Hi @steffenlarsen, I have another question. The pre-ci for testing ABI compatibility is failed due to this PR will indeed break ABI compatibility. However, the ABI compatibility is unimportant to device sanitizer since the user must re-build their code if they want to test their code with device sanitizer. Do you know if there is a way to disable a test under ABI compatibility testing mode?

Hi @steffenlarsen, I have another question. The pre-ci for testing ABI compatibility is failed due to this PR will indeed break ABI compatibility. However, the ABI compatibility is unimportant to device sanitizer since the user must re-build their code if they want to test their code with device sanitizer. Do you know if there is a way to disable a test under ABI compatibility testing mode?

I believe it should be as simple as adding the tests at the top of devops/compat_ci_exclude.sycl-rel-6_3 with a comment stating why it is OK for it to break compatibility.

Hi @steffenlarsen, I have another question. The pre-ci for testing ABI compatibility is failed due to this PR will indeed break ABI compatibility. However, the ABI compatibility is unimportant to device sanitizer since the user must re-build their code if they want to test their code with device sanitizer. Do you know if there is a way to disable a test under ABI compatibility testing mode?

I'm not sure @xtian-github agrees with that: #19719 (comment)

Hi @xtian-github, could you please help review changes in 'devops/compat_ci_exclude.sycl-rel-6_3'. I disabled sanitizer tests since backward compatibility is unimportant for device sanitizer.

Kindly ping @gmlueck, @xtian-github

LGTM.

A question: BMG is not a supported target? What is the engineering cost to enable it? Can the DG2 path be leveraged?

BMG is not a supported target. We do not validate device-side sanitizer on BMG before so we don't know the gap to support it.
We will give you feedback later.

Hi @intel/llvm-gatekeepers, there is a pre-ci failure with hung test. And the hung issue is unrelated to this change. I add comments to issue #19786 to track this problem. Could you please help merge this PR?

Hi @intel/llvm-gatekeepers, this PR is ready to be merged, thanks.