[llvm] ae10b8a - [Attributor][FIX] Give registered simplification callbacks precedence
Johannes Doerfert via llvm-commits
llvm-commits at lists.llvm.org
Thu Jun 9 06:32:34 PDT 2022
Author: Johannes Doerfert
Date: 2022-06-09T15:31:53+02:00
New Revision: ae10b8a58240598c552023437783e71e2d7ef192
URL: https://github.com/llvm/llvm-project/commit/ae10b8a58240598c552023437783e71e2d7ef192
DIFF: https://github.com/llvm/llvm-project/commit/ae10b8a58240598c552023437783e71e2d7ef192.diff
LOG: [Attributor][FIX] Give registered simplification callbacks precedence
We accidentally checked for constants before we looked for registered
simplification callbacks. The latter needs to take precedence though.
Added:
llvm/test/Transforms/OpenMP/spmdization_constant_prop.ll
Modified:
llvm/lib/Transforms/IPO/Attributor.cpp
Removed:
################################################################################
diff --git a/llvm/lib/Transforms/IPO/Attributor.cpp b/llvm/lib/Transforms/IPO/Attributor.cpp
index 0d4cdd7294fd..0a74eab074ff 100644
--- a/llvm/lib/Transforms/IPO/Attributor.cpp
+++ b/llvm/lib/Transforms/IPO/Attributor.cpp
@@ -1025,8 +1025,6 @@ Optional<Constant *>
Attributor::getAssumedConstant(const IRPosition &IRP,
const AbstractAttribute &AA,
bool &UsedAssumedInformation) {
- if (auto *C = dyn_cast<Constant>(&IRP.getAssociatedValue()))
- return C;
// First check all callbacks provided by outside AAs. If any of them returns
// a non-null value that is
diff erent from the associated value, or None, we
// assume it's simplified.
@@ -1038,6 +1036,8 @@ Attributor::getAssumedConstant(const IRPosition &IRP,
return cast<Constant>(*SimplifiedV);
return nullptr;
}
+ if (auto *C = dyn_cast<Constant>(&IRP.getAssociatedValue()))
+ return C;
const auto &ValueSimplifyAA =
getAAFor<AAValueSimplify>(AA, IRP, DepClassTy::NONE);
Optional<Value *> SimplifiedV =
diff --git a/llvm/test/Transforms/OpenMP/spmdization_constant_prop.ll b/llvm/test/Transforms/OpenMP/spmdization_constant_prop.ll
new file mode 100644
index 000000000000..98d852d2e3ba
--- /dev/null
+++ b/llvm/test/Transforms/OpenMP/spmdization_constant_prop.ll
@@ -0,0 +1,326 @@
+; RUN: opt -S -passes=openmp-opt < %s | FileCheck %s
+;
+; Verify we change it to SPMD mode but also avoid propagating the old mode (=generic) into the __kmpc_target_init function.
+;
+; CHECK-NOT: store i32 0, ptr addrspace(3) @IsSPMDMode
+; CHECK: call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @1 to ptr), i8 2, i1 false, i1 false)
+; CHECK-NOT: store i32 0, ptr addrspace(3) @IsSPMDMode
+; CHECK: store i32 1, ptr addrspace(3) @IsSPMDMode
+; CHECK-NOT: store i32 0, ptr addrspace(3) @IsSPMDMode
+;
+target datalayout = "e-p:64:64-p1:64:64-p2:32:32-p3:32:32-p4:64:64-p5:32:32-p6:32:32-i64:64-v16:16-v24:32-v32:32-v48:64-v96:128-v192:256-v256:256-v512:512-v1024:1024-v2048:2048-n32:64-S32-A5-G1-ni:7"
+target triple = "amdgcn-amd-amdhsa"
+
+%struct.ident_t = type { i32, i32, i32, i32, ptr }
+%struct.DeviceEnvironmentTy = type { i32, i32, i32, i32 }
+%"struct.(anonymous namespace)::SharedMemorySmartStackTy" = type { [512 x i8], [1024 x i8] }
+%"struct.(anonymous namespace)::TeamStateTy" = type { %"struct.(anonymous namespace)::ICVStateTy", i32, ptr }
