| /dports/graphics/vapoursynth-waifu2x-ncnn-vulkan/vapoursynth-waifu2x-ncnn-vulkan-r4/deps/ncnn/src/layer/arm/ |
| H A D | convolution_arm.cpp | 203 pd.set(1, kernel_w); in create_pipeline()
238 const int maxk = kernel_w * kernel_h; in create_pipeline()
463 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward()
826 const int maxk = kernel_w * kernel_h; in forward()
837 for (int j = 0; j < kernel_w; j++) in forward()
899 const int maxk = kernel_w * kernel_h; in create_pipeline_fp16s()
1082 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward_fp16s()
1141 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward_fp16sa()
1544 const int maxk = kernel_w * kernel_h; in create_pipeline_bf16s()
1821 const int maxk = kernel_w * kernel_h; in create_pipeline_int8_arm()
[all …]
|
| H A D | convolutiondepthwise_arm.cpp | 119 const int maxk = kernel_w * kernel_h; in create_pipeline()
231 const int maxk = kernel_w * kernel_h; in create_group_ops()
256 pd.set(1, kernel_w); in create_group_ops()
452 const int maxk = kernel_w * kernel_h; in forward()
653 const int maxk = kernel_w * kernel_h; in forward_fp16s()
715 const int maxk = kernel_w * kernel_h; in forward_fp16s()
901 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
963 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
1044 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
1239 const int maxk = kernel_w * kernel_h; in forward_bf16s()
[all …]
|
| /dports/graphics/waifu2x-ncnn-vulkan/waifu2x-ncnn-vulkan-20210521/src/ncnn/src/layer/arm/ |
| H A D | convolution_arm.cpp | 203 pd.set(1, kernel_w); in create_pipeline()
238 const int maxk = kernel_w * kernel_h; in create_pipeline()
463 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward()
826 const int maxk = kernel_w * kernel_h; in forward()
837 for (int j = 0; j < kernel_w; j++) in forward()
899 const int maxk = kernel_w * kernel_h; in create_pipeline_fp16s()
1082 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward_fp16s()
1141 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward_fp16sa()
1544 const int maxk = kernel_w * kernel_h; in create_pipeline_bf16s()
1821 const int maxk = kernel_w * kernel_h; in create_pipeline_int8_arm()
[all …]
|
| H A D | convolutiondepthwise_arm.cpp | 119 const int maxk = kernel_w * kernel_h; in create_pipeline()
231 const int maxk = kernel_w * kernel_h; in create_group_ops()
256 pd.set(1, kernel_w); in create_group_ops()
452 const int maxk = kernel_w * kernel_h; in forward()
653 const int maxk = kernel_w * kernel_h; in forward_fp16s()
715 const int maxk = kernel_w * kernel_h; in forward_fp16s()
901 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
963 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
1044 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
1239 const int maxk = kernel_w * kernel_h; in forward_bf16s()
[all …]
|
| /dports/benchmarks/vkpeak/vkpeak-20210430/ncnn/src/layer/arm/ |
| H A D | convolution_arm.cpp | 203 pd.set(1, kernel_w); in create_pipeline()
238 const int maxk = kernel_w * kernel_h; in create_pipeline()
463 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward()
826 const int maxk = kernel_w * kernel_h; in forward()
837 for (int j = 0; j < kernel_w; j++) in forward()
899 const int maxk = kernel_w * kernel_h; in create_pipeline_fp16s()
1082 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward_fp16s()
1141 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward_fp16sa()
1544 const int maxk = kernel_w * kernel_h; in create_pipeline_bf16s()
1821 const int maxk = kernel_w * kernel_h; in create_pipeline_int8_arm()
[all …]
|
| H A D | convolutiondepthwise_arm.cpp | 119 const int maxk = kernel_w * kernel_h; in create_pipeline()
