bindless_images/examples/example_1_1D_read_write.cpp sporadically failing on BMG

[sarnex]

Describe the bug

Sporadically fails on BMG. Here's the log from a case where it failed.

2025-02-06T23:28:25.2800250Z FAIL: SYCL :: bindless_images/examples/example_1_1D_read_write.cpp (2177 of 2274)
2025-02-06T23:28:25.2800401Z ******************** TEST 'SYCL :: bindless_images/examples/example_1_1D_read_write.cpp' FAILED ********************
2025-02-06T23:28:25.2800451Z Exit Code: 1
2025-02-06T23:28:25.2800453Z 
2025-02-06T23:28:25.2800511Z Command Output (stdout):
2025-02-06T23:28:25.2800555Z --
2025-02-06T23:28:25.2800605Z # RUN: at line 6
2025-02-06T23:28:25.2801048Z /__w/llvm/llvm/toolchain/bin//clang++  -Werror -fsycl -fsycl-targets=spir64  /__w/llvm/llvm/llvm/sycl/test-e2e/bindless_images/examples/example_1_1D_read_write.cpp -o /__w/llvm/llvm/build-e2e/bindless_images/examples/Output/example_1_1D_read_write.cpp.tmp.out
2025-02-06T23:28:25.2801508Z # executed command: /__w/llvm/llvm/toolchain/bin//clang++ -Werror -fsycl -fsycl-targets=spir64 /__w/llvm/llvm/llvm/sycl/test-e2e/bindless_images/examples/example_1_1D_read_write.cpp -o /__w/llvm/llvm/build-e2e/bindless_images/examples/Output/example_1_1D_read_write.cpp.tmp.out
2025-02-06T23:28:25.2801587Z # note: command had no output on stdout or stderr
2025-02-06T23:28:25.2801674Z # RUN: at line 7
2025-02-06T23:28:25.2801890Z env ONEAPI_DEVICE_SELECTOR=level_zero:gpu  /__w/llvm/llvm/build-e2e/bindless_images/examples/Output/example_1_1D_read_write.cpp.tmp.out
2025-02-06T23:28:25.2802131Z # executed command: env ONEAPI_DEVICE_SELECTOR=level_zero:gpu /__w/llvm/llvm/build-e2e/bindless_images/examples/Output/example_1_1D_read_write.cpp.tmp.out
2025-02-06T23:28:25.2802207Z # note: command had no output on stdout or stderr
2025-02-06T23:28:25.2802274Z # error: command failed with exit status: 1
2025-02-06T23:28:25.2802277Z 
2025-02-06T23:28:25.2802326Z --
2025-02-06T23:28:25.2802328Z 

https://github.com/intel/llvm/actions/runs/13188912768/job/36819654925?pr=16910

To reproduce

No response

Environment

No response

Additional context

No response

[DBDuncan]

Swapping the line // RUN: %{run} %t.out with // RUN: %{run} env NEOReadDebugKeys=1 UseBindlessMode=1 UseExternalAllocatorForSshAndDsh=1 %t.out might fix this.

[sarnex]

Manually tried that on the runner and it didn't work. Note that it only seems to fail when tests are run in parallel, if you just run this test alone it seems to pass.

[DBDuncan]

I will be creating an internal ticket to track this and fix. Taking a closer look it seems there is a missing q.wait_and_throw(); between copying data from the device to host and verifying it. Hopefully that will fix the issue.

[sarnex]

Cool, thanks. Glad I don't have to investigate it myself because that sounds like a nightmare to find :)

[DBDuncan]

I am unable to reproduce the failure. I am using a newer driver I believe so that might be why I do not see the failure. Regardless I think the more likely reason for the sporadic failure is to do with the lack of q.wait_and_throw() between the copy from device to host and verifying the results.

Likely what is happening is the device is busy when running all the tests together and does not manage to copy all the data from the device to host before it is verified.

Created #18111 to add the missing wait.