/dports/graphics/realsr-ncnn-vulkan/realsr-ncnn-vulkan-20210210/src/ncnn/src/layer/arm/ |
H A D | convolution_3x3_pack8_fp16s.h | 237 float16x8_t _tmp3m = vaddq_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack8_fp16sa_neon() local 239 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8_fp16sa_neon() 1127 … float16x8_t _tmp3m = vfmaq_n_f16(vfmaq_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack8_fp16sa_neon() local 1130 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8_fp16sa_neon() 1388 … float16x8_t _tmp3m = vfmsq_n_f16(vsubq_f16(_r04, _r02), vsubq_f16(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack8_fp16sa_neon() local 1395 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack8_fp16sa_neon() 2195 … float16x8_t _tmp3m = vfmaq_n_f16(vaddq_f16(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack8_fp16sa_neon() local 2200 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack8_fp16sa_neon()
|
H A D | convolution_3x3_pack4_bf16s.h | 133 float32x4_t _tmp3m = vaddq_f32(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_bf16s_neon() local 135 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_bf16s_neon() 1737 … float32x4_t _tmp3m = vmlaq_n_f32(vmlaq_n_f32(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_bf16s_neon() local 1740 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_bf16s_neon() 1897 … float32x4_t _tmp3m = vmlsq_n_f32(vsubq_f32(_r04, _r02), vsubq_f32(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack4_bf16s_neon() local 1904 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_bf16s_neon() 3418 … float32x4_t _tmp3m = vmlaq_n_f32(vaddq_f32(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack4_bf16s_neon() local 3423 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_bf16s_neon()
|
H A D | convolution_3x3_pack8to4_fp16s.h | 255 float16x8_t _tmp3m = vaddq_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local 257 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() 1129 … float16x4_t _tmp3m = vfma_n_f16(vfma_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local 1132 vst1_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
H A D | convolution_3x3_pack4.h | 364 float32x4_t _tmp3m = vaddq_f32(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_neon() local 366 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_neon() 1968 … float32x4_t _tmp3m = vmlaq_n_f32(vmlaq_n_f32(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_neon() local 1971 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_neon() 2357 … float32x4_t _tmp3m = vmlsq_n_f32(vsubq_f32(_r04, _r02), vsubq_f32(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack4_neon() local 2364 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_neon() 3878 … float32x4_t _tmp3m = vmlaq_n_f32(vaddq_f32(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack4_neon() local 3883 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_neon()
|
H A D | convolution_3x3_pack4_fp16s.h | 355 float16x4_t _tmp3m = vadd_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_fp16sa_neon() local 357 vst1_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_fp16sa_neon() 1073 … float16x4_t _tmp3m = vfma_n_f16(vfma_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_fp16sa_neon() local 1076 vst1_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
/dports/graphics/vapoursynth-waifu2x-ncnn-vulkan/vapoursynth-waifu2x-ncnn-vulkan-r4/deps/ncnn/src/layer/arm/ |
H A D | convolution_3x3_pack8_fp16s.h | 237 float16x8_t _tmp3m = vaddq_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack8_fp16sa_neon() local 239 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8_fp16sa_neon() 1127 … float16x8_t _tmp3m = vfmaq_n_f16(vfmaq_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack8_fp16sa_neon() local 1130 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8_fp16sa_neon() 1388 … float16x8_t _tmp3m = vfmsq_n_f16(vsubq_f16(_r04, _r02), vsubq_f16(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack8_fp16sa_neon() local 1395 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack8_fp16sa_neon() 2195 … float16x8_t _tmp3m = vfmaq_n_f16(vaddq_f16(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack8_fp16sa_neon() local 2200 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack8_fp16sa_neon()
|