231 const int maxk = kernel_w * kernel_h; in create_group_ops()
256 pd.set(1, kernel_w); in create_group_ops()
452 const int maxk = kernel_w * kernel_h; in forward()
653 const int maxk = kernel_w * kernel_h; in forward_fp16s()
715 const int maxk = kernel_w * kernel_h; in forward_fp16s()
901 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
963 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
1044 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
1239 const int maxk = kernel_w * kernel_h; in forward_bf16s()
[all …]
|
| /dports/misc/ncnn/ncnn-20211208/src/layer/arm/ |
| H A D | convolution_arm.cpp | 171 pd.set(1, kernel_w); in create_pipeline()
206 const int maxk = kernel_w * kernel_h; in create_pipeline()
433 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward()
796 const int maxk = kernel_w * kernel_h; in forward()
807 for (int j = 0; j < kernel_w; j++) in forward()
869 const int maxk = kernel_w * kernel_h; in create_pipeline_fp16s()
1051 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward_fp16s()
1110 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward_fp16sa()
1514 const int maxk = kernel_w * kernel_h; in create_pipeline_bf16s()
1791 const int maxk = kernel_w * kernel_h; in create_pipeline_int8_arm()
[all …]
|
| H A D | convolutiondepthwise_arm.cpp | 82 const int maxk = kernel_w * kernel_h; in create_pipeline()
196 const int maxk = kernel_w * kernel_h; in create_group_ops()
221 pd.set(1, kernel_w); in create_group_ops()
419 const int maxk = kernel_w * kernel_h; in forward()
620 const int maxk = kernel_w * kernel_h; in forward_fp16s()
682 const int maxk = kernel_w * kernel_h; in forward_fp16s()
868 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
930 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
1011 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
1207 const int maxk = kernel_w * kernel_h; in forward_bf16s()
[all …]
|
| /dports/misc/ncnn/ncnn-20211208/src/layer/riscv/ |
| H A D | convolution_riscv.cpp | 102 const int maxk = kernel_w * kernel_h; in create_pipeline()
256 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward()
287 …if (kernel_w == 1 && kernel_h == 1 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && strid… in forward()
339 …if (kernel_w == 1 && kernel_h == 1 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && strid… in forward()
392 …if (kernel_w == 1 && kernel_h == 1 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && strid… in forward()
448 const int maxk = kernel_w * kernel_h; in forward()
456 int gap = w * dilation_h - kernel_w * dilation_w; in forward()
459 for (int j = 0; j < kernel_w; j++) in forward()
523 const int maxk = kernel_w * kernel_h; in create_pipeline_fp16s()
617 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward_fp16s()
[all …]
|
| /dports/misc/ncnn/ncnn-20211208/src/layer/mips/ |
| H A D | convolution_mips.cpp | 62 const int maxk = kernel_w * kernel_h; in create_pipeline()
200 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward()
231 …if (kernel_w == 1 && kernel_h == 1 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && strid… in forward()
240 …else if (kernel_w == 1 && kernel_h == 1 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && … in forward()
283 …if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && strid… in forward()
292 …else if (kernel_w == 3 && kernel_h == 3 && dilation_w == 1 && dilation_h == 1 && stride_w == 2 && … in forward()
318 …if (kernel_w == 1 && kernel_h == 1 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && strid… in forward()
354 …if (kernel_w == 1 && kernel_h == 1 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && strid… in forward()
374 const int maxk = kernel_w * kernel_h; in forward()
