[clang] [llvm] [mlir] [Flang][OpenMP] Enable no-loop kernels (PR #155818)

@@ -0,0 +1,43 @@ +! Check if the first OpenMP GPU kernel is promoted to no-loop mode. dhruvachak wrote: Please add more test cases showing thread_limit/num_threads clauses and envars OMP_NUM_TEAMS and OMP_TEAMS_THREAD_LIMIT. Additionally, tests with "omp teams"

[clang] [llvm] [AMDGPU][Clang] Allow amdgpu-waves-per-eu attribute to lower target occupancy range (PR #138284)

@@ -216,41 +216,30 @@ define internal i32 @bitcasted_function() { ret i32 0 } -define internal void @called_from_invalid_bounds_0() { -; CHECK-LABEL: define internal void @called_from_invalid_bounds_0 -; CHECK-SAME: () #[[ATTR1]] { -; CHECK-NEXT:ret void -; +define inter

[clang] [LifetimeSafety] Make the dataflow analysis generic (PR #148222)

dhruvachak wrote: @usx95 Build of https://github.com/ROCm/composable_kernel is failing with the same crash. -Weverything is used there as well. Can you please fix the problem or revert? https://github.com/llvm/llvm-project/pull/148222 ___ cfe-commits

[clang] [Clang][AMDGPU] Stop defaulting to `one-as` for all atomic scopes (PR #120095)

@@ -537,7 +537,11 @@ AMDGPUTargetCodeGenInfo::getLLVMSyncScopeID(const LangOptions &LangOpts, break; } - if (Ordering != llvm::AtomicOrdering::SequentiallyConsistent) { + // OpenCL assumes by default that atomic scopes are per-address space for + // non-sequentially

[clang] [Clang][AMDGPU] Stop defaulting to `one-as` for all atomic scopes (PR #120095)

https://github.com/dhruvachak edited https://github.com/llvm/llvm-project/pull/120095 ___ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits

[clang] [Clang][AMDGPU] Stop defaulting to `one-as` for all atomic scopes (PR #120095)

@@ -537,7 +537,11 @@ AMDGPUTargetCodeGenInfo::getLLVMSyncScopeID(const LangOptions &LangOpts, break; } - if (Ordering != llvm::AtomicOrdering::SequentiallyConsistent) { + // OpenCL assumes by default that atomic scopes are per-address space for + // non-sequentially

[clang] [Clang][AMDGPU] Stop defaulting to `one-as` for all atomic scopes (PR #120095)

@@ -537,7 +537,11 @@ AMDGPUTargetCodeGenInfo::getLLVMSyncScopeID(const LangOptions &LangOpts, break; } - if (Ordering != llvm::AtomicOrdering::SequentiallyConsistent) { + // OpenCL assumes by default that atomic scopes are per-address space for + // non-sequentially

[clang] [llvm] [OpenMP][Clang] Migrate OpenMP UserDefinedMapper from Clang to OMPIRBuilder (PR #110001)

@@ -2801,6 +2801,67 @@ class OpenMPIRBuilder { using GenMapInfoCallbackTy = function_ref; +private: + /// Emit the array initialization or deletion portion for user-defined mapper + /// code generation. First, it evaluates whether an array section is mapped + /// an

[clang] [llvm] [OpenMP][Clang] Migrate OpenMP UserDefinedMapper from Clang to OMPIRBuilder (PR #110001)

@@ -2801,6 +2801,67 @@ class OpenMPIRBuilder { using GenMapInfoCallbackTy = function_ref; +private: + /// Emit the array initialization or deletion portion for user-defined mapper + /// code generation. First, it evaluates whether an array section is mapped + /// an