H A D | convolution_3x3_pack4_bf16s.h | 133 float32x4_t _tmp3m = vaddq_f32(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_bf16s_neon() local 135 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_bf16s_neon() 1737 … float32x4_t _tmp3m = vmlaq_n_f32(vmlaq_n_f32(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_bf16s_neon() local 1740 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_bf16s_neon() 1897 … float32x4_t _tmp3m = vmlsq_n_f32(vsubq_f32(_r04, _r02), vsubq_f32(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack4_bf16s_neon() local 1904 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_bf16s_neon() 3418 … float32x4_t _tmp3m = vmlaq_n_f32(vaddq_f32(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack4_bf16s_neon() local 3423 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_bf16s_neon()
|
H A D | convolution_3x3_pack8to4_fp16s.h | 255 float16x8_t _tmp3m = vaddq_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local 257 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() 1129 … float16x4_t _tmp3m = vfma_n_f16(vfma_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local 1132 vst1_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
H A D | convolution_3x3_pack4.h | 364 float32x4_t _tmp3m = vaddq_f32(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_neon() local 366 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_neon() 1968 … float32x4_t _tmp3m = vmlaq_n_f32(vmlaq_n_f32(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_neon() local 1971 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_neon() 2357 … float32x4_t _tmp3m = vmlsq_n_f32(vsubq_f32(_r04, _r02), vsubq_f32(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack4_neon() local 2364 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_neon() 3878 … float32x4_t _tmp3m = vmlaq_n_f32(vaddq_f32(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack4_neon() local 3883 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_neon()
|
H A D | convolution_3x3_pack4_fp16s.h | 355 float16x4_t _tmp3m = vadd_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_fp16sa_neon() local 357 vst1_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_fp16sa_neon() 1073 … float16x4_t _tmp3m = vfma_n_f16(vfma_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_fp16sa_neon() local 1076 vst1_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
/dports/graphics/waifu2x-ncnn-vulkan/waifu2x-ncnn-vulkan-20210521/src/ncnn/src/layer/arm/ |
H A D | convolution_3x3_pack8_fp16s.h | 237 float16x8_t _tmp3m = vaddq_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack8_fp16sa_neon() local 239 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8_fp16sa_neon() 1127 … float16x8_t _tmp3m = vfmaq_n_f16(vfmaq_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack8_fp16sa_neon() local 1130 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8_fp16sa_neon() 1388 … float16x8_t _tmp3m = vfmsq_n_f16(vsubq_f16(_r04, _r02), vsubq_f16(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack8_fp16sa_neon() local 1395 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack8_fp16sa_neon() 2195 … float16x8_t _tmp3m = vfmaq_n_f16(vaddq_f16(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack8_fp16sa_neon() local 2200 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack8_fp16sa_neon()
|
H A D | convolution_3x3_pack4_bf16s.h | 133 float32x4_t _tmp3m = vaddq_f32(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_bf16s_neon() local 135 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_bf16s_neon() 1737 … float32x4_t _tmp3m = vmlaq_n_f32(vmlaq_n_f32(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_bf16s_neon() local 1740 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_bf16s_neon() 1897 … float32x4_t _tmp3m = vmlsq_n_f32(vsubq_f32(_r04, _r02), vsubq_f32(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack4_bf16s_neon() local 1904 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_bf16s_neon() 3418 … float32x4_t _tmp3m = vmlaq_n_f32(vaddq_f32(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack4_bf16s_neon() local 3423 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_bf16s_neon()
|
H A D | convolution_3x3_pack8to4_fp16s.h | 255 float16x8_t _tmp3m = vaddq_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local 257 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() 1129 … float16x4_t _tmp3m = vfma_n_f16(vfma_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local 1132 vst1_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
