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

[SYCL] Xfail tests failing after driver uplift #364

Merged
merged 1 commit into from
Jul 15, 2021
Merged

[SYCL] Xfail tests failing after driver uplift #364

Show file tree
Hide file tree
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
5 changes: 4 additions & 1 deletion SYCL/AOT/gpu.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,9 @@
// REQUIRES: ocloc, gpu
// UNSUPPORTED: cuda
// CUDA is not compatible with SPIR.

//
// The test is failing with GPU RT 30.0.100.9667
// XFAIL: windows
//
// RUN: %clangxx -fsycl -fsycl-targets=spir64_gen-unknown-unknown-sycldevice -Xsycl-target-backend=spir64_gen-unknown-unknown-sycldevice "-device *" %S/Inputs/aot.cpp -o %t.out
// RUN: %GPU_RUN_PLACEHOLDER %t.out
3 changes: 3 additions & 0 deletions SYCL/AOT/multiple-devices.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,6 +10,9 @@
// UNSUPPORTED: cuda
// CUDA is not compatible with SPIR.

// The test is failing with GPU RT 30.0.100.9667
Copy link

Choose a reason for hiding this comment

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

The test also checks combination without GPU at all and it get's disabled because of unrelated GPU fail. I can't come up with combinations of options that allows to disable concrete RUN lines on concrete OS. We might consider re-arranging this test.

Copy link
Author

@vladimirlaz vladimirlaz Jul 15, 2021 •
edited
Loading

Choose a reason for hiding this comment

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

The test will run only on a machine with CPU+GPU+Accelerator available,
As for my understanding, it is built for all target devices in the single fat binary. Then fat binary is run for every target device to check that every device image is functional.

The test is executed only if all devices are available. No need to disable particular device.

Copy link

Choose a reason for hiding this comment

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

As for my understanding, it is built for all target devices in the single fat binary. Then fat binary is run for every target device to check that every device image is functional.

Yes, it is done on lines 18-22. On next lines device code is compiled manually by each of AOT compilers and different combinations of resulting binaries are supplied to a fat binary via -fsycl-add targets option. There is only CPU&FPGA run starting on line 51.

Copy link
Author

Choose a reason for hiding this comment

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

To address that it we can split the test in separate PR.

// XFAIL: windows

// 1-command compilation case
// Targeting CPU, GPU, FPGA
// RUN: %clangxx -fsycl -fsycl-targets=spir64_x86_64-unknown-unknown-sycldevice,spir64_gen-unknown-unknown-sycldevice,spir64_fpga-unknown-unknown-sycldevice -Xsycl-target-backend=spir64_gen-unknown-unknown-sycldevice "-device *" %S/Inputs/aot.cpp -o %t_all.out
Expand Down
3 changes: 3 additions & 0 deletions SYCL/DeviceCodeSplit/aot-gpu.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,9 @@
// UNSUPPORTED: cuda
// CUDA does neither support device code splitting nor SPIR.
//
// The test is failing with GPU RT 30.0.100.9667
// XFAIL: windows
//
// RUN: %clangxx -fsycl -fsycl-device-code-split=per_source \
// RUN: -fsycl-targets=spir64_gen-unknown-unknown-sycldevice \
// RUN: -Xsycl-target-backend=spir64_gen-unknown-unknown-sycldevice \
Expand Down