_n_s32(vacc${ABC[C:C+4]}, vbicq_s32(vacc${ABC[C:C+4]}, vzero_shift_mask),…
74         vacc${ABC[C:C+4]} = vrshlq_s32(vacc${ABC[C:C+4]}, vright_shift);
78 …const int16x8_t vacc${ABC[C:C+8]} = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc${ABC[C:C+4]}), vacc…
82 …int8x16_t vout${ABC[C:C+16]} = vqmovn_high_s16(vqmovn_s16(vacc${ABC[C:C+8]}), vacc${ABC[C+8:C+16]}…
84           int8x8_t vout${ABC[C:C+8]} = vqmovn_s16(vacc${ABC[C:C+8]});
87 …const int16x8_t vacc${ABC[C:C+8]} = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc${ABC[C:C+4]}), vqmovn_…
91 …int8x16_t vout${ABC[C:C+16]} = vcombine_s8(vqmovn_s16(vacc${ABC[C:C+8]}), vqmovn_s16(vacc${ABC[C+8…
93           int8x8_t vout${ABC[C:C+8]} = vqmovn_s16(vacc${ABC[C:C+8]});
98           vout${ABC[C:C+16]} = vmaxq_s8(vout${ABC[C:C+16]}, voutput_min);
101             vout${ABC[C:C+8]} = vmax_s8(vout${ABC[C:C+8]}, voutput_min);
103             vout${ABC[C:C+8]} = vmax_s8(vout${ABC[C:C+8]}, vget_low_s8(voutput_min));
107           vout${ABC[C:C+16]} = vminq_s8(vout${ABC[C:C+16]}, voutput_max);
110             vout${ABC[C:C+8]} = vmin_s8(vout${ABC[C:C+8]}, voutput_max);
112             vout${ABC[C:C+8]} = vmin_s8(vout${ABC[C:C+8]}, vget_low_s8(voutput_max));
116           vst1q_s8(output, vout${ABC[C:C+16]}); output += 16;
118           vst1_s8(output, vout${ABC[C:C+8]}); output += 8;
124 … int32x4_t vacc${ABC[0:4]} = vld1q_s32(w); w = (const void*) ((uintptr_t) w + 4 * sizeof(int32_t));
125 … int32x4_t vacc${ABC[4:8]} = vld1q_s32(w); w = (const void*) ((uintptr_t) w + 4 * sizeof(int32_t));
129             const int16x8_t vi${K}x${ABC[0:8]} = vmovl_s8(vld1_s8(i${K})); i${K} += 8;
131             const int16x8_t vi${K}x${ABC[0:8]} = vmovl_s8(vld1_s8(i${K}));
134               const int16x8_t vk${K}x${ABC[0:8]} = vmovl_s8(vld1_s8(k)); k += 8;
136 …const int16x8_t vk${K}x${ABC[0:8]} = vmovl_s8(vld1_s8((const void*) (k + ${K * CHANNEL_TILE - 8}))…
139               const int16x8_t vk${K}x${ABC[0:8]} = vmovl_s8(vld1_s8(w));
141 …const int16x8_t vk${K}x${ABC[0:8]} = vmovl_s8(vld1_s8((const void*) ((uintptr_t) w + ${K * CHANNEL…
143 …vacc${ABC[0:4]} = vmlal_s16(vacc${ABC[0:4]}, vget_low_s16(vi${K}x${ABC[0:8]}), vget_low_s16(vk${K}…
144 …vacc${ABC[4:8]} = vmlal_s16(vacc${ABC[4:8]}, vget_high_s16(vi${K}x${ABC[0:8]}), vget_high_s16(vk${…
146         vacc${ABC[0:4]} = vqrdmulhq_s32(vacc${ABC[0:4]}, vmultiplier);
147         vacc${ABC[4:8]} = vqrdmulhq_s32(vacc${ABC[4:8]}, vmultiplier);
149 …  vacc${ABC[0:4]} = vsraq_n_s32(vacc${ABC[0:4]}, vbicq_s32(vacc${ABC[0:4]}, vzero_shift_mask), 31);
150 …  vacc${ABC[4:8]} = vsraq_n_s32(vacc${ABC[4:8]}, vbicq_s32(vacc${ABC[4:8]}, vzero_shift_mask), 31);
152         vacc${ABC[0:4]} = vrshlq_s32(vacc${ABC[0:4]}, vright_shift);
153         vacc${ABC[4:8]} = vrshlq_s32(vacc${ABC[4:8]}, vright_shift);
156 …const int16x8_t vacc${ABC[0:8]} = vqaddq_s16(vqmovn_high_s32(vqmovn_s32(vacc${ABC[0:4]}), vacc${AB…
158         int8x8_t vout${ABC[0:8]} = vqmovn_s16(vacc${ABC[0:8]});
160 …const int16x8_t vacc${ABC[0:8]} = vqaddq_s16(vcombine_s16(vqmovn_s32(vacc${ABC[0:4]}), vqmovn_s32(…
162         int8x8_t vout${ABC[0:8]} = vqmovn_s16(vacc${ABC[0:8]});
166           vout${ABC[0:8]} = vmax_s8(vout${ABC[0:8]}, voutput_min);
167           vout${ABC[0:8]} = vmin_s8(vout${ABC[0:8]}, voutput_max);
169           vout${ABC[0:8]} = vmax_s8(vout${ABC[0:8]}, vget_low_s8(voutput_min));
170           vout${ABC[0:8]} = vmin_s8(vout${ABC[0:8]}, vget_low_s8(voutput_max));
174             vst1_s8(output, vout${ABC[0:8]}); output += 8;
178 …vst1_lane_u32(__builtin_assume_aligned(output, 1), vreinterpret_u32_s8(vout${ABC[0:8]}), 0); outpu…
179               vout${ABC[0:8]} = vext_s8(vout${ABC[0:8]}, vout${ABC[0:8]}, 4);
182 …vst1_lane_u16(__builtin_assume_aligned(output, 1), vreinterpret_u16_s8(vout${ABC[0:8]}), 0); outpu…
183               vout${ABC[0:8]} = vext_s8(vout${ABC[0:8]}, vout${ABC[0:8]}, 2);
186               vst1_lane_s8(output, vout${ABC[0:8]}, 0); output += 1;
192 …vst1_lane_u32(__builtin_assume_aligned(output, 1), vreinterpret_u32_s8(vout${ABC[0:8]}), 0); outpu…
193             vout${ABC[0:8]} = vext_s8(vout${ABC[0:8]}, vout${ABC[0:8]}, 4);
196 …vst1_lane_u16(__builtin_assume_aligned(output, 1), vreinterpret_u16_s8(vout${ABC[0:8]}), 0); outpu…
197             vout${ABC[0:8]} = vext_s8(vout${ABC[0:8]}, vout${ABC[0:8]}, 2);
200             vst1_lane_s8(output, vout${ABC[0:8]}, 0); output += 1;