+%"struct.(anonymous namespace)::ICVStateTy" = type { i32, i32, i32, i32, i32, i32 }
+
+ at __omp_rtl_assume_teams_oversubscription = weak_odr hidden local_unnamed_addr addrspace(1) constant i32 0
+ at __omp_rtl_assume_threads_oversubscription = weak_odr hidden local_unnamed_addr addrspace(1) constant i32 0
+ at 0 = private unnamed_addr constant [23 x i8] c";unknown;unknown;0;0;;\00", align 1
+ at 1 = private unnamed_addr addrspace(1) constant %struct.ident_t { i32 0, i32 2, i32 0, i32 22, ptr @0 }, align 8
+ at __omp_offloading_20_11e3950_main_l12_exec_mode = weak addrspace(1) constant i8 1
+ at __omp_rtl_debug_kind = weak_odr hidden local_unnamed_addr addrspace(1) constant i32 0
+ at __omp_rtl_assume_no_thread_state = weak_odr hidden local_unnamed_addr addrspace(1) constant i32 0
+ at omptarget_device_environment = weak protected addrspace(4) global %struct.DeviceEnvironmentTy undef, align 4
+ at IsSPMDMode = weak hidden addrspace(3) global i32 undef, align 4
+ at .str.12 = private unnamed_addr addrspace(4) constant [47 x i8] c"ValueRAII initialization with wrong old value!\00", align 1
+ at _ZN12_GLOBAL__N_122SharedMemorySmartStackE = internal addrspace(3) global %"struct.(anonymous namespace)::SharedMemorySmartStackTy" undef, align 16
+ at _ZN12_GLOBAL__N_19TeamStateE = internal unnamed_addr addrspace(3) global %"struct.(anonymous namespace)::TeamStateTy" undef, align 8
+ at _ZN12_GLOBAL__N_112ThreadStatesE = internal unnamed_addr addrspace(3) global [1024 x ptr] undef, align 16
+ at .str.12.47 = private unnamed_addr addrspace(4) constant [49 x i8] c"Thread state modified while explicitly disabled!\00", align 1
+ at _ZL29SharedMemVariableSharingSpace = internal unnamed_addr addrspace(3) global [64 x ptr] undef, align 16
+ at G = global i32 undef
+ at llvm.used = appending addrspace(1) global [2 x ptr] [ptr addrspacecast (ptr addrspace(3) @IsSPMDMode to ptr), ptr addrspacecast (ptr addrspace(4) @omptarget_device_environment to ptr)], section "llvm.metadata"
+ at llvm.compiler.used = appending addrspace(1) global [1 x ptr] [ptr addrspacecast (ptr addrspace(1) @__omp_offloading_20_11e3950_main_l12_exec_mode to ptr)], section "llvm.metadata"
+
+; Function Attrs: alwaysinline convergent norecurse nounwind
+define weak_odr amdgpu_kernel void @__omp_offloading_20_11e3950_main_l12(i64 noundef %nxyz, i64 noundef %ng, ptr noundef nonnull align 8 dereferenceable(8) %aa) local_unnamed_addr #0 {
+entry:
+ %ng1 = alloca i32, align 4
+ %captured_vars_addrs = alloca [2 x ptr], align 8, addrspace(5)
+ %0 = call i32 @__kmpc_target_init(ptr addrspacecast (ptr addrspace(1) @1 to ptr), i8 1, i1 true, i1 true)
+ %exec_user_code = icmp eq i32 %0, -1
+ br i1 %exec_user_code, label %user_code.entry, label %common.ret
+
+user_code.entry: ; preds = %entry
+ %captured_vars_addrs.ascast = addrspacecast ptr addrspace(5) %captured_vars_addrs to ptr
+ store ptr %ng1, ptr addrspace(5) %captured_vars_addrs, align 8, !tbaa !7
+ call void @__kmpc_parallel_51(ptr addrspacecast (ptr addrspace(1) @1 to ptr), i32 0, i32 1, i32 -1, i32 -1, ptr nonnull @__omp_outlined__, ptr nonnull @__omp_outlined___wrapper, ptr nonnull %captured_vars_addrs.ascast, i64 2)
+ call void @__kmpc_target_deinit(ptr addrspacecast (ptr addrspace(1) @1 to ptr), i8 1, i1 true)
