| /dports/graphics/vapoursynth-waifu2x-ncnn-vulkan/vapoursynth-waifu2x-ncnn-vulkan-r4/deps/ncnn/src/layer/arm/ |
| H A D | convolution_3x3_pack8to4_fp16s.h | 1191 …float16x4_t _out05 = vadd_f16(_bias0, vadd_f16(vadd_f16(_tmp07, _tmp135a), vfma_n_f16(_tmp135c, _t… in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
1194 vst1_f16(output0 + 20, _out05); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_fp16s.h | 1135 …float16x4_t _out05 = vadd_f16(_bias0, vadd_f16(vadd_f16(_tmp07, _tmp135a), vfma_n_f16(_tmp135c, _t… in conv3x3s1_winograd64_pack4_fp16sa_neon() local
1138 vst1_f16(output0 + 20, _out05); in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
| H A D | convolution_3x3_pack8_fp16s.h | 1189 …float16x8_t _out05 = vaddq_f16(_bias0, vaddq_f16(vaddq_f16(_tmp07, _tmp135a), vfmaq_n_f16(_tmp135c… in conv3x3s1_winograd64_pack8_fp16sa_neon() local
1192 vst1q_f16(output0 + 40, _out05); in conv3x3s1_winograd64_pack8_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_bf16s.h | 1799 …float32x4_t _out05 = vaddq_f32(_bias0, vaddq_f32(vaddq_f32(_tmp07, _tmp135a), vmlaq_n_f32(_tmp135c… in conv3x3s1_winograd64_pack4_bf16s_neon() local
1802 vst1_u16(output0 + 20, vcvt_bf16_f32(_out05)); in conv3x3s1_winograd64_pack4_bf16s_neon()
|
| H A D | convolution_3x3_pack4.h | 2030 …float32x4_t _out05 = vaddq_f32(_bias0, vaddq_f32(vaddq_f32(_tmp07, _tmp135a), vmlaq_n_f32(_tmp135c… in conv3x3s1_winograd64_pack4_neon() local
2033 vst1q_f32(output0 + 20, _out05); in conv3x3s1_winograd64_pack4_neon()
|
| /dports/graphics/waifu2x-ncnn-vulkan/waifu2x-ncnn-vulkan-20210521/src/ncnn/src/layer/arm/ |
| H A D | convolution_3x3_pack8to4_fp16s.h | 1191 …float16x4_t _out05 = vadd_f16(_bias0, vadd_f16(vadd_f16(_tmp07, _tmp135a), vfma_n_f16(_tmp135c, _t… in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
1194 vst1_f16(output0 + 20, _out05); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_fp16s.h | 1135 …float16x4_t _out05 = vadd_f16(_bias0, vadd_f16(vadd_f16(_tmp07, _tmp135a), vfma_n_f16(_tmp135c, _t… in conv3x3s1_winograd64_pack4_fp16sa_neon() local
1138 vst1_f16(output0 + 20, _out05); in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
| H A D | convolution_3x3_pack8_fp16s.h | 1189 …float16x8_t _out05 = vaddq_f16(_bias0, vaddq_f16(vaddq_f16(_tmp07, _tmp135a), vfmaq_n_f16(_tmp135c… in conv3x3s1_winograd64_pack8_fp16sa_neon() local
1192 vst1q_f16(output0 + 40, _out05); in conv3x3s1_winograd64_pack8_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_bf16s.h | 1799 …float32x4_t _out05 = vaddq_f32(_bias0, vaddq_f32(vaddq_f32(_tmp07, _tmp135a), vmlaq_n_f32(_tmp135c… in conv3x3s1_winograd64_pack4_bf16s_neon() local
1802 vst1_u16(output0 + 20, vcvt_bf16_f32(_out05)); in conv3x3s1_winograd64_pack4_bf16s_neon()
|
| H A D | convolution_3x3_pack4.h | 2030 …float32x4_t _out05 = vaddq_f32(_bias0, vaddq_f32(vaddq_f32(_tmp07, _tmp135a), vmlaq_n_f32(_tmp135c… in conv3x3s1_winograd64_pack4_neon() local
2033 vst1q_f32(output0 + 20, _out05); in conv3x3s1_winograd64_pack4_neon()
|
| /dports/benchmarks/vkpeak/vkpeak-20210430/ncnn/src/layer/arm/ |
| H A D | convolution_3x3_pack8to4_fp16s.h | 1191 …float16x4_t _out05 = vadd_f16(_bias0, vadd_f16(vadd_f16(_tmp07, _tmp135a), vfma_n_f16(_tmp135c, _t… in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
1194 vst1_f16(output0 + 20, _out05); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_fp16s.h | 1135 …float16x4_t _out05 = vadd_f16(_bias0, vadd_f16(vadd_f16(_tmp07, _tmp135a), vfma_n_f16(_tmp135c, _t… in conv3x3s1_winograd64_pack4_fp16sa_neon() local
1138 vst1_f16(output0 + 20, _out05); in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
| H A D | convolution_3x3_pack8_fp16s.h | 1189 …float16x8_t _out05 = vaddq_f16(_bias0, vaddq_f16(vaddq_f16(_tmp07, _tmp135a), vfmaq_n_f16(_tmp135c… in conv3x3s1_winograd64_pack8_fp16sa_neon() local
