Skip to content
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.

Sign up for GitHub

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

Jump to bottom

add tests for matrix size be runtime dimension #15429

Draft
wants to merge 8 commits into
base: sycl
Choose a base branch
from
Draft

add tests for matrix size be runtime dimension #15429

Show file tree
Hide file tree
Changes from 2 commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
d3924fe
add tests for matrix size be runtime dimension
YixingZhang007 Sep 18, 2024
4e6593f
update the arg_dim test to reuse the bf16_cache test
YixingZhang007 Sep 18, 2024
b56a34c
update the test to address the commets
YixingZhang007 Sep 18, 2024
a184cbc
add test for runtime matrix dimension
YixingZhang007 Sep 19, 2024
11aa43c
reuse joint_matmul from joint_matrix_bf16_fill_k_cache_impl.hpp in jo…
YixingZhang007 Sep 20, 2024
beb4916
remove extern for testIterations and recordThresh
YixingZhang007 Sep 20, 2024
694db15
remove the link to jira
YixingZhang007 Sep 20, 2024
4b67488
remove joint_matrix_bf16_fill_k_cache_runtime_dim_impl.hpp
YixingZhang007 Sep 20, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 22 additions & 0 deletions sycl/test-e2e/Matrix/joint_matrix_bf16_fill_k_cache_arg_dim.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,22 @@
//==--- joint_matrix_bf16_fill_k_cache_OOB.cpp - DPC++ joint_matrix--------==//
YixingZhang007 marked this conversation as resolved.
Show resolved Hide resolved
//
// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
// See https://llvm.org/LICENSE.txt for license information.
// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
//
//===----------------------------------------------------------------------===//
// REQUIRES: aspect-ext_intel_matrix

// https://jira.devtools.intel.com/browse/GSD-9716
YixingZhang007 marked this conversation as resolved.
Show resolved Hide resolved
// XFAIL: arch-intel_gpu_pvc

// RUN: %{build} -o %t_arg_dim.out -ffp-model=precise -DARG_DIM -DVNNI
YixingZhang007 marked this conversation as resolved.
Show resolved Hide resolved
// RUN: %{run} %t_arg_dim_vnni.out

// RUN: %{build} -o %t_arg_dim.out -ffp-model=precise -DARG_DIM
// RUN: %{run} %t_arg_dim.out

// -ffp-model=precise is added to not depend on compiler defaults.

#include "common.hpp"
#include "joint_matrix_bf16_fill_k_cache_impl.hpp"
19 changes: 19 additions & 0 deletions sycl/test-e2e/Matrix/joint_matrix_bf16_fill_k_cache_impl.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,11 +36,23 @@ static constexpr void manually_unroll_loop(F &&f) {

template <size_t TM, size_t TN, size_t TK> class MatMul;

#ifdef ARG_DIM
template <size_t vnniFactor, typename TOperand, typename TResult, size_t TM,
size_t TN, size_t TK, size_t MCache1, size_t NCache1, size_t KCache1,
size_t MCache2, size_t NCache2, size_t KCache2>
#else // ARG_DIM
template <size_t rowsA, size_t colsA, size_t rowsB, size_t colsB,
size_t vnniFactor, typename TOperand, typename TResult, size_t TM,
size_t TN, size_t TK, size_t MCache1, size_t NCache1, size_t KCache1,
size_t MCache2, size_t NCache2, size_t KCache2>
YixingZhang007 marked this conversation as resolved.
Show resolved Hide resolved
#endif // ARG_DIM

#ifdef ARG_DIM
double joint_matmul(TOperand *A, TOperand *B, TResult *C, queue &q, int i, size_t rowsA, size_t colsA, size_t rowsB, size_t colsB) {
#else // ARG_DIM
double joint_matmul(TOperand *A, TOperand *B, TResult *C, queue &q, int i) {
#endif // ARG_DIM

size_t sgSize = get_sg_size<MatMul<TM, TN, TK>>(q);
range<2> global{rowsA / MCache1, (colsB / NCache1) * sgSize};
range<2> cachelocal{MCache2 / MCache1, NCache2 / NCache1 * sgSize};
Expand Down Expand Up @@ -381,10 +393,17 @@ void test() {
// run testIterations time, aggregate and calculate average run time
double totalDuration = 0;
for (unsigned int i = 0; i < testIterations; i++) {
#ifdef ARG_DIM
double duration =
joint_matmul<vnniFactor, T, TResult, TM, TN, TK, MCache1, NCache1,
KCache1, MCache2, NCache2, KCache2>(A, B, C, q, i,
MATRIX_SIZE, MATRIX_SIZE, MATRIX_SIZE, MATRIX_SIZE);
#else // ARG_DIM
double duration =
joint_matmul<MATRIX_SIZE, MATRIX_SIZE, MATRIX_SIZE, MATRIX_SIZE,
vnniFactor, T, TResult, TM, TN, TK, MCache1, NCache1,
KCache1, MCache2, NCache2, KCache2>(A, B, C, q, i);
#endif // ARG_DIM
if (i >= recordThresh) {
totalDuration += duration;
}
Expand Down
Loading