Skip to content

[llvm][OpenMPOpt] Remove no-op ptr-to-ptr bitcast (NFC) #73869

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Merged
merged 1 commit into from
Nov 30, 2023
Merged

[llvm][OpenMPOpt] Remove no-op ptr-to-ptr bitcast (NFC) #73869

merged 1 commit into from
Nov 30, 2023
+68 −131

Conversation

JOE1994
Copy link
Member

@JOE1994 JOE1994 commented Nov 29, 2023

  • Remove a call to CreatePointerBitCastOrAddrSpaceCast which merely adds a no-op ptr-to-ptr bitcast.

  • Most of the diff is from removing checks for no-op ptr-to-ptr bitcasts in relevant LIT tests

@JOE1994
[llvm][OpenMPOpt] Remove no-op ptr-to-ptr bitcast (NFC)
705ec05 
* Remove a call to CreatePointerBitCastOrAddrSpaceCast which merely adds a
  no-op ptr-to-ptr bitcast.

* Most of the diff is from removing checks for no-op ptr-to-ptr bitcasts in
  relevant LIT tests
@JOE1994 JOE1994 requested review from nikic and jdoerfert November 29, 2023 23:28
@llvmbot llvmbot added llvm:transforms clang:openmp OpenMP related changes to Clang labels Nov 29, 2023
@llvmbot
Copy link
Member

llvmbot commented Nov 29, 2023

@llvm/pr-subscribers-llvm-transforms

Author: Youngsuk Kim (JOE1994)

Changes

  • Remove a call to CreatePointerBitCastOrAddrSpaceCast which merely adds a no-op ptr-to-ptr bitcast.

  • Most of the diff is from removing checks for no-op ptr-to-ptr bitcasts in relevant LIT tests

Patch is 100.14 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/73869.diff

7 Files Affected:

  • (modified) llvm/lib/Transforms/IPO/OpenMPOpt.cpp (+2-5)
  • (modified) llvm/test/Transforms/OpenMP/custom_state_machines.ll (+18-30)
  • (modified) llvm/test/Transforms/OpenMP/gpu_state_machine_function_ptr_replacement.ll (+2-2)
  • (modified) llvm/test/Transforms/OpenMP/spmdization.ll (+40-80)
  • (modified) llvm/test/Transforms/OpenMP/spmdization_guarding.ll (-1)
  • (modified) llvm/test/Transforms/OpenMP/spmdization_indirect.ll (+4-8)
  • (modified) llvm/test/Transforms/OpenMP/spmdization_no_guarding_two_reaching_kernels.ll (+2-5)
diff --git a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
index 5b42f215fb40ca0..b2665161c090dfb 100644
--- a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
+++ b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
@@ -4527,9 +4527,6 @@ struct AAKernelInfoFunction : AAKernelInfo {
     FunctionType *ParallelRegionFnTy = FunctionType::get(
         Type::getVoidTy(Ctx), {Type::getInt16Ty(Ctx), Type::getInt32Ty(Ctx)},
         false);
-    Value *WorkFnCast = BitCastInst::CreatePointerBitCastOrAddrSpaceCast(
-        WorkFn, ParallelRegionFnTy->getPointerTo(), "worker.work_fn.addr_cast",
-        StateMachineBeginBB);
 
