| /dports/misc/ncnn/ncnn-20211208/src/layer/arm/ |
| H A D | requantize_relu_pack8.h | 123 _bias0 = vmulq_f32(_bias0, _scale_out0); in requantize_relu_pack8_neon()
139 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
141 _v2 = vfmaq_f32(_bias0, _v2, _scale0); in requantize_relu_pack8_neon()
143 _v4 = vfmaq_f32(_bias0, _v4, _scale0); in requantize_relu_pack8_neon()
145 _v6 = vfmaq_f32(_bias0, _v6, _scale0); in requantize_relu_pack8_neon()
165 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
167 _v2 = vfmaq_f32(_bias0, _v2, _scale0); in requantize_relu_pack8_neon()
170 _v0 = vmlaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
172 _v2 = vmlaq_f32(_bias0, _v2, _scale0); in requantize_relu_pack8_neon()
187 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
[all …]
|
| H A D | convolutiondepthwise_3x3_pack4.h | 215 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
315 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
395 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
509 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
610 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
688 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
761 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
950 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
1085 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
1166 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
[all …]
|
| H A D | convolutiondepthwise_3x3_pack4_bf16s.h | 287 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
433 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
538 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
690 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
823 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
919 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
1010 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
1016 float32x4_t _sum0 = _bias0; in convdw3x3s1_pack4_bf16s_neon()
1252 "w"(_bias0) // %17 in convdw3x3s2_pack4_bf16s_neon()
1360 "w"(_bias0) // %17 in convdw3x3s2_pack4_bf16s_neon()
[all …]
|
| H A D | requantize_leakyrelu_pack8.h | 124 _bias0 = vmulq_f32(_bias0, _scale_out0); in requantize_leakyrelu_pack8_neon()
141 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_leakyrelu_pack8_neon()
143 _v2 = vfmaq_f32(_bias0, _v2, _scale0); in requantize_leakyrelu_pack8_neon()
145 _v4 = vfmaq_f32(_bias0, _v4, _scale0); in requantize_leakyrelu_pack8_neon()
147 _v6 = vfmaq_f32(_bias0, _v6, _scale0); in requantize_leakyrelu_pack8_neon()
167 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_leakyrelu_pack8_neon()
169 _v2 = vfmaq_f32(_bias0, _v2, _scale0); in requantize_leakyrelu_pack8_neon()
172 _v0 = vmlaq_f32(_bias0, _v0, _scale0); in requantize_leakyrelu_pack8_neon()
174 _v2 = vmlaq_f32(_bias0, _v2, _scale0); in requantize_leakyrelu_pack8_neon()
189 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_leakyrelu_pack8_neon()
[all …]
|
| H A D | convolutiondepthwise_3x3_pack8_fp16s.h | 31 float16x8_t _bias0 = bias ? vld1q_f16(bias + g * 8) : vdupq_n_f16((__fp16)0.f); in convdw3x3s1_pack8_fp16sa_neon() local
211 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
306 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
385 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
490 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
557 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
614 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
643 float16x8_t _bias0 = bias ? vld1q_f16(bias + g * 8) : vdupq_n_f16((__fp16)0.f); in convdw3x3s2_pack8_fp16sa_neon() local
768 "w"(_bias0) // %17 in convdw3x3s2_pack8_fp16sa_neon()
843 "w"(_bias0) // %17 in convdw3x3s2_pack8_fp16sa_neon()
[all …]
|
| H A D | requantize_relu_pack4.h | 115 _bias0 = vmulq_f32(_bias0, _scale_out0); in requantize_relu_pack4_neon()
130 _v00 = vfmaq_f32(_bias0, _v00, _scale0); in requantize_relu_pack4_neon()
131 _v01 = vfmaq_f32(_bias0, _v01, _scale0); in requantize_relu_pack4_neon()
132 _v02 = vfmaq_f32(_bias0, _v02, _scale0); in requantize_relu_pack4_neon()
133 _v03 = vfmaq_f32(_bias0, _v03, _scale0); in requantize_relu_pack4_neon()
155 _v00 = vfmaq_f32(_bias0, _v00, _scale0); in requantize_relu_pack4_neon()
156 _v01 = vfmaq_f32(_bias0, _v01, _scale0); in requantize_relu_pack4_neon()
160 _v00 = vmlaq_f32(_bias0, _v00, _scale0); in requantize_relu_pack4_neon()
161 _v01 = vmlaq_f32(_bias0, _v01, _scale0); in requantize_relu_pack4_neon()
177 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack4_neon()
[all …]
|
| H A D | requantize_leakyrelu_pack4.h | 116 _bias0 = vmulq_f32(_bias0, _scale_out0); in requantize_leakyrelu_pack4_neon()
132 _v00 = vfmaq_f32(_bias0, _v00, _scale0); in requantize_leakyrelu_pack4_neon()