+ br label %common.ret
+
+common.ret: ; preds = %user_code.entry, %entry
+ ret void
+}
+
+; Function Attrs: argmemonly nocallback nofree nosync nounwind willreturn
+declare void @llvm.lifetime.start.p5(i64 immarg, ptr addrspace(5) nocapture) #1
+
+; Function Attrs: alwaysinline mustprogress nofree norecurse nosync nounwind readnone willreturn
+define internal void @__omp_outlined__(ptr noalias nocapture %.global_tid., ptr noalias nocapture %.bound_tid., ptr nocapture nonnull align 4 %ng, ptr nocapture nonnull align 8 %aa) #2 {
+entry:
+ %isspmd = load i32, ptr addrspace(3) @IsSPMDMode
+ store i32 %isspmd, ptr @G
+ ret void
+}
+
+; Function Attrs: argmemonly nocallback nofree nosync nounwind willreturn
+declare void @llvm.lifetime.end.p5(i64 immarg, ptr addrspace(5) nocapture) #1
+
+; Function Attrs: mustprogress nofree norecurse nosync nounwind readnone willreturn
+define internal void @__omp_outlined___wrapper(i16 zeroext %0, i32 noundef %1) #3 {
+entry:
+ %isspmd = load i32, ptr addrspace(3) @IsSPMDMode
+ store i32 %isspmd, ptr @G
+ ret void
+}
+
+; Function Attrs: nounwind readnone speculatable willreturn
+declare i32 @llvm.amdgcn.workitem.id.x() #4
+
+; Function Attrs: nounwind readnone speculatable willreturn
+declare i32 @llvm.amdgcn.workgroup.id.x() #4
+
+; Function Attrs: nounwind readnone speculatable willreturn
+declare align 4 ptr addrspace(4) @llvm.amdgcn.dispatch.ptr() #4
+
+; Function Attrs: nocallback nofree nosync nounwind readnone speculatable willreturn
+declare i32 @llvm.umin.i32(i32, i32) #5
+
+; Function Attrs: inaccessiblememonly nocallback nofree nosync nounwind willreturn
+declare void @llvm.assume(i1 noundef) #6
+
+; Function Attrs: convergent nounwind willreturn
+declare void @llvm.amdgcn.s.barrier() #7
+
+; Function Attrs: convergent mustprogress noinline nounwind willreturn
+define internal fastcc void @_ZN4_OMP11synchronize14threadsAlignedEv() unnamed_addr #8 {
+entry:
+ call void @llvm.amdgcn.s.barrier() #13
+ ret void
+}
+
+; Function Attrs: convergent nounwind
+define internal i32 @__kmpc_target_init(ptr nocapture noundef readnone %Ident, i8 noundef signext %Mode, i1 noundef zeroext %UseGenericStateMachine, i1 noundef zeroext %0) local_unnamed_addr #9 {
+entry:
+ %1 = and i8 %Mode, 2
+ %tobool.not = icmp eq i8 %1, 0
+ br i1 %tobool.not, label %if.else, label %if.then
+
+if.then: ; preds = %entry
+ %2 = call i32 @llvm.amdgcn.workitem.id.x() #13, !range !11
+ %3 = call i32 @llvm.amdgcn.workgroup.id.x() #13
+ %4 = call align 4 dereferenceable(64) ptr addrspace(4) @llvm.amdgcn.dispatch.ptr() #13
+ %5 = getelementptr i8, ptr addrspace(4) %4, i64 12
+ %6 = load i32, ptr addrspace(4) %5, align 4, !invariant.load !12
+ %7 = getelementptr i8, ptr addrspace(4) %4, i64 4
+ %8 = load i16, ptr addrspace(4) %7, align 4, !range !13, !invariant.load !12
+ %conv.i.i7.i.i.i = zext i16 %8 to i32
+ %mul.i.i8.i.i.i = mul i32 %3, %conv.i.i7.i.i.i
+ %sub.i.i9.i.i.i = sub i32 %6, %mul.i.i8.i.i.i
+ %9 = call i32 @llvm.umin.i32(i32 %sub.i.i9.i.i.i, i32 %conv.i.i7.i.i.i) #13
+ %cmp4.i.i.i = icmp ult i32 %2, %9
+ call void @llvm.assume(i1 %cmp4.i.i.i) #13
+ %cmp.i.i8 = icmp eq i32 %2, 0
+ br i1 %cmp.i.i8, label %if.then.i, label %_ZN4_OMP5state4initEb.exit.critedge