382 int gap = w * dilation_h - kernel_w * dilation_w; in forward()
[all …]
|
| /dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/tests/python/contrib/test_arm_compute_lib/ |
| H A D | test_conv2d.py | 38 kernel_w, argument
66 kernel_size=(kernel_h, kernel_w),
94 kernel_max * kernel_h * kernel_w * channels * input_max,
109 kernel_w, argument
147 kernel_size=(kernel_h, kernel_w),
179 kernel_w, argument
210 "kernel_size": [[str(kernel_h), str(kernel_w)]],
309 kernel_w,
327 "kernel size": (kernel_h, kernel_w),
429 kernel_w,
[all …]
|
| /dports/graphics/realsr-ncnn-vulkan/realsr-ncnn-vulkan-20210210/src/ncnn/src/layer/arm/ |
| H A D | convolutiondepthwise_arm.cpp | 114 const int maxk = kernel_w * kernel_h; in create_pipeline()
220 const int maxk = kernel_w * kernel_h; in create_group_ops()
248 pd.set(1, kernel_w); in create_group_ops()
427 const int maxk = kernel_w * kernel_h; in forward()
624 const int maxk = kernel_w * kernel_h; in forward_fp16s()
686 const int maxk = kernel_w * kernel_h; in forward_fp16s()
872 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
934 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
1015 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
1229 const int maxk = kernel_w * kernel_h; in forward_bf16s()
[all …]
|
| H A D | convolution_arm.cpp | 165 pd.set(1, kernel_w); in create_pipeline()
200 const int maxk = kernel_w * kernel_h; in create_pipeline()
545 const int maxk = kernel_w * kernel_h; in forward()
556 for (int j = 0; j < kernel_w; j++) in forward()
1069 const int maxk = kernel_w * kernel_h; in create_pipeline_fp16s()
1246 const int maxk = kernel_w * kernel_h; in forward_fp16s()
1516 const int maxk = kernel_w * kernel_h; in forward_fp16sa()
2250 const int maxk = kernel_w * kernel_h; in create_pipeline_bf16s()
2489 const int maxk = kernel_w * kernel_h; in forward_bf16s()
2892 const int maxk = kernel_w * kernel_h; in create_pipeline_int8_arm()
[all …]
|
| /dports/misc/py-mxnet/incubator-mxnet-1.9.0/src/operator/contrib/nn/ |
| H A D | modulated_deformable_im2col.cuh | 186 const int height, const int width, const int kernel_h, const int kernel_w, in modulated_deformable_im2col_gpu_kernel() argument
200 const int c_col = c_im * kernel_h * kernel_w; in modulated_deformable_im2col_gpu_kernel()
216 for (int j = 0; j < kernel_w; ++j) { in modulated_deformable_im2col_gpu_kernel()
296 const int kernel_h, const int kernel_w, in modulated_deformable_col2im_gpu_kernel() argument
305 const int j = (index / width_col / height_col / batch_size) % kernel_w; in modulated_deformable_col2im_gpu_kernel()
307 const int c = index / width_col / height_col / batch_size / kernel_w / kernel_h; in modulated_deformable_col2im_gpu_kernel()
406 const int kernel_h, const int kernel_w, in modulated_deformable_col2im_coord_gpu_kernel() argument
422 const int deformable_group_index = c / (2 * kernel_h * kernel_w); in modulated_deformable_col2im_coord_gpu_kernel()
423 const int col_step = kernel_h * kernel_w; in modulated_deformable_col2im_coord_gpu_kernel()
430 const int offset_c = c - deformable_group_index * 2 * kernel_h * kernel_w; in modulated_deformable_col2im_coord_gpu_kernel()
[all …]
|
| /dports/misc/mxnet/incubator-mxnet-1.9.0/src/operator/contrib/nn/ |
| H A D | modulated_deformable_im2col.cuh | 186 const int height, const int width, const int kernel_h, const int kernel_w, in modulated_deformable_im2col_gpu_kernel() argument
200 const int c_col = c_im * kernel_h * kernel_w; in modulated_deformable_im2col_gpu_kernel()
216 for (int j = 0; j < kernel_w; ++j) { in modulated_deformable_im2col_gpu_kernel()
296 const int kernel_h, const int kernel_w, in modulated_deformable_col2im_gpu_kernel() argument
