Repository navigation
feat: Compatible K when not a multiple of 8 #11537
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Merged
Merged
Changes from all commits
Commits
Show all changes
6 commits
Select commit
Hold shift + click to select a range
3c6c293
Add A,B,C vector sizes as parameters to GemmConfig
fee9935
Add tests and address review comments
9d94c74
Broaden arch scope beyond gfx1250 (address review)
3e30eba
Merge branch 'develop' into users/fjimenez/ck/meta_gfx1250_K_dimension
JP-Fernando 63b4de9
Fix test compilation in CI
5970868
Merge branch 'develop' into users/fjimenez/ck/meta_gfx1250_K_dimension
JP-Fernando File filter
Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
There are no files selected for viewing
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
157 changes: 157 additions & 0 deletions
157
projects/composablekernel/test/ck_tile/gemm/test_gemm_unaligned_k.cpp
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -0,0 +1,157 @@ | ||
| // Copyright (c) Advanced Micro Devices, Inc., or its affiliates. | ||
| // SPDX-License-Identifier: MIT | ||
|
|
||
| #include "gemm_utils.hpp" | ||
| #include "run_gemm_example.inc" | ||
| #include "universal_gemm_invoker.hpp" | ||
|
|
||
| #include "ck_tile/host.hpp" | ||
|
|
||
| #include "gtest/gtest.h" | ||
|
|
||
| #include <tuple> | ||
|
|
||
| using Row = ck_tile::tensor_layout::gemm::RowMajor; | ||
| using Col = ck_tile::tensor_layout::gemm::ColumnMajor; | ||
|
|
||
| template <typename PrecType, typename CDataType> | ||
| using KPaddingConfig = | ||
| GemmConfigFixedVectorSize<GemmConfigComputeV3_WMMA<PrecType>, 1, 1, kMaxVectorElems<CDataType>>; | ||
|
|
||
| template <typename Tuple> | ||
| class TestGemmUnalignedK : public ::testing::Test | ||
| { | ||
| protected: | ||
| using ADataType = std::tuple_element_t<0, Tuple>; | ||
| using BDataType = std::tuple_element_t<1, Tuple>; | ||
| using AccDataType = std::tuple_element_t<2, Tuple>; | ||
| using CDataType = std::tuple_element_t<3, Tuple>; | ||
|
|
||
| static constexpr bool check_data_type() | ||
| { | ||
| #if defined(ARCH_GFX12) | ||
| #if defined(CK_USE_GFX1250) | ||
| using DeviceIp = ck_tile::gfx125_t; | ||
| #else | ||
| using DeviceIp = ck_tile::gfx120_t; | ||
| #endif | ||
| #elif defined(ARCH_GFX11) | ||
| using DeviceIp = ck_tile::gfx11_t; | ||
| #else | ||
| #error "Unsupported architecture for WMMA" | ||
| #endif | ||
| using Config = KPaddingConfig<ADataType, CDataType>; | ||
| return ck_tile::has_wmma_traits_v<DeviceIp, | ||
| ADataType, | ||
| BDataType, | ||
| AccDataType, | ||
| Config::M_Warp_Tile, | ||
| Config::N_Warp_Tile, | ||
| Config::K_Warp_Tile>; | ||
| } | ||
|
|
||
| void RunAndVerify(int K, int k_batch = 1) | ||
| { | ||
| if constexpr(!check_data_type()) | ||
| { | ||
| GTEST_SKIP() << "Unsupported data type combination for this architecture."; | ||
| } | ||
| else | ||
| { | ||
| constexpr int M = 128; | ||
| constexpr int N = 128; | ||
|
|
||
| const ck_tile::index_t stride_A = K; | ||
| const ck_tile::index_t stride_B = K; | ||
| const ck_tile::index_t stride_C = N; | ||
|
|
||
| // Host tensors | ||
| ck_tile::HostTensor<ADataType> a_m_k( | ||
| ck_tile::host_tensor_descriptor(M, K, stride_A, ck_tile::bool_constant<true>{})); | ||
| ck_tile::HostTensor<BDataType> b_k_n( | ||
| ck_tile::host_tensor_descriptor(K, N, stride_B, ck_tile::bool_constant<false>{})); | ||