+
+if.then.i: ; preds = %if.then
+ store i32 1, ptr addrspace(3) @IsSPMDMode, align 4, !tbaa !14
+ store i8 0, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::SharedMemorySmartStackTy", ptr addrspace(3) @_ZN12_GLOBAL__N_122SharedMemorySmartStackE, i32 0, i32 1, i32 0), align 16, !tbaa !18
+ store i32 %9, ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, align 8, !tbaa !19
+ store i32 0, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 1), align 4, !tbaa !23
+ store i32 0, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 2), align 8, !tbaa !24
+ store i32 1, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 3), align 4, !tbaa !25
+ store i32 1, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 4), align 8, !tbaa !26
+ store i32 1, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 5), align 4, !tbaa !27
+ store i32 1, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 1), align 8, !tbaa !28
+ store ptr null, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 2), align 8, !tbaa !29
+ br label %_ZN4_OMP5state4initEb.exit
+
+_ZN4_OMP5state4initEb.exit.critedge: ; preds = %if.then
+ %arrayidx.i.i.c = getelementptr inbounds [1024 x i8], ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::SharedMemorySmartStackTy", ptr addrspace(3) @_ZN12_GLOBAL__N_122SharedMemorySmartStackE, i32 0, i32 1, i32 0), i32 0, i32 %2
+ store i8 0, ptr addrspace(3) %arrayidx.i.i.c, align 1, !tbaa !18
+ br label %_ZN4_OMP5state4initEb.exit
+
+_ZN4_OMP5state4initEb.exit: ; preds = %_ZN4_OMP5state4initEb.exit.critedge, %if.then.i
+ %arrayidx.i = getelementptr inbounds [1024 x ptr], ptr addrspace(3) @_ZN12_GLOBAL__N_112ThreadStatesE, i32 0, i32 %2
+ store ptr null, ptr addrspace(3) %arrayidx.i, align 8, !tbaa !30
+ call fastcc void @_ZN4_OMP11synchronize14threadsAlignedEv() #14
+ br label %if.end
+
+if.else: ; preds = %entry
+ %10 = call i32 @llvm.amdgcn.workgroup.id.x() #13
+ %11 = call align 4 dereferenceable(64) ptr addrspace(4) @llvm.amdgcn.dispatch.ptr() #13
+ %12 = getelementptr i8, ptr addrspace(4) %11, i64 12
+ %13 = load i32, ptr addrspace(4) %12, align 4, !invariant.load !12
+ %14 = getelementptr i8, ptr addrspace(4) %11, i64 4
+ %15 = load i16, ptr addrspace(4) %14, align 4, !range !13, !invariant.load !12
+ %conv.i.i.i.i.i.i = zext i16 %15 to i32
+ %mul.i.i.i.i.i.i = mul i32 %10, %conv.i.i.i.i.i.i
+ %sub.i.i.i.i.i.i = sub i32 %13, %mul.i.i.i.i.i.i
+ %16 = call i32 @llvm.umin.i32(i32 %sub.i.i.i.i.i.i, i32 %conv.i.i.i.i.i.i) #13
+ %17 = call i32 @llvm.amdgcn.workitem.id.x() #13
+ %cmp.i.i.i.i26 = icmp ult i32 %17, %16
+ call void @llvm.assume(i1 %cmp.i.i.i.i26) #13
+ %sub.i.i.i27 = add nsw i32 %16, -1
+ %and.i.i.i28 = and i32 %sub.i.i.i27, -64
+ %cmp.i2.i.i29 = icmp eq i32 %17, %and.i.i.i28
+ br i1 %cmp.i2.i.i29, label %if.then.i30, label %_ZN4_OMP5state4initEb.exit55.critedge
+
+if.then.i30: ; preds = %if.else
+ store i32 0, ptr addrspace(3) @IsSPMDMode, align 4, !tbaa !14
+ %arrayidx.i.i46 = getelementptr inbounds [1024 x i8], ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::SharedMemorySmartStackTy", ptr addrspace(3) @_ZN12_GLOBAL__N_122SharedMemorySmartStackE, i32 0, i32 1, i32 0), i32 0, i32 %17
+ store i8 0, ptr addrspace(3) %arrayidx.i.i46, align 1, !tbaa !18