305 const int j = (index / width_col / height_col / batch_size) % kernel_w; in modulated_deformable_col2im_gpu_kernel()
307 const int c = index / width_col / height_col / batch_size / kernel_w / kernel_h; in modulated_deformable_col2im_gpu_kernel()
406 const int kernel_h, const int kernel_w, in modulated_deformable_col2im_coord_gpu_kernel() argument
422 const int deformable_group_index = c / (2 * kernel_h * kernel_w); in modulated_deformable_col2im_coord_gpu_kernel()
423 const int col_step = kernel_h * kernel_w; in modulated_deformable_col2im_coord_gpu_kernel()
430 const int offset_c = c - deformable_group_index * 2 * kernel_h * kernel_w; in modulated_deformable_col2im_coord_gpu_kernel()
[all …]
|
| /dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/tests/python/contrib/test_ethosn/ |
| H A D | test_conv2d.py | 46 kernel_w, argument
73 p = _get_same_padding((shape[1], shape[2]), (kernel_h, kernel_w), dilation, strides)
75 weight_shape = (kernel_h, kernel_w, shape[3] // groups, out_channels)
77 weight_shape = (kernel_h, kernel_w, out_channels, 1)
91 kernel_size=(kernel_h, kernel_w),
137 for shape, out_channels, kernel_h, kernel_w, pad, stride, dilation in trials:
141 kernel_w = kernel_h
143 stride = (1, 1) if kernel_w == 1 else (2, 2)
162 kernel_w,
309 kernel_w,
[all …]
|
| /dports/graphics/realsr-ncnn-vulkan/realsr-ncnn-vulkan-20210210/src/ncnn/src/layer/x86/ |
| H A D | convolutiondepthwise_x86.cpp | 122 const int maxk = kernel_w * kernel_h; in create_pipeline()
202 const int maxk = kernel_w * kernel_h; in create_group_ops()
230 pd.set(1, kernel_w); in create_group_ops()
322 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward()
421 const int maxk = kernel_w * kernel_h; in forward()
429 int gap = w * dilation_h - kernel_w * dilation_w; in forward()
432 for (int j = 0; j < kernel_w; j++) in forward()
486 const int maxk = kernel_w * kernel_h; in forward()
497 for (int j = 0; j < kernel_w; j++) in forward()
640 const int maxk = kernel_w * kernel_h; in create_pipeline_int8_x86()
[all …]
|
| /dports/misc/ncnn/ncnn-20211208/src/layer/x86/ |
| H A D | convolution_x86.cpp | 92 int kernel_size = kernel_w * kernel_h; in create_pipeline()
102 pd.set(1, kernel_w); in create_pipeline()
172 const int maxk = kernel_w * kernel_h; in create_pipeline()
211 …if (kernel_w == 1 && kernel_h == 1 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && strid… in create_pipeline()
313 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward()
353 const int maxk = kernel_w * kernel_h; in forward()
361 int gap = w * dilation_h - kernel_w * dilation_w; in forward()
364 for (int j = 0; j < kernel_w; j++) in forward()
1070 const int maxk = kernel_w * kernel_h; in create_pipeline_int8_x86()
1177 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward_int8_x86()
[all …]
|
| /dports/misc/py-tvm/incubator-tvm-0.6.1/src/relay/qnn/op/ |
| H A D | convolution.cc | 88 int out_channels, kernel_h, kernel_w; in GetWorkload() local
94 kernel_w = get_const_int(kernel_shape[3]); in GetWorkload()
100 kernel_w = get_const_int(kernel_shape[1]); in GetWorkload()
107 kernel_w = get_const_int(kernel_shape[1]); in GetWorkload()
116 return std::make_tuple(batch_size, in_channels, out_channels, kernel_h, kernel_w, in GetWorkload()
201 int kernel_w, int channel_multiplier) { in DepthwiseConv2DSecondTerm() argument
212 if (kernel_h * kernel_w != 1) { in DepthwiseConv2DSecondTerm()
342 int kernel_w, int out_channels) { in Conv2DSecondTerm() argument
367 if (kernel_h * kernel_w != 1) { in Conv2DSecondTerm()
585 auto term4 = DepthwiseConv2DFourthTerm(param, kernel_h, kernel_w); in QnnConv2DCanonicalize()
[all …]
|
| /dports/misc/tvm/incubator-tvm-0.6.1/src/relay/qnn/op/ |
| H A D | convolution.cc | 88 int out_channels, kernel_h, kernel_w; in GetWorkload() local
