Skip to content
This repository was archived by the owner on Mar 28, 2023. It is now read-only.

[SYCL][CUDA] Enabled unified matrix tests #1334

Merged
againull merged 9 commits into from
Jan 4, 2023
Merged

[SYCL][CUDA] Enabled unified matrix tests #1334

Show file tree
Hide file tree
Changes from 5 commits
Commits
Show all changes
9 commits
Select commit Hold shift + click to select a range
5a56101
enabled tests.
JackAKirk Oct 19, 2022
21e3bb1
updated get_wi_marray() and joint_matrix.
JackAKirk Nov 4, 2022
7471d95
format
JackAKirk Nov 4, 2022
29f94c2
Use joint_matrix constructor without arg.
JackAKirk Nov 10, 2022
5b8e940
format
JackAKirk Nov 10, 2022
072c439
Updated tests to use get_wi_data free function.
JackAKirk Dec 2, 2022
d47ce3f
Merge branch 'intel' into enable-unified-matrix-tests
JackAKirk Dec 2, 2022
7714dbb
take account of bfloat16 moved out of experimental.
JackAKirk Dec 2, 2022
e2038c2
format.
JackAKirk Dec 2, 2022
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
2 changes: 0 additions & 2 deletions SYCL/Matrix/element_wise_all_ops_cuda.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,8 +6,6 @@
//
//===----------------------------------------------------------------------===//
// REQUIRES: cuda
// Temp xfail: test was merged early.
// XFAIL: cuda
// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple -Xsycl-target-backend --cuda-gpu-arch=sm_80 -DSYCL_EXT_ONEAPI_MATRIX_VERSION=4 %s -o %t.out
// RUN: %t.out

Expand Down
8 changes: 3 additions & 5 deletions SYCL/Matrix/element_wise_wi_marray.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -6,8 +6,6 @@
//
//===----------------------------------------------------------------------===//
// REQUIRES: cuda
// Temp xfail: test was merged early.
// XFAIL: cuda
// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple -Xsycl-target-backend --cuda-gpu-arch=sm_80 -DSYCL_EXT_ONEAPI_MATRIX_VERSION=4 %s -o %t.out
// RUN: %t.out

Expand Down Expand Up @@ -41,10 +39,10 @@ template <typename T, size_t M, size_t K> void verify_wi_marray(queue q) {
for (int i = 0; i < wi_slice_a.length(); i++) {
wi_slice_a[i] = fabs(wi_slice_a[i]);
}
sub_a_2.wi_marray = fabs(sub_a_2.wi_marray);
sub_a_2.get_wi_marray() = fabs(sub_a_2.get_wi_marray());

for (int i = 0; i < sub_a_2.wi_marray.size(); i++) {
if (sub_a_2.wi_marray[i] != wi_slice_a[i]) {
for (int i = 0; i < sub_a_2.get_wi_marray().size(); i++) {
if (sub_a_2.get_wi_marray()[i] != wi_slice_a[i]) {
ERR[0] = 1;
}
}
Expand Down
12 changes: 6 additions & 6 deletions SYCL/Matrix/joint_matrix_tensorcores.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,7 +1,5 @@

// REQUIRES: cuda
// Temp xfail: test was merged early.
// XFAIL: cuda
// RUN: %clangxx -fsycl -fsycl-targets=%sycl_triple -Xsycl-target-backend --cuda-gpu-arch=sm_80 -DSYCL_EXT_ONEAPI_MATRIX_VERSION=4 %s -o %t.out
// RUN: %t.out
//
Expand Down Expand Up @@ -165,11 +163,13 @@ void test(queue &q) {

// round values to correct precision if using tf32
if constexpr (std::is_same<T3, precision::tf32>::value) {
auto wi_size = sub_a.wi_marray.size();
assert(wi_size == sub_b.wi_marray.size());
auto wi_size = sub_a.get_wi_marray().size();
assert(wi_size == sub_b.get_wi_marray().size());
for (auto i = 0; i < wi_size; ++i) {
sub_a.wi_marray[i] = round_to_tf32(sub_a.wi_marray[i]);
sub_b.wi_marray[i] = round_to_tf32(sub_b.wi_marray[i]);
sub_a.get_wi_marray()[i] =
round_to_tf32(sub_a.get_wi_marray()[i]);
sub_b.get_wi_marray()[i] =
round_to_tf32(sub_b.get_wi_marray()[i]);
}
}

Expand Down