+ %sub.i.i = add nsw i32 %16, -64
+ store i32 %sub.i.i, ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, align 8, !tbaa !19
+ store i32 0, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 1), align 4, !tbaa !23
+ store i32 0, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 2), align 8, !tbaa !24
+ store i32 1, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 3), align 4, !tbaa !25
+ store i32 1, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 4), align 8, !tbaa !26
+ store i32 1, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 5), align 4, !tbaa !27
+ store i32 1, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 1), align 8, !tbaa !28
+ store ptr null, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 2), align 8, !tbaa !29
+ br label %_ZN4_OMP5state4initEb.exit55
+
+_ZN4_OMP5state4initEb.exit55.critedge: ; preds = %if.else
+ %arrayidx.i.i46.c = getelementptr inbounds [1024 x i8], ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::SharedMemorySmartStackTy", ptr addrspace(3) @_ZN12_GLOBAL__N_122SharedMemorySmartStackE, i32 0, i32 1, i32 0), i32 0, i32 %17
+ store i8 0, ptr addrspace(3) %arrayidx.i.i46.c, align 1, !tbaa !18
+ br label %_ZN4_OMP5state4initEb.exit55
+
+_ZN4_OMP5state4initEb.exit55: ; preds = %_ZN4_OMP5state4initEb.exit55.critedge, %if.then.i30
+ %arrayidx.i53 = getelementptr inbounds [1024 x ptr], ptr addrspace(3) @_ZN12_GLOBAL__N_112ThreadStatesE, i32 0, i32 %17
+ store ptr null, ptr addrspace(3) %arrayidx.i53, align 8, !tbaa !30
+ br label %if.end
+
+if.end: ; preds = %_ZN4_OMP5state4initEb.exit55, %_ZN4_OMP5state4initEb.exit
+ %18 = call i32 @llvm.amdgcn.workgroup.id.x() #13
+ %19 = call align 4 dereferenceable(64) ptr addrspace(4) @llvm.amdgcn.dispatch.ptr() #13
+ %20 = getelementptr i8, ptr addrspace(4) %19, i64 12
+ %21 = load i32, ptr addrspace(4) %20, align 4, !invariant.load !12
+ %22 = getelementptr i8, ptr addrspace(4) %19, i64 4
+ %23 = load i16, ptr addrspace(4) %22, align 4, !range !13, !invariant.load !12
+ %conv.i.i.i.i.i73 = zext i16 %23 to i32
+ %mul.i.i.i.i.i74 = mul i32 %18, %conv.i.i.i.i.i73
+ %sub.i.i.i.i.i75 = sub i32 %21, %mul.i.i.i.i.i74
+ %24 = call i32 @llvm.umin.i32(i32 %sub.i.i.i.i.i75, i32 %conv.i.i.i.i.i73) #13
+ %25 = call i32 @llvm.amdgcn.workitem.id.x() #13
+ %cmp.i.i.i79 = icmp ult i32 %25, %24
+ call void @llvm.assume(i1 %cmp.i.i.i79) #13
+ br i1 %tobool.not, label %_ZN4_OMP7mapping23isInitialThreadInLevel0Eb.exit, label %_ZN4_OMP7mapping12getBlockSizeEb.exit.i64
+
+_ZN4_OMP7mapping12getBlockSizeEb.exit.i64: ; preds = %if.end
+ %26 = load i32, ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, align 8
+ %cmp.i.i.i63 = icmp eq i32 %24, %26
+ call void @llvm.assume(i1 %cmp.i.i.i63) #13
+ %27 = load i32, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 1), align 4
+ %cmp9.i.i.i = icmp eq i32 %27, 0
+ call void @llvm.assume(i1 %cmp9.i.i.i) #13
+ %28 = load i32, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 2), align 8
+ %cmp19.i.i.i = icmp eq i32 %28, 0
+ call void @llvm.assume(i1 %cmp19.i.i.i) #13
+ %29 = load i32, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 3), align 4
+ %cmp29.i.i.i = icmp eq i32 %29, 1
+ call void @llvm.assume(i1 %cmp29.i.i.i) #13
