ROCm/composable_kernel
1
0
Fork 0
mirror of https://github.com/ROCm/composable_kernel.git synced 2026-05-17 19:40:04 +00:00
Code Issues Packages Projects Releases Wiki Activity

Integrate universal gemm with conv bwd data and add SplitK (#1315)

Browse Source 
* Integrate universal gemm with conv bwd data

* Fix multi d kernel

* Add splitK support

* instances refactor

* instances refactor

* refactor

* fixeS

* fixes

* 16x16 instnaces

* Fixes

* Fix

* Fix

* Fix

* Fix

* Fix

* Fixes

* fix

* fix

[ROCm/composable_kernel commit: 4094ad158a]
This commit is contained in:
Bartłomiej Kocot
2025-04-28 23:54:49 +02:00
committed by GitHub
parent 02ef8bcfb1
commit 05f9b2dde3
69 changed files with 2262 additions and 349 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
test/grouped_convnd_bwd_data/CMakeLists.txt
View File

@@ -2,6 +2,11 @@ add_gtest_executable(test_grouped_convnd_bwd_data_xdl test_grouped_convnd_bwd_da
if(result EQUAL 0)
target_link_libraries(test_grouped_convnd_bwd_data_xdl PRIVATE utility device_grouped_conv2d_bwd_data_instance device_grouped_conv3d_bwd_data_instance)
endif()
if(GPU_TARGETS MATCHES "gfx9")
add_executable(test_grouped_convnd_bwd_data_xdl_large_cases test_grouped_convnd_bwd_data_xdl_large_cases.cpp)
target_compile_options(test_grouped_convnd_bwd_data_xdl_large_cases PRIVATE -Wno-global-constructors -Wno-undef)
target_link_libraries(test_grouped_convnd_bwd_data_xdl_large_cases PRIVATE gtest_main getopt::getopt utility device_grouped_conv2d_bwd_data_instance device_grouped_conv3d_bwd_data_instance)
endif()
add_gtest_executable(test_grouped_convnd_bwd_data_wmma test_grouped_convnd_bwd_data_wmma.cpp)
if(result EQUAL 0)
target_link_libraries(test_grouped_convnd_bwd_data_wmma PRIVATE utility device_grouped_conv2d_bwd_data_instance device_grouped_conv3d_bwd_data_instance)