H A D | convolution_3x3_pack4.h | 364 float32x4_t _tmp3m = vaddq_f32(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_neon() local 366 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_neon() 1968 … float32x4_t _tmp3m = vmlaq_n_f32(vmlaq_n_f32(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_neon() local 1971 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_neon() 2357 … float32x4_t _tmp3m = vmlsq_n_f32(vsubq_f32(_r04, _r02), vsubq_f32(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack4_neon() local 2364 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_neon() 3878 … float32x4_t _tmp3m = vmlaq_n_f32(vaddq_f32(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack4_neon() local 3883 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_neon()
|
H A D | convolution_3x3_pack4_fp16s.h | 355 float16x4_t _tmp3m = vadd_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_fp16sa_neon() local 357 vst1_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_fp16sa_neon() 1073 … float16x4_t _tmp3m = vfma_n_f16(vfma_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_fp16sa_neon() local 1076 vst1_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
/dports/benchmarks/vkpeak/vkpeak-20210430/ncnn/src/layer/arm/ |
H A D | convolution_3x3_pack8_fp16s.h | 237 float16x8_t _tmp3m = vaddq_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack8_fp16sa_neon() local 239 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8_fp16sa_neon() 1127 … float16x8_t _tmp3m = vfmaq_n_f16(vfmaq_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack8_fp16sa_neon() local 1130 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8_fp16sa_neon() 1388 … float16x8_t _tmp3m = vfmsq_n_f16(vsubq_f16(_r04, _r02), vsubq_f16(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack8_fp16sa_neon() local 1395 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack8_fp16sa_neon() 2195 … float16x8_t _tmp3m = vfmaq_n_f16(vaddq_f16(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack8_fp16sa_neon() local 2200 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack8_fp16sa_neon()
|
H A D | convolution_3x3_pack4_bf16s.h | 133 float32x4_t _tmp3m = vaddq_f32(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_bf16s_neon() local 135 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_bf16s_neon() 1737 … float32x4_t _tmp3m = vmlaq_n_f32(vmlaq_n_f32(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_bf16s_neon() local 1740 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_bf16s_neon() 1897 … float32x4_t _tmp3m = vmlsq_n_f32(vsubq_f32(_r04, _r02), vsubq_f32(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack4_bf16s_neon() local 1904 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_bf16s_neon() 3418 … float32x4_t _tmp3m = vmlaq_n_f32(vaddq_f32(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack4_bf16s_neon() local 3423 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_bf16s_neon()
|
H A D | convolution_3x3_pack8to4_fp16s.h | 255 float16x8_t _tmp3m = vaddq_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local 257 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() 1129 … float16x4_t _tmp3m = vfma_n_f16(vfma_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local 1132 vst1_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
H A D | convolution_3x3_pack4.h | 364 float32x4_t _tmp3m = vaddq_f32(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_neon() local 366 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_neon() 1968 … float32x4_t _tmp3m = vmlaq_n_f32(vmlaq_n_f32(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_neon() local 1971 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_neon() 2357 … float32x4_t _tmp3m = vmlsq_n_f32(vsubq_f32(_r04, _r02), vsubq_f32(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack4_neon() local 2364 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_neon() 3878 … float32x4_t _tmp3m = vmlaq_n_f32(vaddq_f32(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack4_neon() local 3883 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_neon()
|
/dports/misc/ncnn/ncnn-20211208/src/layer/arm/ |