+ %30 = load i32, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 4), align 8
+ %cmp39.i.i.i = icmp eq i32 %30, 1
+ call void @llvm.assume(i1 %cmp39.i.i.i) #13
+ %31 = load i32, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 0, i32 5), align 4
+ %cmp49.i.i.i = icmp eq i32 %31, 1
+ call void @llvm.assume(i1 %cmp49.i.i.i) #13
+ %32 = load i32, ptr addrspace(3) getelementptr inbounds (%"struct.(anonymous namespace)::TeamStateTy", ptr addrspace(3) @_ZN12_GLOBAL__N_19TeamStateE, i32 0, i32 1), align 8
+ %cmp.i.i67 = icmp eq i32 %32, 1
+ call void @llvm.assume(i1 %cmp.i.i67) #13
+ %33 = load i32, ptr addrspace(3) @IsSPMDMode, align 4, !tbaa !14
+ %tobool.i59.i = icmp ne i32 %33, 0
+ call void @llvm.assume(i1 %tobool.i59.i) #13
+ br label %_ZN14DebugEntryRAIID2Ev.exit250
+
+_ZN4_OMP7mapping23isInitialThreadInLevel0Eb.exit: ; preds = %if.end
+ %sub.i.i83 = add nsw i32 %24, -1
+ %and.i.i84 = and i32 %sub.i.i83, -64
+ %cmp.i2.i = icmp eq i32 %25, %and.i.i84
+ br i1 %cmp.i2.i, label %_ZN14DebugEntryRAIID2Ev.exit250, label %if.end10
+
+if.end10: ; preds = %_ZN4_OMP7mapping23isInitialThreadInLevel0Eb.exit
+ %sub.i = add nsw i32 %24, -64
+ %cmp = icmp ult i32 %25, %sub.i
+ %or.cond251 = select i1 %UseGenericStateMachine, i1 %cmp, i1 false
+ br i1 %or.cond251, label %do.body.i, label %_ZN14DebugEntryRAIID2Ev.exit250
+
+do.body.i: ; preds = %if.end10
+ call void @llvm.amdgcn.s.barrier() #13
+ br label %_ZN14DebugEntryRAIID2Ev.exit250
+
+_ZN14DebugEntryRAIID2Ev.exit250: ; preds = %do.body.i, %if.end10, %_ZN4_OMP7mapping23isInitialThreadInLevel0Eb.exit, %_ZN4_OMP7mapping12getBlockSizeEb.exit.i64
+ %retval.0 = phi i32 [ -1, %_ZN4_OMP7mapping12getBlockSizeEb.exit.i64 ], [ -1, %_ZN4_OMP7mapping23isInitialThreadInLevel0Eb.exit ], [ %25, %do.body.i ], [ %25, %if.end10 ]
+ ret i32 %retval.0
+}
+
+; Function Attrs: nounwind
+define internal void @__kmpc_target_deinit(ptr nocapture noundef readnone %Ident, i8 noundef signext %Mode, i1 noundef zeroext %0) local_unnamed_addr #10 {
+ ret void
+}
+
+; Function Attrs: convergent nounwind
+declare void @__kmpc_parallel_51(ptr nocapture noundef readnone %ident, i32 noundef %0, i32 noundef %if_expr, i32 noundef %num_threads, i32 noundef %proc_bind, ptr noundef %fn, ptr noundef %wrapper_fn, ptr noundef %args, i64 noundef %nargs)
+
+; Function Attrs: argmemonly nofree nounwind willreturn
+declare void @llvm.memcpy.p0.p0.i64(ptr noalias nocapture writeonly, ptr noalias nocapture readonly, i64, i1 immarg) #12
+
+attributes #0 = { alwaysinline convergent norecurse nounwind "frame-pointer"="none" "kernel" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx908" "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+mai-insts,+s-memrealtime,+s-memtime-inst" }
+attributes #1 = { argmemonly nocallback nofree nosync nounwind willreturn }
+attributes #2 = { alwaysinline mustprogress nofree norecurse nosync nounwind readnone willreturn "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx908" "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+mai-insts,+s-memrealtime,+s-memtime-inst" }
+attributes #3 = { mustprogress nofree norecurse nosync nounwind readnone willreturn "frame-pointer"="none" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx908" "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+mai-insts,+s-memrealtime,+s-memtime-inst" }