133 _v01 = vfmaq_f32(_bias0, _v01, _scale0); in requantize_leakyrelu_pack4_neon()
134 _v02 = vfmaq_f32(_bias0, _v02, _scale0); in requantize_leakyrelu_pack4_neon()
135 _v03 = vfmaq_f32(_bias0, _v03, _scale0); in requantize_leakyrelu_pack4_neon()
157 _v00 = vfmaq_f32(_bias0, _v00, _scale0); in requantize_leakyrelu_pack4_neon()
158 _v01 = vfmaq_f32(_bias0, _v01, _scale0); in requantize_leakyrelu_pack4_neon()
162 _v00 = vmlaq_f32(_bias0, _v00, _scale0); in requantize_leakyrelu_pack4_neon()
163 _v01 = vmlaq_f32(_bias0, _v01, _scale0); in requantize_leakyrelu_pack4_neon()
179 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_leakyrelu_pack4_neon()
[all …]
|
| /dports/graphics/vapoursynth-waifu2x-ncnn-vulkan/vapoursynth-waifu2x-ncnn-vulkan-r4/deps/ncnn/src/layer/arm/ |
| H A D | convolutiondepthwise_3x3_pack4.h | 215 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
315 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
395 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
509 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
610 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
688 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
761 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
950 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
1085 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
1166 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
[all …]
|
| H A D | convolutiondepthwise_3x3_pack4_bf16s.h | 287 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
433 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
538 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
690 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
823 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
919 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
1010 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
1016 float32x4_t _sum0 = _bias0; in convdw3x3s1_pack4_bf16s_neon()
1252 "w"(_bias0) // %17 in convdw3x3s2_pack4_bf16s_neon()
1360 "w"(_bias0) // %17 in convdw3x3s2_pack4_bf16s_neon()
[all …]
|
| H A D | requantize_relu_pack8.h | 118 …float32x4_t _bias0 = bias_data_size == 1 ? vdupq_n_f32(bias_data[0]) : vld1q_f32((const float*)bia… in requantize_relu_pack8_neon() local
123 _bias0 = vmulq_f32(_bias0, _scale_out0); in requantize_relu_pack8_neon()
139 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
141 _v2 = vfmaq_f32(_bias0, _v2, _scale0); in requantize_relu_pack8_neon()
143 _v4 = vfmaq_f32(_bias0, _v4, _scale0); in requantize_relu_pack8_neon()
145 _v6 = vfmaq_f32(_bias0, _v6, _scale0); in requantize_relu_pack8_neon()
165 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
167 _v2 = vfmaq_f32(_bias0, _v2, _scale0); in requantize_relu_pack8_neon()
231 "w"(_bias0), // %6 in requantize_relu_pack8_neon()
241 _v0 = vmlaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
[all …]
|
| H A D | convolutiondepthwise_3x3_pack8_fp16s.h | 31 float16x8_t _bias0 = bias ? vld1q_f16(bias + g * 8) : vdupq_n_f16((__fp16)0.f); in convdw3x3s1_pack8_fp16sa_neon() local
211 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
306 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
385 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
490 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
557 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
614 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
643 float16x8_t _bias0 = bias ? vld1q_f16(bias + g * 8) : vdupq_n_f16((__fp16)0.f); in convdw3x3s2_pack8_fp16sa_neon() local
768 "w"(_bias0) // %17 in convdw3x3s2_pack8_fp16sa_neon()
843 "w"(_bias0) // %17 in convdw3x3s2_pack8_fp16sa_neon()
[all …]
|
| H A D | requantize_relu_pack4.h | 110 …float32x4_t _bias0 = bias_data_size == 1 ? vdupq_n_f32(bias_data[0]) : vld1q_f32((const float*)bia… in requantize_relu_pack4_neon() local
115 _bias0 = vmulq_f32(_bias0, _scale_out0); in requantize_relu_pack4_neon()
130 _v00 = vfmaq_f32(_bias0, _v00, _scale0); in requantize_relu_pack4_neon()
131 _v01 = vfmaq_f32(_bias0, _v01, _scale0); in requantize_relu_pack4_neon()
132 _v02 = vfmaq_f32(_bias0, _v02, _scale0); in requantize_relu_pack4_neon()
133 _v03 = vfmaq_f32(_bias0, _v03, _scale0); in requantize_relu_pack4_neon()
155 _v00 = vfmaq_f32(_bias0, _v00, _scale0); in requantize_relu_pack4_neon()
156 _v01 = vfmaq_f32(_bias0, _v01, _scale0); in requantize_relu_pack4_neon()
225 "w"(_bias0), // %8 in requantize_relu_pack4_neon()
235 _v0 = vmlaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack4_neon()
[all …]
|
| /dports/graphics/waifu2x-ncnn-vulkan/waifu2x-ncnn-vulkan-20210521/src/ncnn/src/layer/arm/ |
| H A D | convolutiondepthwise_3x3_pack4.h | 215 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
315 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