70
test/grouped_convnd_bwd_data/test_grouped_convnd_bwd_data_xdl.cpp
View File

@@ -21,26 +21,31 @@ class TestGroupedConvndBwdDataXdl : public ::testing::Test
using InLayout = std::tuple_element_t<3, Tuple>;
std::vector<ck::utils::conv::ConvParam> conv_params;
std::vector<ck::index_t> split_ks{1, 2};
template <ck::index_t NDimSpatial>
void Run()
{
EXPECT_FALSE(conv_params.empty());
bool pass = true;
for(auto& param : conv_params)
for(auto split_k : split_ks)
{
pass = pass && ck::profiler::profile_grouped_conv_bwd_data_impl<NDimSpatial,
OutLayout,
WeiLayout,
InLayout,
DataType,
DataType,
DataType>(
true, // do_verification
1, // init_method: integer value
false, // do_log
false, // time_kernel
param);
for(auto& param : conv_params)
{
pass = pass && ck::profiler::profile_grouped_conv_bwd_data_impl<NDimSpatial,
OutLayout,
WeiLayout,
InLayout,
DataType,
DataType,
DataType>(
true, // do_verification
1, // init_method: integer value
false, // do_log
false, // time_kernel
param,
split_k);
}
}
EXPECT_TRUE(pass);
}
@@ -92,19 +97,16 @@ TYPED_TEST(TestGroupedConvndBwdDataXdl2d, Test2D)
this->conv_params.clear();
this->conv_params.push_back(
{2, 2, 4, 192, 192, {3, 3}, {28, 28}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
{2, 2, 2, 192, 192, {3, 3}, {28, 28}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
this->conv_params.push_back(
{2, 2, 128, 128, 256, {3, 3}, {14, 14}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
{2, 2, 2, 128, 256, {3, 3}, {14, 14}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
this->conv_params.push_back(
{2, 2, 128, 128, 256, {1, 1}, {7, 7}, {2, 2}, {1, 1}, {0, 0}, {0, 0}});
{2, 2, 2, 128, 256, {1, 1}, {7, 7}, {2, 2}, {1, 1}, {0, 0}, {0, 0}});
this->conv_params.push_back(
{2, 2, 128, 128, 256, {1, 1}, {3, 3}, {1, 1}, {1, 1}, {0, 0}, {0, 0}});
this->conv_params.push_back({2, 1, 1, 1, 32, {8, 8}, {32, 32}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
this->conv_params.push_back({2, 1, 1, 64, 3, {8, 8}, {32, 32}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
this->conv_params.push_back({2, 1, 1, 1, 1, {8, 8}, {32, 32}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
// SplitN case
this->conv_params.push_back(
{2, 1, 128, 4, 192, {2, 2}, {224, 224}, {224, 224}, {1, 1}, {0, 0}, {0, 0}});
{2, 2, 2, 128, 256, {1, 1}, {3, 3}, {1, 1}, {1, 1}, {0, 0}, {0, 0}});
this->conv_params.push_back({2, 1, 1, 1, 32, {8, 8}, {16, 16}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
this->conv_params.push_back({2, 1, 1, 64, 3, {8, 8}, {16, 16}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
this->conv_params.push_back({2, 1, 1, 1, 1, {8, 8}, {16, 16}, {1, 1}, {1, 1}, {1, 1}, {1, 1}});
this->template Run<2>();
}
@@ -112,28 +114,16 @@ TYPED_TEST(TestGroupedConvndBwdDataXdl3d, Test3D)
{
this->conv_params.clear();
this->conv_params.push_back(
{3, 2, 16, 128, 256, {1, 1, 1}, {7, 7, 7}, {2, 2, 2}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}});
{3, 2, 2, 128, 256, {1, 1, 1}, {7, 7, 7}, {2, 2, 2}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}});
this->conv_params.push_back(
{3, 2, 2, 128, 256, {3, 3, 3}, {14, 14, 3}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}});
this->conv_params.push_back(
{3, 2, 32, 128, 256, {1, 1, 1}, {3, 3, 3}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}});
{3, 2, 2, 128, 256, {1, 1, 1}, {3, 3, 3}, {1, 1, 1}, {1, 1, 1}, {0, 0, 0}, {0, 0, 0}});
this->conv_params.push_back(
{3, 1, 1, 1, 32, {3, 3, 3}, {32, 32, 32}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}});
{3, 1, 1, 1, 32, {3, 3, 3}, {4, 16, 16}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}});
this->conv_params.push_back(
{3, 1, 1, 64, 3, {3, 3, 3}, {32, 32, 32}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}});
{3, 1, 1, 64, 3, {3, 3, 3}, {4, 16, 16}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}});
this->conv_params.push_back(
{3, 1, 1, 1, 1, {3, 3, 3}, {32, 32, 32}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}});
// SplitN case
this->conv_params.push_back({3,
1,
128,
4,
192,
{2, 2, 2},
{2, 224, 224},
{1, 224, 224},
{1, 1, 1},
{0, 0, 0},
{0, 0, 0}});
{3, 1, 1, 1, 1, {3, 3, 3}, {4, 16, 16}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}, {1, 1, 1}});
this->template Run<3>();
}

120
test/grouped_convnd_bwd_data/test_grouped_convnd_bwd_data_xdl_large_cases.cpp Normal file
View File

