| /dports/graphics/vapoursynth-waifu2x-ncnn-vulkan/vapoursynth-waifu2x-ncnn-vulkan-r4/deps/ncnn/src/layer/arm/ |
| H A D | convolution_3x3_pack8to1_fp16s.h | 1058 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack8to1_fp16sa_neon() local
1078 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
1087 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
|
| H A D | convolution_3x3_pack4to1_bf16s.h | 1973 const float* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack4to1_bf16s_neon() local
1993 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
2002 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
|
| H A D | convolution_3x3_pack8to4_fp16s.h | 1083 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
1097 float16x4_t _out0tm7 = vld1_f16(output0_tm_7); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
1146 output0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_fp16s.h | 1027 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4_fp16sa_neon() local
1041 float16x4_t _out0tm7 = vld1_f16(output0_tm_7); in conv3x3s1_winograd64_pack4_fp16sa_neon()
1090 output0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4to1.h | 2230 const float* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack4to1_neon() local
2250 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack4to1_neon()
2259 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack4to1_neon()
|
| /dports/graphics/waifu2x-ncnn-vulkan/waifu2x-ncnn-vulkan-20210521/src/ncnn/src/layer/arm/ |
| H A D | convolution_3x3_pack8to1_fp16s.h | 1058 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack8to1_fp16sa_neon() local
1078 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
1087 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
|
| H A D | convolution_3x3_pack4to1_bf16s.h | 1973 const float* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack4to1_bf16s_neon() local
1993 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
2002 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
|
| H A D | convolution_3x3_pack8to4_fp16s.h | 1083 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
1097 float16x4_t _out0tm7 = vld1_f16(output0_tm_7); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
1146 output0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_fp16s.h | 1027 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4_fp16sa_neon() local
1041 float16x4_t _out0tm7 = vld1_f16(output0_tm_7); in conv3x3s1_winograd64_pack4_fp16sa_neon()
1090 output0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4to1.h | 2230 const float* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack4to1_neon() local
2250 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack4to1_neon()
2259 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack4to1_neon()
|
| /dports/benchmarks/vkpeak/vkpeak-20210430/ncnn/src/layer/arm/ |
| H A D | convolution_3x3_pack8to1_fp16s.h | 1058 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack8to1_fp16sa_neon() local
1078 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
1087 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
|
| H A D | convolution_3x3_pack4to1_bf16s.h | 1973 const float* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack4to1_bf16s_neon() local
1993 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
2002 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
|
| H A D | convolution_3x3_pack8to4_fp16s.h | 1083 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
1097 float16x4_t _out0tm7 = vld1_f16(output0_tm_7); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
1146 output0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_fp16s.h | 1027 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4_fp16sa_neon() local
1041 float16x4_t _out0tm7 = vld1_f16(output0_tm_7); in conv3x3s1_winograd64_pack4_fp16sa_neon()
1090 output0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4to1.h | 2230 const float* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack4to1_neon() local
2250 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack4to1_neon()
2259 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack4to1_neon()
|
| /dports/misc/ncnn/ncnn-20211208/src/layer/arm/ |
| H A D | convolution_3x3_pack8to1_fp16s.h | 1058 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack8to1_fp16sa_neon() local
1078 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
1087 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
|
| H A D | convolution_3x3_pack4to1_bf16s.h | 1973 const float* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack4to1_bf16s_neon() local
1993 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
2002 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
|
| H A D | convolution_3x3_pack8to4_fp16s.h | 1083 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
1097 float16x4_t _out0tm7 = vld1_f16(output0_tm_7); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
1146 output0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_fp16s.h | 1027 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4_fp16sa_neon() local
1041 float16x4_t _out0tm7 = vld1_f16(output0_tm_7); in conv3x3s1_winograd64_pack4_fp16sa_neon()
1090 output0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4to1.h | 2230 const float* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack4to1_neon() local
2250 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack4to1_neon()
2259 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack4to1_neon()
|
| /dports/graphics/realsr-ncnn-vulkan/realsr-ncnn-vulkan-20210210/src/ncnn/src/layer/arm/ |
| H A D | convolution_3x3_pack8to1_fp16s.h | 1058 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack8to1_fp16sa_neon() local
1078 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
1087 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack8to1_fp16sa_neon()
|
| H A D | convolution_3x3_pack4to1_bf16s.h | 1973 const float* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack4to1_bf16s_neon() local
1993 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
2002 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack4to1_bf16s_neon()
|
| H A D | convolution_3x3_pack8to4_fp16s.h | 1083 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
1097 float16x4_t _out0tm7 = vld1_f16(output0_tm_7); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
1146 output0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_fp16s.h | 1027 const __fp16* output0_tm_7 = output0_tm_0 + tiles * 28; in conv3x3s1_winograd64_pack4_fp16sa_neon() local
1041 float16x4_t _out0tm7 = vld1_f16(output0_tm_7); in conv3x3s1_winograd64_pack4_fp16sa_neon()
1090 output0_tm_7 += tiles * 32; in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4to1.h | 2230 const float* output0_tm_7 = output0_tm_0 + tiles * 7; in conv3x3s1_winograd64_pack4to1_neon() local
2250 tmp[5][m] = output0_tm_7[0] + tmp135a + tmp135b * 32 + tmp135c; in conv3x3s1_winograd64_pack4to1_neon()
2259 output0_tm_7 += tiles * 8; in conv3x3s1_winograd64_pack4to1_neon()
|