H A D | convolution_3x3_pack8_fp16s.h | 237 float16x8_t _tmp3m = vaddq_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack8_fp16sa_neon() local 239 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8_fp16sa_neon() 1127 … float16x8_t _tmp3m = vfmaq_n_f16(vfmaq_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack8_fp16sa_neon() local 1130 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8_fp16sa_neon() 1388 … float16x8_t _tmp3m = vfmsq_n_f16(vsubq_f16(_r04, _r02), vsubq_f16(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack8_fp16sa_neon() local 1395 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack8_fp16sa_neon() 2195 … float16x8_t _tmp3m = vfmaq_n_f16(vaddq_f16(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack8_fp16sa_neon() local 2200 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack8_fp16sa_neon()
|
H A D | convolution_3x3_pack4_bf16s.h | 133 float32x4_t _tmp3m = vaddq_f32(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_bf16s_neon() local 135 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_bf16s_neon() 1737 … float32x4_t _tmp3m = vmlaq_n_f32(vmlaq_n_f32(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_bf16s_neon() local 1740 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_bf16s_neon() 1897 … float32x4_t _tmp3m = vmlsq_n_f32(vsubq_f32(_r04, _r02), vsubq_f32(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack4_bf16s_neon() local 1904 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_bf16s_neon() 3418 … float32x4_t _tmp3m = vmlaq_n_f32(vaddq_f32(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack4_bf16s_neon() local 3423 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_bf16s_neon()
|
H A D | convolution_3x3_pack8to4_fp16s.h | 255 float16x8_t _tmp3m = vaddq_f16(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local 257 vst1q_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() 1129 … float16x4_t _tmp3m = vfma_n_f16(vfma_n_f16(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local 1132 vst1_f16(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
H A D | convolution_3x3_pack4.h | 364 float32x4_t _tmp3m = vaddq_f32(_tmp34a, _tmp34b); in conv3x3s1_winograd64_pack4_neon() local 366 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_neon() 1968 … float32x4_t _tmp3m = vmlaq_n_f32(vmlaq_n_f32(_tmp135a, _tmp135b, 8.f), _tmp135c, 4.f); in conv3x3s1_winograd64_pack4_neon() local 1971 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd64_pack4_neon() 2357 … float32x4_t _tmp3m = vmlsq_n_f32(vsubq_f32(_r04, _r02), vsubq_f32(_r01, _r03), 2.f); in conv3x3s1_winograd42_pack4_neon() local 2364 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_neon() 3878 … float32x4_t _tmp3m = vmlaq_n_f32(vaddq_f32(_out0tm5, _tmp13a), _tmp13b, 8.f); in conv3x3s1_winograd42_pack4_neon() local 3883 vst1q_f32(tmp[3][m], _tmp3m); in conv3x3s1_winograd42_pack4_neon()
|
/dports/misc/ncnn/ncnn-20211208/src/layer/riscv/ |
H A D | convolution_3x3_packn.h | 208 vfloat32m1_t _tmp3m = vfadd_vv_f32m1(_tmp34a, _tmp34b, vl); in conv3x3s1_winograd64_packn_rvv() local 210 vse32_v_f32m1(tmp[3][m], _tmp3m, vl); in conv3x3s1_winograd64_packn_rvv() 684 …vfloat32m1_t _tmp3m = vfmacc_vf_f32m1(vfmacc_vf_f32m1(_tmp135a, 8.f, _tmp135b, vl), 4.f, _tmp135c,… in conv3x3s1_winograd64_packn_rvv() local 687 vse32_v_f32m1(tmp[3][m], _tmp3m, vl); in conv3x3s1_winograd64_packn_rvv() 907 …vfloat32m1_t _tmp3m = vfmacc_vf_f32m1(vfsub_vv_f32m1(_r04, _r02, vl), -2.f, vfsub_vv_f32m1(_r01, _… in conv3x3s1_winograd42_packn_rvv() local 914 vse32_v_f32m1(tmp[3][m], _tmp3m, vl); in conv3x3s1_winograd42_packn_rvv() 1343 … vfloat32m1_t _tmp3m = vfmacc_vf_f32m1(vfadd_vv_f32m1(_out0tm5, _tmp13a, vl), 8.f, _tmp13b, vl); in conv3x3s1_winograd42_packn_rvv() local 1348 vse32_v_f32m1(tmp[3][m], _tmp3m, vl); in conv3x3s1_winograd42_packn_rvv()
|
H A D | convolution_3x3_packn_fp16s.h | 208 vfloat16m1_t _tmp3m = vfadd_vv_f16m1(_tmp34a, _tmp34b, vl); in conv3x3s1_winograd64_packn_fp16sa_rvv() local 210 vse16_v_f16m1(tmp[3][m], _tmp3m, vl); in conv3x3s1_winograd64_packn_fp16sa_rvv() 684 …vfloat16m1_t _tmp3m = vfmacc_vf_f16m1(vfmacc_vf_f16m1(_tmp135a, 8.f, _tmp135b, vl), 4.f, _tmp135c,… in conv3x3s1_winograd64_packn_fp16sa_rvv() local 687 vse16_v_f16m1(tmp[3][m], _tmp3m, vl); in conv3x3s1_winograd64_packn_fp16sa_rvv() 907 …vfloat16m1_t _tmp3m = vfmacc_vf_f16m1(vfsub_vv_f16m1(_r04, _r02, vl), -2.f, vfsub_vv_f16m1(_r01, _… in conv3x3s1_winograd42_packn_fp16sa_rvv() local 914 vse16_v_f16m1(tmp[3][m], _tmp3m, vl); in conv3x3s1_winograd42_packn_fp16sa_rvv() 1343 … vfloat16m1_t _tmp3m = vfmacc_vf_f16m1(vfadd_vv_f16m1(_out0tm5, _tmp13a, vl), 8.f, _tmp13b, vl); in conv3x3s1_winograd42_packn_fp16sa_rvv() local 1348 vse16_v_f16m1(tmp[3][m], _tmp3m, vl); in conv3x3s1_winograd42_packn_fp16sa_rvv()
|