[llvm] 196fe99 - [OpenMP] Fold `__kmpc_is_generic_main_thread_id` if possible
via llvm-commits
llvm-commits at lists.llvm.org
Wed Jul 21 18:18:44 PDT 2021
Author: Joseph Huber
Date: 2021-07-21T21:18:22-04:00
New Revision: 196fe994b8cccd986f4a63530d494cdd04cd04db
URL: https://github.com/llvm/llvm-project/commit/196fe994b8cccd986f4a63530d494cdd04cd04db
DIFF: https://github.com/llvm/llvm-project/commit/196fe994b8cccd986f4a63530d494cdd04cd04db.diff
LOG: [OpenMP] Fold `__kmpc_is_generic_main_thread_id` if possible
This patch adds the ability to fold `__kmpc_is_generic_main_thread_id` if we
know for a fact that it is executed by the initial thread using
AAExecutionDomain. This combined with folding `__kmpc_is_spmd_exec_mode` will
allow us to fully fold `__kmpc_is_generic_main_thread`.
Depends on D106438 D106437
Reviewed By: jdoerfert
Differential Revision: https://reviews.llvm.org/D106439
Added:
llvm/test/Transforms/OpenMP/fold_generic_main_thread.ll
Modified:
llvm/lib/Transforms/IPO/OpenMPOpt.cpp
Removed:
################################################################################
diff --git a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
index d6ca4087eebf..85b11e80c2ed 100644
--- a/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
+++ b/llvm/lib/Transforms/IPO/OpenMPOpt.cpp
@@ -3514,6 +3514,9 @@ struct AAFoldRuntimeCallCallSiteReturned : AAFoldRuntimeCall {
case OMPRTL___kmpc_is_spmd_exec_mode:
Changed |= foldIsSPMDExecMode(A);
break;
+ case OMPRTL___kmpc_is_generic_main_thread_id:
+ Changed |= foldIsGenericMainThread(A);
+ break;
default:
llvm_unreachable("Unhandled OpenMP runtime function!");
}
@@ -3528,6 +3531,10 @@ struct AAFoldRuntimeCallCallSiteReturned : AAFoldRuntimeCall {
Instruction &CB = *getCtxI();
A.changeValueAfterManifest(CB, **SimplifiedValue);
A.deleteAfterManifest(CB);
+
+ LLVM_DEBUG(dbgs() << TAG << "Folding runtime call: " << CB << " with "
+ << **SimplifiedValue << "\n");
+
Changed = ChangeStatus::CHANGED;
}
@@ -3604,6 +3611,28 @@ struct AAFoldRuntimeCallCallSiteReturned : AAFoldRuntimeCall {
: ChangeStatus::CHANGED;
}
+ /// Fold __kmpc_is_generic_main_thread_id into a constant if possible.
+ ChangeStatus foldIsGenericMainThread(Attributor &A) {
+ Optional<Value *> SimplifiedValueBefore = SimplifiedValue;
+
+ CallBase &CB = cast<CallBase>(getAssociatedValue());
+ Function *F = CB.getFunction();
+ const auto &ExecutionDomainAA = A.getAAFor<AAExecutionDomain>(
+ *this, IRPosition::function(*F), DepClassTy::REQUIRED);
+
+ if (!ExecutionDomainAA.isValidState())
+ return indicatePessimisticFixpoint();
+
+ auto &Ctx = getAnchorValue().getContext();
+ if (ExecutionDomainAA.isExecutedByInitialThreadOnly(CB))
+ SimplifiedValue = ConstantInt::get(Type::getInt8Ty(Ctx), true);
+ else
+ return indicatePessimisticFixpoint();
+
+ return SimplifiedValue == SimplifiedValueBefore ? ChangeStatus::UNCHANGED
+ : ChangeStatus::CHANGED;
+ }
+
/// An optional value the associated value is assumed to fold to. That is, we
/// assume the associated value (which is a call) can be replaced by this
/// simplified value.
