Add PVC tests and move XMX8 tests to a new folder #1347
Conversation
|
@yubingex007-a11y, @myler please review |
|
failed test SYCL :: KernelAndProgram/multiple-kernel-linking.cpp unrelated to this change |
|
The |
| @@ -89,14 +89,14 @@ void matrix_sum_rows(queue q, big_matrix<T, M, N> &B, nd_range<2> &r) { | |||
| for (int i = 0; i < data.length() / (TK / 4); i++) { // 4 per row | |||
| // i*SG_SIZE index is found based on the round robin | |||
| // distribution we are using in the implementation | |||
| sum_local_rows[row + global_idx * (TK / 4)] += data[i + row * 4] | |||
| sum_local_rows[row + global_idx * (TK / 4)] += data[i + row * 4]; | |||
There was a problem hiding this comment.
@myler, this code was missing the semi colon but we were not getting the analysis that this test is failing at compile time.
How can we improve sharing the analysis of failures on the matrix code?
|
@againull, SYCL :: KernelAndProgram/kernel-bundle-merge-options-env.cpp is currently failing and unrelated to the changes I am making here. |
|
@intel/llvm-reviewers-runtime, can you please help merge this? |
|
Am I reading the tests correctly that the only differences between the existing tests and their XMX variants are the For example for SYCL/Matrix/XMX8/element_wise_all_ops_bf16.cpp and SYCL/Matrix/element_wise_all_ops_bf16.cpp you would add a common header element_wise_all_ops_bf16_impl.hpp with the test implementation without //==----------- element_wise_all_ops_bf16.cpp - DPC++ joint_matrix---------==//
//
// 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: matrix
// RUN: %clangxx -fsycl %s -o %t.out
// RUN: %CPU_RUN_PLACEHOLDER %t.out
// RUN: %GPU_RUN_PLACEHOLDER %t.out
#define SG_SIZE 16
#include "element_wise_all_ops_bf16_impl.hpp" |
This is a very good suggestion. I just made the change. |
|
@dkhaldi Could tyu please fix clang-format failure. |
|
Failures are unrelated and are most likely due to out-of-sync build being used with newer tests. Merging this. |
Disable ESIMD/vec_arg_call_conv* for xmain and xmain-rel
No description provided.