ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-11 08:50:17 +00:00
Code Issues Packages Projects Releases Wiki Activity

Conv3d new (#94)

Browse Source 
* conv3d compiles but has memory error

* conv3d works

* fix performance issue by using __builtin_amdgc_readfirstlane

* change MakeBlock2CTileMap to MakeDefaultBlock2CTileMap; change c_blockid_to* to cblockid_to*

* clang-format

* remove CK_EXPERIMENTAL_PASS_TENSOR_DECRIPTOR_BY_*; moved wrapper into DeviceConv3d

* format

* remove useless marc

* add comment

Co-authored-by: Chao Liu <chao.liu2@amd.com>
This commit is contained in:
Jianfeng Yan
2022-02-22 22:45:28 -06:00
committed by GitHub
parent 19c5d6e651
commit 6dfb92bbef
61 changed files with 2255 additions and 1725 deletions
Download Patch File Download Diff File Expand all files Collapse all files

29
example/4_conv2d_fwd_xdl/conv2d_fwd_xdl.cpp
View File

@@ -14,6 +14,7 @@
#include "element_wise_operation.hpp"
#include "device_conv2d_fwd_xdl_c_shuffle_nhwc_kyxc_nhwk.hpp"
#include "reference_conv_fwd.hpp"
#include "convolution_utility.hpp"
using InDataType = ck::half_t;
using WeiDataType = ck::half_t;
@@ -138,16 +139,20 @@ int main(int argc, char* argv[])
exit(0);
}
const ck::index_t YEff = (Y - 1) * conv_dilation_h + 1;
const ck::index_t XEff = (X - 1) * conv_dilation_w + 1;
const std::vector<ck::index_t> conv_filter_strides{conv_stride_h, conv_stride_w};
const std::vector<ck::index_t> conv_filter_dilations{conv_dilation_h, conv_dilation_w};
const std::vector<ck::index_t> input_left_pads{in_left_pad_h, in_left_pad_w};
const std::vector<ck::index_t> input_right_pads{in_right_pad_h, in_right_pad_w};
const auto output_spatial_lengths =
ck::tensor_operation::ConvolutionUtility::ComputeOutputSpatialLengths({Hi, Wi},
{Y, X},
conv_filter_strides,
conv_filter_dilations,
input_left_pads,
input_right_pads);
const ck::index_t Ho = (Hi + in_left_pad_h + in_right_pad_h - YEff) / conv_stride_h + 1;
const ck::index_t Wo = (Wi + in_left_pad_w + in_right_pad_w - XEff) / conv_stride_w + 1;
const std::vector<ck::index_t> conv_filter_strides{{conv_stride_h, conv_stride_w}};
const std::vector<ck::index_t> conv_filter_dilations{{conv_dilation_h, conv_dilation_w}};
const std::vector<ck::index_t> input_left_pads{{in_left_pad_h, in_left_pad_w}};
const std::vector<ck::index_t> input_right_pads{{in_right_pad_h, in_right_pad_w}};
const ck::index_t Ho = output_spatial_lengths[0];
const ck::index_t Wo = output_spatial_lengths[1];
// tensor layout
auto f_host_tensor_descriptor = [](std::size_t N_,
@@ -214,9 +219,9 @@ int main(int argc, char* argv[])
N,
K,
C,
std::vector<ck::index_t>{{Hi, Wi}},
std::vector<ck::index_t>{{Y, X}},
std::vector<ck::index_t>{{Ho, Wo}},
std::vector<ck::index_t>{Hi, Wi},
std::vector<ck::index_t>{Y, X},
std::vector<ck::index_t>{Ho, Wo},
conv_filter_strides,
conv_filter_dilations,
input_left_pads,