ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-20 21:09:08 +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>

[ROCm/composable_kernel commit: 6dfb92bbef]
This commit is contained in:
Jianfeng Yan
2022-02-22 22:45:28 -06:00
committed by GitHub
parent 34ecda63ec
commit 8592df2211
61 changed files with 2255 additions and 1725 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
test/conv2d_fwd/main.cpp
View File

@@ -130,13 +130,13 @@ int main(int argc, char* argv[])
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> input_spatial_lengths{{Hi, Wi}};
const std::vector<ck::index_t> filter_spatial_lengths{{Y, X}};
const std::vector<ck::index_t> output_spatial_lengths{{Ho, Wo}};
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 std::vector<ck::index_t> input_spatial_lengths{Hi, Wi};
const std::vector<ck::index_t> filter_spatial_lengths{Y, X};
const std::vector<ck::index_t> output_spatial_lengths{Ho, Wo};
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};
auto f_host_tensor_descriptor =
[](std::size_t N_, std::size_t C_, std::size_t H, std::size_t W) {

28
test/magic_number_division/main.cpp
View File

@@ -41,6 +41,19 @@ gpu_naive_division(int32_t divisor, const int32_t* p_dividend, int32_t* p_result
}
}
__host__ void cpu_magic_number_division(uint32_t magic_multiplier,
uint32_t magic_shift,
const int32_t* p_dividend,
int32_t* p_result,
uint64_t num)
{
for(uint64_t data_id = 0; data_id < num; ++data_id)
{
p_result[data_id] =
ck::MagicDivision::DoMagicDivision(p_dividend[data_id], magic_multiplier, magic_shift);
}
}
template <typename T>
T check_error(const std::vector<T>& ref, const std::vector<T>& result)
{
@@ -90,6 +103,7 @@ int main(int, char*[])
std::vector<int32_t> naive_result_host(num_dividend);
std::vector<int32_t> magic_result_host(num_dividend);
std::vector<int32_t> magic_result_host2(num_dividend);
dividends_dev_buf.ToDevice(dividends_host.data());
@@ -128,6 +142,20 @@ int main(int, char*[])
pass = false;
continue;
}
cpu_magic_number_division(magic_multiplier,
magic_shift,
dividends_host.data(),
magic_result_host2.data(),
num_dividend);
max_diff = check_error(naive_result_host, magic_result_host2);
if(max_diff != 0)
{
pass = false;
continue;
}
}
if(pass)