[llvm-branch-commits] [openmp] 0d4e55b - [OpenMP] Recover non-blocking target nowait disabled by D141232
Author: Ye Luo
Date: 2023-02-14T15:48:38-06:00
New Revision: 0d4e55ba69c1dd04913c59d39c764c8070e55c60
URL:
https://github.com/llvm/llvm-project/commit/0d4e55ba69c1dd04913c59d39c764c8070e55c60
DIFF:
https://github.com/llvm/llvm-project/commit/0d4e55ba69c1dd04913c59d39c764c8070e55c60.diff
LOG: [OpenMP] Recover non-blocking target nowait disabled by D141232
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D143871
Added:
Modified:
openmp/libomptarget/src/interface.cpp
Removed:
diff --git a/openmp/libomptarget/src/interface.cpp
b/openmp/libomptarget/src/interface.cpp
index 91dda62b0c2d5..beea0c26210bc 100644
--- a/openmp/libomptarget/src/interface.cpp
+++ b/openmp/libomptarget/src/interface.cpp
@@ -299,8 +299,12 @@ EXTERN int __tgt_target_kernel(ident_t *Loc, int64_t
DeviceId, int32_t NumTeams,
int32_t ThreadLimit, void *HostPtr,
KernelArgsTy *KernelArgs) {
TIMESCOPE_WITH_IDENT(Loc);
- return targetKernel(Loc, DeviceId, NumTeams, ThreadLimit,
- HostPtr, KernelArgs);
+ if (KernelArgs->Flags.NoWait)
+return targetKernel(
+Loc, DeviceId, NumTeams, ThreadLimit, HostPtr, KernelArgs);
+ else
+return targetKernel(Loc, DeviceId, NumTeams, ThreadLimit,
+ HostPtr, KernelArgs);
}
/// Implements a target kernel entry that replays a pre-recorded kernel.
___
llvm-branch-commits mailing list
llvm-branch-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [openmp] release/19.x: [OpenMP] Create versioned libgomp softlinks (#112973) (PR #115944)
ye-luo wrote: > Are we still releasing 19? yes. Point releases 19.1.x https://github.com/llvm/llvm-project/pull/115944 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang] [Clang] Improve type traits recognition in `__has_builtin` (#111516) (PR #111660)
ye-luo wrote: > Just to double-check assumptions -- we don't know of any builtins which Clang > now reports true for which Clang does not actually support? And we don't know > of any builtins which Clang now reports false for which Clang does actually > support? IOW, `__has_builtin` is not lying after landing these changes, right? I have little knowledge about builtin. At least `__has_builtin` is not lying about `__reference_converts_from_temporary`. So, probably I will just live with this change in the point release. https://github.com/llvm/llvm-project/pull/111660 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang] [Clang] Improve type traits recognition in `__has_builtin` (#111516) (PR #111660)
ye-luo wrote: This change breaks use of llvm 19.1.2 and 19.1.3 https://github.com/llvm/llvm-project/issues/113125 Can it be reverted on the release branch? https://github.com/llvm/llvm-project/pull/111660 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [clang] [Clang] Improve type traits recognition in `__has_builtin` (#111516) (PR #111660)
ye-luo wrote: I tried __reference_converts_from_temporary https://godbolt.org/z/TPjbqWT67 18.1.0, 19.1.0 both return false 19.1.2 and main returns true. I have no issue with __has_builtin fixed in the main but the resulted behavior change in a point release is a bit trouble some. https://github.com/llvm/llvm-project/pull/111660 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [openmp] release/19.x: [OpenMP] Create versioned libgomp softlinks (#112973) (PR #115944)
ye-luo wrote: > Hmm. Is this really a regression? Which is what we usually only fix at this > point. It seems like a "feature". What's the risk of this creating problems > for users already adopted 19.x? It is not a regression but a bug fix. The issue has been there for years. NVHPC ships the similar softlinks to their OpenMP runtime for years. It should be low risk but I’ll let you decide to apply the change to 19.x or not. https://github.com/llvm/llvm-project/pull/115944 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
[llvm-branch-commits] [llvm] release/20.x: [OpenMP] Add pre sm_70 load hack back in (#138589) (PR #138626)
ye-luo wrote: I manually verified the effectiveness of this patch on 20.x release branch. https://github.com/llvm/llvm-project/pull/138626 ___ llvm-branch-commits mailing list llvm-branch-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-branch-commits