     Instruction *IsDone =
         ICmpInst::Create(ICmpInst::ICmp, llvm::CmpInst::ICMP_EQ, WorkFn,
@@ -4576,7 +4573,7 @@ struct AAKernelInfoFunction : AAKernelInfo {
       Value *IsPR;
       if (I + 1 < E || !ReachedUnknownParallelRegions.empty()) {
         Instruction *CmpI = ICmpInst::Create(
-            ICmpInst::ICmp, llvm::CmpInst::ICMP_EQ, WorkFnCast, ParallelRegion,
+            ICmpInst::ICmp, llvm::CmpInst::ICMP_EQ, WorkFn, ParallelRegion,
             "worker.check_parallel_region", StateMachineIfCascadeCurrentBB);
         CmpI->setDebugLoc(DLoc);
         IsPR = CmpI;
@@ -4596,7 +4593,7 @@ struct AAKernelInfoFunction : AAKernelInfo {
     if (!ReachedUnknownParallelRegions.empty()) {
       StateMachineIfCascadeCurrentBB->setName(
           "worker_state_machine.parallel_region.fallback.execute");
-      CallInst::Create(ParallelRegionFnTy, WorkFnCast, {ZeroArg, GTid}, "",
+      CallInst::Create(ParallelRegionFnTy, WorkFn, {ZeroArg, GTid}, "",
                        StateMachineIfCascadeCurrentBB)
           ->setDebugLoc(DLoc);
     }
diff --git a/llvm/test/Transforms/OpenMP/custom_state_machines.ll b/llvm/test/Transforms/OpenMP/custom_state_machines.ll
index 1a6624b46f10dc7..b0d1842ca2e8ba4 100644
--- a/llvm/test/Transforms/OpenMP/custom_state_machines.ll
+++ b/llvm/test/Transforms/OpenMP/custom_state_machines.ll
@@ -998,7 +998,6 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]])
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
 ; AMDGPU:       worker_state_machine.finished:
@@ -1006,7 +1005,7 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU:       worker_state_machine.is_active.check:
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.check:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__2_wrapper.ID
+; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__2_wrapper.ID
 ; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.execute:
 ; AMDGPU-NEXT:    call void @__omp_outlined__2_wrapper(i16 0, i32 [[TMP0]])
@@ -1117,7 +1116,6 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]])
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
 ; AMDGPU:       worker_state_machine.finished:
@@ -1125,13 +1123,13 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU:       worker_state_machine.is_active.check:
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.check:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__17_wrapper
+; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__17_wrapper
 ; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.execute:
 ; AMDGPU-NEXT:    call void @__omp_outlined__17_wrapper(i16 0, i32 [[TMP0]])
 ; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.check1:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__5_wrapper.ID
+; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__5_wrapper.ID
 ; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION4]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK3:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.execute2:
 ; AMDGPU-NEXT:    call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]])
@@ -1259,7 +1257,6 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]])
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
 ; AMDGPU:       worker_state_machine.finished:
@@ -1267,19 +1264,19 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU:       worker_state_machine.is_active.check:
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.check:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__7_wrapper.ID
+; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__7_wrapper.ID
 ; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.execute:
 ; AMDGPU-NEXT:    call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]])
 ; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.check1:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__8_wrapper.ID
+; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__8_wrapper.ID
 ; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION4]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.execute2:
 ; AMDGPU-NEXT:    call void @__omp_outlined__8_wrapper(i16 0, i32 [[TMP0]])
 ; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
 ; AMDGPU:       worker_state_machine.parallel_region.fallback.execute:
-; AMDGPU-NEXT:    call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
+; AMDGPU-NEXT:    call void [[WORKER_WORK_FN]](i16 0, i32 [[TMP0]])
 ; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
 ; AMDGPU:       worker_state_machine.parallel_region.end:
 ; AMDGPU-NEXT:    call void @__kmpc_kernel_end_parallel()
@@ -1379,7 +1376,6 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]])
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
 ; AMDGPU:       worker_state_machine.finished:
@@ -1387,7 +1383,7 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU:       worker_state_machine.is_active.check:
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.check:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__10_wrapper.ID
+; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__10_wrapper.ID
 ; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.execute:
 ; AMDGPU-NEXT:    call void @__omp_outlined__10_wrapper(i16 0, i32 [[TMP0]])
@@ -1497,7 +1493,6 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]])
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
 ; AMDGPU:       worker_state_machine.finished:
@@ -1505,7 +1500,7 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU:       worker_state_machine.is_active.check:
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.check:
-; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__13_wrapper.ID
+; AMDGPU-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__13_wrapper.ID
 ; AMDGPU-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.execute:
 ; AMDGPU-NEXT:    call void @__omp_outlined__13_wrapper(i16 0, i32 [[TMP0]])
@@ -1686,7 +1681,6 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_GENERIC:%.*]] = addrspacecast ptr addrspace(5) [[WORKER_WORK_FN_ADDR]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR_GENERIC]])
 ; AMDGPU-NEXT:    [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR_GENERIC]], align 8
-; AMDGPU-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr
 ; AMDGPU-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
 ; AMDGPU:       worker_state_machine.finished:
@@ -1694,7 +1688,7 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; AMDGPU:       worker_state_machine.is_active.check:
 ; AMDGPU-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.fallback.execute:
-; AMDGPU-NEXT:    call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
+; AMDGPU-NEXT:    call void [[WORKER_WORK_FN]](i16 0, i32 [[TMP0]])
 ; AMDGPU-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
 ; AMDGPU:       worker_state_machine.parallel_region.end:
 ; AMDGPU-NEXT:    call void @__kmpc_kernel_end_parallel()
@@ -1908,7 +1902,6 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; NVPTX-NEXT:    call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
 ; NVPTX-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]])
 ; NVPTX-NEXT:    [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr
 ; NVPTX-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null
 ; NVPTX-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
 ; NVPTX:       worker_state_machine.finished:
@@ -1916,7 +1909,7 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; NVPTX:       worker_state_machine.is_active.check:
 ; NVPTX-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
 ; NVPTX:       worker_state_machine.parallel_region.check:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__2_wrapper.ID
+; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__2_wrapper.ID
 ; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
 ; NVPTX:       worker_state_machine.parallel_region.execute:
 ; NVPTX-NEXT:    call void @__omp_outlined__2_wrapper(i16 0, i32 [[TMP0]])
@@ -2026,7 +2019,6 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; NVPTX-NEXT:    call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
 ; NVPTX-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]])
 ; NVPTX-NEXT:    [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr
 ; NVPTX-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null
 ; NVPTX-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
 ; NVPTX:       worker_state_machine.finished:
@@ -2034,13 +2026,13 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; NVPTX:       worker_state_machine.is_active.check:
 ; NVPTX-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
 ; NVPTX:       worker_state_machine.parallel_region.check:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__17_wrapper
+; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__17_wrapper
 ; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
 ; NVPTX:       worker_state_machine.parallel_region.execute:
 ; NVPTX-NEXT:    call void @__omp_outlined__17_wrapper(i16 0, i32 [[TMP0]])
 ; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
 ; NVPTX:       worker_state_machine.parallel_region.check1:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__5_wrapper.ID
+; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__5_wrapper.ID
 ; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION4]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK3:%.*]]
 ; NVPTX:       worker_state_machine.parallel_region.execute2:
 ; NVPTX-NEXT:    call void @__omp_outlined__5_wrapper(i16 0, i32 [[TMP0]])
@@ -2167,7 +2159,6 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; NVPTX-NEXT:    call void @__kmpc_barrier_simple_generic(ptr @[[GLOB1]], i32 [[TMP0]])
 ; NVPTX-NEXT:    [[WORKER_IS_ACTIVE:%.*]] = call i1 @__kmpc_kernel_parallel(ptr [[WORKER_WORK_FN_ADDR]])
 ; NVPTX-NEXT:    [[WORKER_WORK_FN:%.*]] = load ptr, ptr [[WORKER_WORK_FN_ADDR]], align 8
-; NVPTX-NEXT:    [[WORKER_WORK_FN_ADDR_CAST:%.*]] = bitcast ptr [[WORKER_WORK_FN]] to ptr
 ; NVPTX-NEXT:    [[WORKER_IS_DONE:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], null
 ; NVPTX-NEXT:    br i1 [[WORKER_IS_DONE]], label [[WORKER_STATE_MACHINE_FINISHED]], label [[WORKER_STATE_MACHINE_IS_ACTIVE_CHECK:%.*]]
 ; NVPTX:       worker_state_machine.finished:
@@ -2175,19 +2166,19 @@ attributes #9 = { convergent nounwind readonly willreturn }
 ; NVPTX:       worker_state_machine.is_active.check:
 ; NVPTX-NEXT:    br i1 [[WORKER_IS_ACTIVE]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK:%.*]], label [[WORKER_STATE_MACHINE_DONE_BARRIER:%.*]]
 ; NVPTX:       worker_state_machine.parallel_region.check:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__7_wrapper.ID
+; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__7_wrapper.ID
 ; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_CHECK1:%.*]]
 ; NVPTX:       worker_state_machine.parallel_region.execute:
 ; NVPTX-NEXT:    call void @__omp_outlined__7_wrapper(i16 0, i32 [[TMP0]])
 ; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END:%.*]]
 ; NVPTX:       worker_state_machine.parallel_region.check1:
-; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq ptr [[WORKER_WORK_FN_ADDR_CAST]], @__omp_outlined__8_wrapper.ID
+; NVPTX-NEXT:    [[WORKER_CHECK_PARALLEL_REGION4:%.*]] = icmp eq ptr [[WORKER_WORK_FN]], @__omp_outlined__8_wrapper.ID
 ; NVPTX-NEXT:    br i1 [[WORKER_CHECK_PARALLEL_REGION4]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_EXECUTE2:%.*]], label [[WORKER_STATE_MACHINE_PARALLEL_REGION_FALLBACK_EXECUTE:%.*]]
 ; NVPTX:       worker_state_machine.parallel_region.execute2:
 ; NVPTX-NEXT:    call void @__omp_outlined__8_wrapper(i16 0, i32 [[TMP0]])
 ; NVPTX-NEXT:    br label [[WORKER_STATE_MACHINE_PARALLEL_REGION_END]]
 ; NVPTX:       worker_state_machine.parallel_region.fallback.execute:
-; NVPTX-NEXT:    call void [[WORKER_WORK_FN_ADDR_CAST]](i16 0, i32 [[TMP0]])
+; NVPTX-NEXT:    call void [[WORKER_WORK_FN]](i16 0, i32 [[TMP0]])
 ; NVPTX-NEXT:    br...
[truncated]

@JOE1994 JOE1994 merged commit c57ef2c into llvm:main Nov 30, 2023
@JOE1994 JOE1994 deleted the openmpopt_remove_ptrtoptrbitcast branch November 30, 2023 01:47
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@aeubanks aeubanks aeubanks approved these changes

@nikic nikic Awaiting requested review from nikic

@jdoerfert jdoerfert Awaiting requested review from jdoerfert

Assignees
No one assigned
Labels
clang:openmp OpenMP related changes to Clang llvm:transforms
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
3 participants
@JOE1994 @llvmbot @aeubanks