@@ -3630,6 +3659,19 @@ void OpenMPOpt::registerAAs(bool IsModulePass) {
DepClassTy::NONE, /* ForceUpdate */ false,
/* UpdateAfterInit */ false);
+ auto &IsMainRFI =
+ OMPInfoCache.RFIs[OMPRTL___kmpc_is_generic_main_thread_id];
+ IsMainRFI.foreachUse(SCC, [&](Use &U, Function &F) {
+ CallInst *CI = OpenMPOpt::getCallIfRegularCall(U, &IsMainRFI);
+ if (!CI)
+ return false;
+ A.getOrCreateAAFor<AAFoldRuntimeCall>(
+ IRPosition::callsite_returned(*CI), /* QueryingAA */ nullptr,
+ DepClassTy::NONE, /* ForceUpdate */ false,
+ /* UpdateAfterInit */ false);
+ return false;
+ });
+
auto &IsSPMDRFI = OMPInfoCache.RFIs[OMPRTL___kmpc_is_spmd_exec_mode];
IsSPMDRFI.foreachUse(SCC, [&](Use &U, Function &) {
CallInst *CI = OpenMPOpt::getCallIfRegularCall(U, &IsSPMDRFI);
diff --git a/llvm/test/Transforms/OpenMP/fold_generic_main_thread.ll b/llvm/test/Transforms/OpenMP/fold_generic_main_thread.ll
new file mode 100644
index 000000000000..73df88c791e3
--- /dev/null
+++ b/llvm/test/Transforms/OpenMP/fold_generic_main_thread.ll
@@ -0,0 +1,114 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --function-signature
+; RUN: opt -S -passes='openmp-opt' < %s | FileCheck %s
+; ModuleID = 'single_threaded_exeuction.c'
+
+%struct.ident_t = type { i32, i32, i32, i32, i8* }
+
+ at 0 = private unnamed_addr constant [1 x i8] c"\00", align 1
+ at 1 = private unnamed_addr constant %struct.ident_t { i32 0, i32 2, i32 0, i32 0, i8* getelementptr inbounds ([1 x i8], [1 x i8]* @0, i32 0, i32 0) }, align 8
+
+define void @kernel() {
+; CHECK-LABEL: define {{[^@]+}}@kernel() {
+; CHECK-NEXT: [[CALL:%.*]] = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @[[GLOB1:[0-9]+]], i1 false, i1 false, i1 false)
+; CHECK-NEXT: [[CMP:%.*]] = icmp eq i32 [[CALL]], -1
+; CHECK-NEXT: br i1 [[CMP]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
+; CHECK: if.then:
+; CHECK-NEXT: call void @foo()
+; CHECK-NEXT: br label [[IF_END:%.*]]
+; CHECK: if.else:
+; CHECK-NEXT: call void @bar()
+; CHECK-NEXT: br label [[IF_END]]
+; CHECK: if.end:
+; CHECK-NEXT: call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 true)
+; CHECK-NEXT: ret void
+;
+ %call = call i32 @__kmpc_target_init(%struct.ident_t* nonnull @1, i1 false, i1 false, i1 false)
+ %cmp = icmp eq i32 %call, -1
+ br i1 %cmp, label %if.then, label %if.else
+if.then:
+ call void @foo()
+ br label %if.end
+if.else:
+ call void @bar()
+ br label %if.end
+if.end:
+ call void @__kmpc_target_deinit(%struct.ident_t* null, i1 false, i1 true)
+ ret void
+}
+
+; Function Attrs: noinline
+define internal void @foo() {
+; CHECK-LABEL: define {{[^@]+}}@foo() {
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TID:%.*]] = call i32 @__kmpc_get_hardware_thread_id()
+; CHECK-NEXT: br label [[IF_THEN:%.*]]
+; CHECK: if.then:
+; CHECK-NEXT: call void @baz()
+; CHECK-NEXT: br label [[IF_END:%.*]]
+; CHECK: if.end:
+; CHECK-NEXT: ret void
+;
+entry:
+ %tid = call i32 @__kmpc_get_hardware_thread_id();
+ %ismain = call signext i8 @__kmpc_is_generic_main_thread_id(i32 %tid)
+ %pred = icmp eq i8 %ismain, 1
+ br i1 %pred, label %if.then, label %if.end
+
+if.then:
+ call void @baz()
+ br label %if.end
+
+if.end:
+ ret void
+}
+
+; Function Attrs: noinline
+define internal void @bar() {
+; CHECK-LABEL: define {{[^@]+}}@bar() {
+; CHECK-NEXT: entry:
+; CHECK-NEXT: [[TID:%.*]] = call i32 @__kmpc_get_hardware_thread_id()
+; CHECK-NEXT: [[ISMAIN:%.*]] = call signext i8 @__kmpc_is_generic_main_thread_id(i32 [[TID]])
+; CHECK-NEXT: [[PRED:%.*]] = icmp eq i8 [[ISMAIN]], 1
+; CHECK-NEXT: br i1 [[PRED]], label [[IF_THEN:%.*]], label [[IF_END:%.*]]
+; CHECK: if.then:
+; CHECK-NEXT: call void @baz()
+; CHECK-NEXT: br label [[IF_END]]
+; CHECK: if.end:
+; CHECK-NEXT: ret void
+;
+entry:
+ %tid = call i32 @__kmpc_get_hardware_thread_id();
+ %ismain = call signext i8 @__kmpc_is_generic_main_thread_id(i32 %tid)
+ %pred = icmp eq i8 %ismain, 1
+ br i1 %pred, label %if.then, label %if.end
+
+if.then:
+ call void @baz()
+ br label %if.end
+
+if.end:
+ ret void
+}
+
+declare void @baz()
+
+declare i8 @__kmpc_is_generic_main_thread_id(i32)
+
+declare i32 @__kmpc_get_hardware_thread_id()
+
+declare i32 @__kmpc_target_init(%struct.ident_t*, i1, i1, i1)
+
+declare void @__kmpc_target_deinit(%struct.ident_t*, i1, i1)
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4, !5, !6}
+!nvvm.annotations = !{!7}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 13.0.0", isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, splitDebugInlining: false, nameTableKind: None)
+!1 = !DIFile(filename: "fold_generic_main_thread.c", directory: "/tmp/fold_generic_main_thread.c")
+!2 = !{}
+!3 = !{i32 2, !"Debug Info Version", i32 3}
+!4 = !{i32 1, !"wchar_size", i32 4}
+!5 = !{i32 7, !"openmp", i32 50}
+!6 = !{i32 7, !"openmp-device", i32 50}
+!7 = !{void ()* @kernel, !"kernel", i32 1}
More information about the llvm-commits
mailing list