1192 vst1q_f16(output0 + 40, _out05); in conv3x3s1_winograd64_pack8_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_bf16s.h | 1799 …float32x4_t _out05 = vaddq_f32(_bias0, vaddq_f32(vaddq_f32(_tmp07, _tmp135a), vmlaq_n_f32(_tmp135c… in conv3x3s1_winograd64_pack4_bf16s_neon() local
1802 vst1_u16(output0 + 20, vcvt_bf16_f32(_out05)); in conv3x3s1_winograd64_pack4_bf16s_neon()
|
| /dports/misc/ncnn/ncnn-20211208/src/layer/arm/ |
| H A D | convolution_3x3_pack8to4_fp16s.h | 1191 …float16x4_t _out05 = vadd_f16(_bias0, vadd_f16(vadd_f16(_tmp07, _tmp135a), vfma_n_f16(_tmp135c, _t… in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
1194 vst1_f16(output0 + 20, _out05); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_fp16s.h | 1135 …float16x4_t _out05 = vadd_f16(_bias0, vadd_f16(vadd_f16(_tmp07, _tmp135a), vfma_n_f16(_tmp135c, _t… in conv3x3s1_winograd64_pack4_fp16sa_neon() local
1138 vst1_f16(output0 + 20, _out05); in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
| H A D | convolution_3x3_pack8_fp16s.h | 1189 …float16x8_t _out05 = vaddq_f16(_bias0, vaddq_f16(vaddq_f16(_tmp07, _tmp135a), vfmaq_n_f16(_tmp135c… in conv3x3s1_winograd64_pack8_fp16sa_neon() local
1192 vst1q_f16(output0 + 40, _out05); in conv3x3s1_winograd64_pack8_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_bf16s.h | 1799 …float32x4_t _out05 = vaddq_f32(_bias0, vaddq_f32(vaddq_f32(_tmp07, _tmp135a), vmlaq_n_f32(_tmp135c… in conv3x3s1_winograd64_pack4_bf16s_neon() local
1802 vst1_u16(output0 + 20, vcvt_bf16_f32(_out05)); in conv3x3s1_winograd64_pack4_bf16s_neon()
|
| /dports/graphics/realsr-ncnn-vulkan/realsr-ncnn-vulkan-20210210/src/ncnn/src/layer/arm/ |
| H A D | convolution_3x3_pack8to4_fp16s.h | 1191 …float16x4_t _out05 = vadd_f16(_bias0, vadd_f16(vadd_f16(_tmp07, _tmp135a), vfma_n_f16(_tmp135c, _t… in conv3x3s1_winograd64_pack8to4_fp16sa_neon() local
1194 vst1_f16(output0 + 20, _out05); in conv3x3s1_winograd64_pack8to4_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_fp16s.h | 1135 …float16x4_t _out05 = vadd_f16(_bias0, vadd_f16(vadd_f16(_tmp07, _tmp135a), vfma_n_f16(_tmp135c, _t… in conv3x3s1_winograd64_pack4_fp16sa_neon() local
1138 vst1_f16(output0 + 20, _out05); in conv3x3s1_winograd64_pack4_fp16sa_neon()
|
| H A D | convolution_3x3_pack8_fp16s.h | 1189 …float16x8_t _out05 = vaddq_f16(_bias0, vaddq_f16(vaddq_f16(_tmp07, _tmp135a), vfmaq_n_f16(_tmp135c… in conv3x3s1_winograd64_pack8_fp16sa_neon() local
1192 vst1q_f16(output0 + 40, _out05); in conv3x3s1_winograd64_pack8_fp16sa_neon()
|
| H A D | convolution_3x3_pack4_bf16s.h | 1799 …float32x4_t _out05 = vaddq_f32(_bias0, vaddq_f32(vaddq_f32(_tmp07, _tmp135a), vmlaq_n_f32(_tmp135c… in conv3x3s1_winograd64_pack4_bf16s_neon() local
1802 vst1_u16(output0 + 20, vcvt_bf16_f32(_out05)); in conv3x3s1_winograd64_pack4_bf16s_neon()
|
| /dports/misc/ncnn/ncnn-20211208/src/layer/riscv/ |
| H A D | convolution_3x3_packn.h | 729 …vfloat32m1_t _out05 = vfadd_vv_f32m1(_bias0, vfadd_vv_f32m1(vfadd_vv_f32m1(_tmp07, _tmp135a, vl), … in conv3x3s1_winograd64_packn_rvv() local
732 vse32_v_f32m1(output0 + packn * 5, _out05, vl); in conv3x3s1_winograd64_packn_rvv()
|
| H A D | convolution_3x3_packn_fp16s.h | 729 …vfloat16m1_t _out05 = vfadd_vv_f16m1(_bias0, vfadd_vv_f16m1(vfadd_vv_f16m1(_tmp07, _tmp135a, vl), … in conv3x3s1_winograd64_packn_fp16sa_rvv() local
732 vse16_v_f16m1(output0 + packn * 5, _out05, vl); in conv3x3s1_winograd64_packn_fp16sa_rvv()
|
| /dports/graphics/realsr-ncnn-vulkan/realsr-ncnn-vulkan-20210210/src/ncnn/src/layer/x86/ |
| H A D | convolution_3x3_pack8.h | 1471 …__m256 _out05 = _mm256_add_ps(_bias0, _mm256_add_ps(_mm256_add_ps(_tmp07, _tmp135a), _mm256_fmadd_… in conv3x3s1_winograd64_pack8_avx() local
1474 _mm256_storeu_ps(output0 + 40, _out05); in conv3x3s1_winograd64_pack8_avx()
|