[SYCL] fix for __sycl_unregister_lib() on Windows and tests #19633
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Signed-off-by: Chris Perkins <[email protected]>
cperkinsintel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
cperkinsintel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
cperkinsintel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
Signed-off-by: Chris Perkins <[email protected]>
cperkinsintel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
cperkinsintel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
cperkinsintel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
Signed-off-by: Chris Perkins <[email protected]>
cperkinsintel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
cperkinsintel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
cperkinsintel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
Signed-off-by: Chris Perkins <[email protected]>
cperkinsintel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
cperkinsintel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
cperkinsintel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
|
PVC failures is unrelated: #19662 |
Signed-off-by: Chris Perkins <[email protected]>
cperkinsintel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
cperkinsintel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
cperkinsintel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
cperkinsintel
had a problem deploying
to
WindowsCILock
GitHub Actions
Failure
cperkinsintel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
cperkinsintel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
Signed-off-by: Chris Perkins <[email protected]>
cperkinsintel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
cperkinsintel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
cperkinsintel
temporarily deployed
to
WindowsCILock
GitHub Actions
Inactive
againull
reviewed
Aug 6, 2025
| // DEFINE: %{shared_lib_ext} = %if windows %{dll%} %else %{so%} | ||
|
|
||
| // clang-format off | ||
| // IMPORTANT -DSO_PATH='R"(%T)"' WTF ?? |
There was a problem hiding this comment.
Suggested change
| // IMPORTANT -DSO_PATH='R"(%T)"' WTF ?? | |
| // IMPORTANT -DSO_PATH='R"(%T)"' |
| // Because on Windows, the path delimiters are \, | ||
| // which C++ preprocessor converts to escape sequences, | ||
| // which becomes a nightmare. | ||
| // SO the hack here is to put heredoc in the definition |
There was a problem hiding this comment.
Suggested change
| // SO the hack here is to put heredoc in the definition | |
| // So the hack here is to put heredoc in the definition |
Comment on lines
+1
to
+4
| // UNSUPPORTED: cuda || hip | ||
| // UNSUPPORTED-TRACKER: CMPLRLLVM-69415 | ||
|
|
||
| // REQUIRES: level_zero |
There was a problem hiding this comment.
Should we remove "REQUIRES: level_zero"? It looks like test is supposed to work on OpenCL backend as well.
If you want to additionally check leaks on level_zero backend, then probably something like this should be done:
// RUN: %if level_zero %{%{l0_leak_check} %{run} %t.out 2>&1 | FileCheck %s --implicit-check-not=LEAK %}
| RegBuilder.CreateRetVoid(); | ||
|
|
||
| // Add this function to global destructors. | ||
| // Match priority of __tgt_register_lib |
There was a problem hiding this comment.
Suggested change
| // Match priority of __tgt_register_lib | |
| // Match priority of __sycl_register_lib |
Comment on lines
-306
to
-315
| { | ||
| sycl::detail::DeviceGlobalMap &DeviceGlobalMap = PM.getDeviceGlobals(); | ||
| EXPECT_EQ(DeviceGlobalMap.size(), ExpectedCount) << Comment; | ||
| EXPECT_TRUE(DeviceGlobalMap.count(generateRefName("A", "DeviceGlobal")) > 0) | ||
| << Comment; | ||
| EXPECT_TRUE(DeviceGlobalMap.count(generateRefName("B", "DeviceGlobal")) > 0) | ||
| << Comment; | ||
| EXPECT_EQ(DeviceGlobalMap.getPointerMap().size(), ExpectedCount) << Comment; | ||
| } | ||
|
|
There was a problem hiding this comment.
Is it possible to add a test for new behavior to compensate the removed test?
For my own education, could you please point me to some info regarding this limitation. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
__sycl_unregister_lib()is not being called on Windows when using shared libraries due to a limitation in LLVM/clang To work around this, on Windows we register both__sycl_register_lib()and anstd::atexithandler that will call__sycl_unregister_lib(). Further, it was discovered that freeing of the device globals during device images destruction is duplicate and unnecessary. The~context_impldestructor handles that (and handles it correctly, because a context is needed to free USM memory). So we remove the unneeded duplication. Adding a test that stresses__sycl_unregister_lib()and makes sure there are no resource leaks