@@ -0,0 +1,120 @@
// SPDX-License-Identifier: MIT
// Copyright (c) 2025, Advanced Micro Devices, Inc. All rights reserved.
#include <cstdlib>
#include <iostream>
#include <initializer_list>
#include <tuple>
#include <vector>
#include <gtest/gtest.h>
#include "profiler/profile_grouped_conv_bwd_data_impl.hpp"
template <typename Tuple>
class TestGroupedConvndBwdDataXdl : public ::testing::Test
{
protected:
using DataType = std::tuple_element_t<0, Tuple>;
using OutLayout = std::tuple_element_t<1, Tuple>;
using WeiLayout = std::tuple_element_t<2, Tuple>;
using InLayout = std::tuple_element_t<3, Tuple>;
std::vector<ck::utils::conv::ConvParam> conv_params;
std::vector<ck::index_t> split_ks{1, 2};
template <ck::index_t NDimSpatial>
void Run()
{
EXPECT_FALSE(conv_params.empty());
bool pass = true;
for(auto split_k : split_ks)
{
for(auto& param : conv_params)
{
pass = pass && ck::profiler::profile_grouped_conv_bwd_data_impl<NDimSpatial,
OutLayout,
WeiLayout,
InLayout,
DataType,
DataType,
DataType>(
true, // do_verification
1, // init_method: integer value
false, // do_log
false, // time_kernel
param,
split_k);
}
}
EXPECT_TRUE(pass);
}
};
using namespace ck::tensor_layout::convolution;
using KernelTypes2d = ::testing::Types<std::tuple<float, GNHWK, GKYXC, GNHWC>,
std::tuple<ck::half_t, GNHWK, GKYXC, GNHWC>,
std::tuple<ck::bhalf_t, GNHWK, GKYXC, GNHWC>,
std::tuple<float, NGKHW, GKYXC, NGCHW>,
std::tuple<ck::half_t, NGKHW, GKYXC, NGCHW>,
std::tuple<ck::bhalf_t, NGKHW, GKYXC, NGCHW>,
std::tuple<float, NGKHW, GKCYX, NGCHW>,
std::tuple<ck::half_t, NGKHW, GKCYX, NGCHW>,
std::tuple<ck::bhalf_t, NGKHW, GKCYX, NGCHW>,
std::tuple<float, NHWGK, GKYXC, NHWGC>,
std::tuple<ck::half_t, NHWGK, GKYXC, NHWGC>,
std::tuple<ck::bhalf_t, NHWGK, GKYXC, NHWGC>>;
using KernelTypes3d = ::testing::Types<std::tuple<float, GNDHWK, GKZYXC, GNDHWC>,
std::tuple<ck::half_t, GNDHWK, GKZYXC, GNDHWC>,
std::tuple<ck::bhalf_t, GNDHWK, GKZYXC, GNDHWC>,
std::tuple<float, NGKDHW, GKZYXC, NGCDHW>,
std::tuple<ck::half_t, NGKDHW, GKZYXC, NGCDHW>,
std::tuple<ck::bhalf_t, NGKDHW, GKZYXC, NGCDHW>,
std::tuple<float, NGKDHW, GKCZYX, NGCDHW>,
std::tuple<ck::half_t, NGKDHW, GKCZYX, NGCDHW>,
std::tuple<ck::bhalf_t, NGKDHW, GKCZYX, NGCDHW>,
std::tuple<float, NDHWGK, GKZYXC, NDHWGC>,
std::tuple<ck::half_t, NDHWGK, GKZYXC, NDHWGC>,
std::tuple<ck::bhalf_t, NDHWGK, GKZYXC, NDHWGC>>;
template <typename Tuple>
class TestGroupedConvndBwdDataXdl2d : public TestGroupedConvndBwdDataXdl<Tuple>
{
};
template <typename Tuple>
class TestGroupedConvndBwdDataXdl3d : public TestGroupedConvndBwdDataXdl<Tuple>
{
};
TYPED_TEST_SUITE(TestGroupedConvndBwdDataXdl2d, KernelTypes2d);
TYPED_TEST_SUITE(TestGroupedConvndBwdDataXdl3d, KernelTypes3d);
TYPED_TEST(TestGroupedConvndBwdDataXdl2d, Test2D)
{
this->conv_params.clear();
// SplitN case
this->conv_params.push_back(
{2, 1, 128, 4, 192, {2, 2}, {224, 224}, {224, 224}, {1, 1}, {0, 0}, {0, 0}});
this->template Run<2>();
}
TYPED_TEST(TestGroupedConvndBwdDataXdl3d, Test3D)
{
this->conv_params.clear();
// SplitN case
this->conv_params.push_back({3,
1,
128,
4,
192,
{2, 2, 2},
{2, 224, 224},
{1, 224, 224},
{1, 1, 1},
{0, 0, 0},
{0, 0, 0}});
this->template Run<3>();
}