395 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
509 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
610 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
688 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
761 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
950 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
1085 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
1166 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
[all …]
|
| H A D | convolutiondepthwise_3x3_pack4_bf16s.h | 287 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
433 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
538 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
690 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
823 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
919 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
1010 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
1016 float32x4_t _sum0 = _bias0; in convdw3x3s1_pack4_bf16s_neon()
1252 "w"(_bias0) // %17 in convdw3x3s2_pack4_bf16s_neon()
1360 "w"(_bias0) // %17 in convdw3x3s2_pack4_bf16s_neon()
[all …]
|
| H A D | requantize_relu_pack8.h | 118 …float32x4_t _bias0 = bias_data_size == 1 ? vdupq_n_f32(bias_data[0]) : vld1q_f32((const float*)bia… in requantize_relu_pack8_neon() local
123 _bias0 = vmulq_f32(_bias0, _scale_out0); in requantize_relu_pack8_neon()
139 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
141 _v2 = vfmaq_f32(_bias0, _v2, _scale0); in requantize_relu_pack8_neon()
143 _v4 = vfmaq_f32(_bias0, _v4, _scale0); in requantize_relu_pack8_neon()
145 _v6 = vfmaq_f32(_bias0, _v6, _scale0); in requantize_relu_pack8_neon()
165 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
167 _v2 = vfmaq_f32(_bias0, _v2, _scale0); in requantize_relu_pack8_neon()
231 "w"(_bias0), // %6 in requantize_relu_pack8_neon()
241 _v0 = vmlaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
[all …]
|
| H A D | convolutiondepthwise_3x3_pack8_fp16s.h | 31 float16x8_t _bias0 = bias ? vld1q_f16(bias + g * 8) : vdupq_n_f16((__fp16)0.f); in convdw3x3s1_pack8_fp16sa_neon() local
211 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
306 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
385 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
490 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
557 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
614 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
643 float16x8_t _bias0 = bias ? vld1q_f16(bias + g * 8) : vdupq_n_f16((__fp16)0.f); in convdw3x3s2_pack8_fp16sa_neon() local
768 "w"(_bias0) // %17 in convdw3x3s2_pack8_fp16sa_neon()
843 "w"(_bias0) // %17 in convdw3x3s2_pack8_fp16sa_neon()
[all …]
|
| H A D | requantize_relu_pack4.h | 110 …float32x4_t _bias0 = bias_data_size == 1 ? vdupq_n_f32(bias_data[0]) : vld1q_f32((const float*)bia… in requantize_relu_pack4_neon() local
115 _bias0 = vmulq_f32(_bias0, _scale_out0); in requantize_relu_pack4_neon()
130 _v00 = vfmaq_f32(_bias0, _v00, _scale0); in requantize_relu_pack4_neon()
131 _v01 = vfmaq_f32(_bias0, _v01, _scale0); in requantize_relu_pack4_neon()
132 _v02 = vfmaq_f32(_bias0, _v02, _scale0); in requantize_relu_pack4_neon()
133 _v03 = vfmaq_f32(_bias0, _v03, _scale0); in requantize_relu_pack4_neon()
155 _v00 = vfmaq_f32(_bias0, _v00, _scale0); in requantize_relu_pack4_neon()
156 _v01 = vfmaq_f32(_bias0, _v01, _scale0); in requantize_relu_pack4_neon()
225 "w"(_bias0), // %8 in requantize_relu_pack4_neon()
235 _v0 = vmlaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack4_neon()
[all …]
|
| /dports/benchmarks/vkpeak/vkpeak-20210430/ncnn/src/layer/arm/ |
| H A D | convolutiondepthwise_3x3_pack4.h | 215 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
315 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
395 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
509 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
610 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
688 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
761 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
950 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
1085 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
1166 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
[all …]
|
| H A D | convolutiondepthwise_3x3_pack4_bf16s.h | 287 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
433 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
538 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
690 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
823 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
919 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
1010 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