| ck_tile::HostTensor<CDataType> gpu_result( | ||
| ck_tile::host_tensor_descriptor(M, N, stride_C, ck_tile::bool_constant<true>{})); | ||
|
|
||
| ck_tile::FillUniformDistributionIntegerValue<ADataType>{-5, 5, 11939}(a_m_k); | ||
| ck_tile::FillUniformDistributionIntegerValue<BDataType>{-5, 5, 11940}(b_k_n); | ||
|
|
||
| // Device buffers | ||
| ck_tile::DeviceMem a_buf(a_m_k.get_element_space_size_in_bytes()); | ||
| ck_tile::DeviceMem b_buf(b_k_n.get_element_space_size_in_bytes()); | ||
| ck_tile::DeviceMem c_buf(gpu_result.get_element_space_size_in_bytes()); | ||
| a_buf.ToDevice(a_m_k.data()); | ||
| b_buf.ToDevice(b_k_n.data()); | ||
| c_buf.SetZero(); | ||
|
|
||
| ck_tile::GemmHostArgs args = {a_buf.GetDeviceBuffer(), | ||
| b_buf.GetDeviceBuffer(), | ||
| c_buf.GetDeviceBuffer(), | ||
| k_batch, | ||
| M, | ||
| N, | ||
| K, | ||
| stride_A, | ||
| stride_B, | ||
| stride_C}; | ||
|
|
||
| // Run UniversalInvoker::gemm() | ||
| UniversalInvoker::gemm<KPaddingConfig<ADataType, CDataType>, | ||
| ADataType, | ||
| BDataType, | ||
| ck_tile::tuple<>, | ||
| AccDataType, | ||
| CDataType, | ||
| Row, | ||
| Col, | ||
| ck_tile::tuple<>, | ||
| Row, | ||
| /*Persistent=*/false, | ||
| ck_tile::element_wise::PassThrough>( | ||
| args, ck_tile::stream_config{nullptr, false}); | ||
|
|
||
| c_buf.FromDevice(gpu_result.data()); | ||
|
|
||
| ck_tile::HostTensor<CDataType> host_reference( | ||
| ck_tile::host_tensor_descriptor(M, N, stride_C, ck_tile::bool_constant<true>{})); | ||
| host_reference.SetZero(); | ||
| ck_tile::reference_gemm<ADataType, BDataType, AccDataType, CDataType>( | ||
| a_m_k, b_k_n, host_reference); | ||
|
|
||
| const float max_accumulated_value = | ||
| *std::max_element(host_reference.mData.begin(), host_reference.mData.end()); | ||
| const auto rtol_atol = | ||
| calculate_rtol_atol<ADataType, BDataType, AccDataType, CDataType>( | ||
| K, k_batch, max_accumulated_value); | ||
|
|
||
| // Compare both results | ||
| EXPECT_TRUE(do_verify(gpu_result, host_reference, rtol_atol, "GPU")) | ||
| << "K=" << K << ", k_batch=" << k_batch; | ||
| } | ||
| } | ||
| }; | ||
|
|
||
| using UnalignedKDataTypes = | ||
| ::testing::Types<std::tuple<ck_tile::half_t, ck_tile::half_t, float, ck_tile::half_t>, | ||
| std::tuple<ck_tile::bf16_t, ck_tile::bf16_t, float, ck_tile::bf16_t>, | ||
| std::tuple<ck_tile::fp8_t, ck_tile::fp8_t, float, ck_tile::half_t>, | ||
| std::tuple<ck_tile::bf8_t, ck_tile::bf8_t, float, ck_tile::half_t>, | ||
| std::tuple<ck_tile::int8_t, ck_tile::int8_t, int32_t, int32_t>>; | ||
| TYPED_TEST_SUITE(TestGemmUnalignedK, UnalignedKDataTypes); | ||
|
|
||
| TYPED_TEST(TestGemmUnalignedK, SmallKAndTail) | ||
| { | ||
| // Cover a K smaller than the maximum load width and a full K tile plus a tail. | ||
| constexpr int KTile = GemmConfigComputeV3_WMMA<typename TestFixture::ADataType>::K_Tile; | ||
| for(int K : {5, KTile + 5}) | ||
| { | ||
| this->RunAndVerify(K); | ||
| } | ||
| } | ||
|
|
||
| TYPED_TEST(TestGemmUnalignedK, SplitK) | ||
| { | ||
| // The first split is a full warp tile; the second contains the K tail. | ||
| constexpr int K = GemmConfigComputeV3_WMMA<typename TestFixture::ADataType>::K_Warp_Tile + 5; | ||
| this->RunAndVerify(K, /*k_batch=*/2); | ||
| } |
Oops, something went wrong.
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Uh oh!
There was an error while loading. Please reload this page.