94 kernel_w = get_const_int(kernel_shape[3]); in GetWorkload()
100 kernel_w = get_const_int(kernel_shape[1]); in GetWorkload()
107 kernel_w = get_const_int(kernel_shape[1]); in GetWorkload()
116 return std::make_tuple(batch_size, in_channels, out_channels, kernel_h, kernel_w, in GetWorkload()
201 int kernel_w, int channel_multiplier) { in DepthwiseConv2DSecondTerm() argument
212 if (kernel_h * kernel_w != 1) { in DepthwiseConv2DSecondTerm()
342 int kernel_w, int out_channels) { in Conv2DSecondTerm() argument
367 if (kernel_h * kernel_w != 1) { in Conv2DSecondTerm()
585 auto term4 = DepthwiseConv2DFourthTerm(param, kernel_h, kernel_w); in QnnConv2DCanonicalize()
[all …]
|
| /dports/misc/tvm/incubator-tvm-0.6.1/topi/python/topi/nn/ |
| H A D | deformable_conv2d.py | 78 out_channel, channel, kernel_h, kernel_w = get_const_tuple(kernel.shape)
86 dilated_kernel_w = (kernel_w - 1) * dilation_w + 1
91 rx = tvm.reduce_axis((0, kernel_w), name='rx')
101 tvm.compute((batch, in_channel, kernel_h, kernel_w, out_height, out_width),
105 offset[n, c // ic_per_dgroup * (kernel_w*kernel_h*2) +
106 (kh * kernel_w + kw) * 2, y, x],
108 offset[n, c // ic_per_dgroup * (kernel_w*kernel_h*2) +
109 (kh * kernel_w + kw) * 2 + 1, y, x]))
|
| /dports/misc/py-tvm/incubator-tvm-0.6.1/topi/python/topi/nn/ |
| H A D | deformable_conv2d.py | 78 out_channel, channel, kernel_h, kernel_w = get_const_tuple(kernel.shape)
86 dilated_kernel_w = (kernel_w - 1) * dilation_w + 1
91 rx = tvm.reduce_axis((0, kernel_w), name='rx')
101 tvm.compute((batch, in_channel, kernel_h, kernel_w, out_height, out_width),
105 offset[n, c // ic_per_dgroup * (kernel_w*kernel_h*2) +
106 (kh * kernel_w + kw) * 2, y, x],
108 offset[n, c // ic_per_dgroup * (kernel_w*kernel_h*2) +
109 (kh * kernel_w + kw) * 2 + 1, y, x]))
|
| /dports/graphics/opencv/opencv-4.5.3/modules/dnn/src/opencl/ |
| H A D | conv_spatial_helper.cl | 54 const int kernel_w,
68 int filter = sX / (kernel_w*kernel_h*channels);
69 int kernel_X = sX % kernel_w;
70 int kernel_Y = (sX / kernel_w) % kernel_h;
71 int kernel_C = (sX / (kernel_w * kernel_h)) % channels;
76 …weightOut[FP*(kernel_w*kernel_h*channels*swizzleFactor) + kernel_C*(kernel_w*kernel_h*swizzleFacto…
77 …= weightIn[filter*(kernel_w*kernel_h*channels) + kernel_C*(kernel_w*kernel_h) + kernel_Y*kernel_w …
|
| /dports/misc/mxnet/incubator-mxnet-1.9.0/3rdparty/tvm/python/tvm/topi/nn/ |
| H A D | deformable_conv2d.py | 80 out_channel, channel, kernel_h, kernel_w = get_const_tuple(kernel.shape)
88 dilated_kernel_w = (kernel_w - 1) * dilation_w + 1
92 rx = te.reduce_axis((0, kernel_w), name="rx")
102 (batch, in_channel, kernel_h, kernel_w, out_height, out_width),
110 n, c // ic_per_dgroup * (kernel_w * kernel_h * 2) + (kh * kernel_w + kw) * 2, y, x
117 c // ic_per_dgroup * (kernel_w * kernel_h * 2) + (kh * kernel_w + kw) * 2 + 1,
|
| /dports/graphics/vapoursynth-waifu2x-ncnn-vulkan/vapoursynth-waifu2x-ncnn-vulkan-r4/deps/ncnn/src/layer/x86/ |
| H A D | convolution_x86.cpp | 134 int kernel_size = kernel_w * kernel_h; in create_pipeline()
144 pd.set(1, kernel_w); in create_pipeline()
214 const int maxk = kernel_w * kernel_h; in create_pipeline()
253 …if (kernel_w == 1 && kernel_h == 1 && dilation_w == 1 && dilation_h == 1 && stride_w == 1 && strid… in create_pipeline()
350 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward()
390 const int maxk = kernel_w * kernel_h; in forward()
398 int gap = w * dilation_h - kernel_w * dilation_w; in forward()
401 for (int j = 0; j < kernel_w; j++) in forward()
1074 const int maxk = kernel_w * kernel_h; in create_pipeline_int8_x86()
1182 const int kernel_extent_w = dilation_w * (kernel_w - 1) + 1; in forward_int8_x86()
[all …]
|