1016 float32x4_t _sum0 = _bias0; in convdw3x3s1_pack4_bf16s_neon()
1252 "w"(_bias0) // %17 in convdw3x3s2_pack4_bf16s_neon()
1360 "w"(_bias0) // %17 in convdw3x3s2_pack4_bf16s_neon()
[all …]
|
| H A D | requantize_relu_pack8.h | 118 …float32x4_t _bias0 = bias_data_size == 1 ? vdupq_n_f32(bias_data[0]) : vld1q_f32((const float*)bia… in requantize_relu_pack8_neon() local
123 _bias0 = vmulq_f32(_bias0, _scale_out0); in requantize_relu_pack8_neon()
139 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
141 _v2 = vfmaq_f32(_bias0, _v2, _scale0); in requantize_relu_pack8_neon()
143 _v4 = vfmaq_f32(_bias0, _v4, _scale0); in requantize_relu_pack8_neon()
145 _v6 = vfmaq_f32(_bias0, _v6, _scale0); in requantize_relu_pack8_neon()
165 _v0 = vfmaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
167 _v2 = vfmaq_f32(_bias0, _v2, _scale0); in requantize_relu_pack8_neon()
231 "w"(_bias0), // %6 in requantize_relu_pack8_neon()
241 _v0 = vmlaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack8_neon()
[all …]
|
| H A D | convolutiondepthwise_3x3_pack8_fp16s.h | 31 float16x8_t _bias0 = bias ? vld1q_f16(bias + g * 8) : vdupq_n_f16((__fp16)0.f); in convdw3x3s1_pack8_fp16sa_neon() local
211 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
306 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
385 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
490 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
557 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
614 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
643 float16x8_t _bias0 = bias ? vld1q_f16(bias + g * 8) : vdupq_n_f16((__fp16)0.f); in convdw3x3s2_pack8_fp16sa_neon() local
768 "w"(_bias0) // %17 in convdw3x3s2_pack8_fp16sa_neon()
843 "w"(_bias0) // %17 in convdw3x3s2_pack8_fp16sa_neon()
[all …]
|
| H A D | requantize_relu_pack4.h | 110 …float32x4_t _bias0 = bias_data_size == 1 ? vdupq_n_f32(bias_data[0]) : vld1q_f32((const float*)bia… in requantize_relu_pack4_neon() local
115 _bias0 = vmulq_f32(_bias0, _scale_out0); in requantize_relu_pack4_neon()
130 _v00 = vfmaq_f32(_bias0, _v00, _scale0); in requantize_relu_pack4_neon()
131 _v01 = vfmaq_f32(_bias0, _v01, _scale0); in requantize_relu_pack4_neon()
132 _v02 = vfmaq_f32(_bias0, _v02, _scale0); in requantize_relu_pack4_neon()
133 _v03 = vfmaq_f32(_bias0, _v03, _scale0); in requantize_relu_pack4_neon()
155 _v00 = vfmaq_f32(_bias0, _v00, _scale0); in requantize_relu_pack4_neon()
156 _v01 = vfmaq_f32(_bias0, _v01, _scale0); in requantize_relu_pack4_neon()
225 "w"(_bias0), // %8 in requantize_relu_pack4_neon()
235 _v0 = vmlaq_f32(_bias0, _v0, _scale0); in requantize_relu_pack4_neon()
[all …]
|
| /dports/graphics/realsr-ncnn-vulkan/realsr-ncnn-vulkan-20210210/src/ncnn/src/layer/arm/ |
| H A D | convolutiondepthwise_3x3_pack4.h | 215 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
315 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
395 "w"(_bias0) // %21 in convdw3x3s1_pack4_neon()
509 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
610 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
688 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
761 "w"(_bias0) // %17 in convdw3x3s1_pack4_neon()
950 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
1085 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
1166 "w"(_bias0) // %17 in convdw3x3s2_pack4_neon()
[all …]
|
| H A D | convolutiondepthwise_3x3_pack4_bf16s.h | 287 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
433 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
538 "w"(_bias0) // %21 in convdw3x3s1_pack4_bf16s_neon()
690 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
823 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
919 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
1010 "w"(_bias0) // %17 in convdw3x3s1_pack4_bf16s_neon()
1016 float32x4_t _sum0 = _bias0; in convdw3x3s1_pack4_bf16s_neon()
1252 "w"(_bias0) // %17 in convdw3x3s2_pack4_bf16s_neon()
1360 "w"(_bias0) // %17 in convdw3x3s2_pack4_bf16s_neon()
[all …]
|
| H A D | convolutiondepthwise_3x3_pack8_fp16s.h | 31 float16x8_t _bias0 = bias ? vld1q_f16(bias + g * 8) : vdupq_n_f16((__fp16)0.f); in convdw3x3s1_pack8_fp16sa_neon() local
211 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
306 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
385 "w"(_bias0) // %21 in convdw3x3s1_pack8_fp16sa_neon()
490 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
557 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
614 "w"(_bias0) // %17 in convdw3x3s1_pack8_fp16sa_neon()
643 float16x8_t _bias0 = bias ? vld1q_f16(bias + g * 8) : vdupq_n_f16((__fp16)0.f); in convdw3x3s2_pack8_fp16sa_neon() local
768 "w"(_bias0) // %17 in convdw3x3s2_pack8_fp16sa_neon()
843 "w"(_bias0) // %17 in convdw3x3s2_pack8_fp16sa_neon()
[all …]
|