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

Fix batching test for PR 5977 #978

Merged
merged 1 commit into from
Apr 12, 2022
Merged

Fix batching test for PR 5977 #978

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
4 changes: 2 additions & 2 deletions SYCL/Plugin/level_zero_dynamic_batch_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -14,7 +14,7 @@
// raise the batch size up 3 times.
//
// Then the test starts enqueueing only 4 kernels before doing a wait, and
// it does that 20 times. That should cause the batch size to
// it does that 25 times. That should cause the batch size to
// be lowered to be less than 4.
//
// CKDYN: Raising QueueBatchSize to 5
Expand Down Expand Up @@ -166,7 +166,7 @@ int main(int argc, char *argv[]) {
validate(Y1, Z7, M * N);
validate(Y1, Z8, M * N);

for (size_t i = 0; i < 20; i++) {
for (size_t i = 0; i < 25; i++) {
Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Could you please clarify why this change is needed. Does test fail when 20 iterations? Or do you want to test some specific behaviour with 25 iterations?

Copy link

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Oh, I see from the comment to intel/llvm PR that test is failing when 20 iterations. Is it clear why it fails and is it expected to fail with 20 iterations?

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

In PR 5977, we are changing the thresholds that control how batching is done. This change is expected to give some performance improvements. It does change the way the batch size is lowered/raised and that's what we test here. So, a change is needed here. Thanks.

q.submit([&](sycl::handler &h) {
h.parallel_for<class u32_copy9>(sycl::range<2>{M, N},
[=](sycl::id<2> it) {
Expand Down