+attributes #4 = { nounwind readnone speculatable willreturn }
+attributes #5 = { nocallback nofree nosync nounwind readnone speculatable willreturn }
+attributes #6 = { inaccessiblememonly nocallback nofree nosync nounwind willreturn }
+attributes #7 = { convergent nounwind willreturn }
+attributes #8 = { convergent mustprogress noinline nounwind willreturn "frame-pointer"="none" "llvm.assume"="ompx_aligned_barrier,ompx_no_call_asm,ompx_no_call_asm" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx908" "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+mai-insts,+s-memrealtime,+s-memtime-inst" }
+attributes #9 = { convergent nounwind "frame-pointer"="none" "llvm.assume"="ompx_no_call_asm,ompx_no_call_asm" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx908" "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+mai-insts,+s-memrealtime,+s-memtime-inst" }
+attributes #10 = { nounwind "frame-pointer"="none" "llvm.assume"="ompx_no_call_asm,ompx_no_call_asm" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx908" "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+mai-insts,+s-memrealtime,+s-memtime-inst" }
+attributes #11 = { convergent nounwind "frame-pointer"="none" "llvm.assume"="ompx_no_call_asm" "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="gfx908" "target-features"="+16-bit-insts,+ci-insts,+dl-insts,+dot1-insts,+dot2-insts,+dot3-insts,+dot4-insts,+dot5-insts,+dot6-insts,+dot7-insts,+dpp,+flat-address-space,+gfx8-insts,+gfx9-insts,+mai-insts,+s-memrealtime,+s-memtime-inst" }
+attributes #12 = { argmemonly nofree nounwind willreturn }
+attributes #13 = { nounwind }
+attributes #14 = { convergent nounwind "llvm.assume"="ompx_aligned_barrier,ompx_no_call_asm" }
+attributes #15 = { convergent nounwind }
+
+!omp_offload.info = !{!0}
+!nvvm.annotations = !{!1}
+!llvm.module.flags = !{!2, !3, !4, !5}
+!llvm.ident = !{!6}
+
+!0 = !{i32 0, i32 32, i32 18757968, !"main", i32 12, i32 0}
+!1 = !{ptr @__omp_offloading_20_11e3950_main_l12, !"kernel", i32 1}
+!2 = !{i32 1, !"wchar_size", i32 4}
+!3 = !{i32 7, !"openmp", i32 50}
+!4 = !{i32 7, !"openmp-device", i32 50}
+!5 = !{i32 7, !"PIC Level", i32 2}
+!6 = !{!"clang version 15.0.0"}
+!7 = !{!8, !8, i64 0}
+!8 = !{!"any pointer", !9, i64 0}
+!9 = !{!"omnipotent char", !10, i64 0}
+!10 = !{!"Simple C/C++ TBAA"}
+!11 = !{i32 0, i32 1024}
+!12 = !{}
+!13 = !{i16 1, i16 1025}
+!14 = !{!15, !15, i64 0}
+!15 = !{!"int", !16, i64 0}
+!16 = !{!"omnipotent char", !17, i64 0}
+!17 = !{!"Simple C++ TBAA"}
+!18 = !{!16, !16, i64 0}
+!19 = !{!20, !15, i64 0}
+!20 = !{!"_ZTSN12_GLOBAL__N_111TeamStateTyE", !21, i64 0, !15, i64 24, !22, i64 32}
+!21 = !{!"_ZTSN12_GLOBAL__N_110ICVStateTyE", !15, i64 0, !15, i64 4, !15, i64 8, !15, i64 12, !15, i64 16, !15, i64 20}
+!22 = !{!"any pointer", !16, i64 0}
+!23 = !{!20, !15, i64 4}
+!24 = !{!20, !15, i64 8}
+!25 = !{!20, !15, i64 12}
+!26 = !{!20, !15, i64 16}
+!27 = !{!20, !15, i64 20}
+!28 = !{!20, !15, i64 24}
+!29 = !{!20, !22, i64 32}
+!30 = !{!22, !22, i64 0}
+!31 = !{!"branch_weights", i32 2000, i32 1}
More information about the llvm-commits
mailing list