[DA] conservatively mark the join of every divergent branch

For a loop, a join block is a block that is reachable along multiple
disjoint paths from the exiting block of a loop. If the exit condition
of the loop is divergent, then such join blocks must also be marked
divergent. This currently fails in some cases because not all join
blocks are identified correctly.

The workaround is to conservatively mark every join block of any
branch (not necessarily the exiting block of a loop) as divergent.

https://bugs.llvm.org/show_bug.cgi?id=46372

Reviewed By: simoll

Differential Revision: https://reviews.llvm.org/D81806

lib/Analysis/DivergenceAnalysis.cpp

@@ -295,14 +295,11 @@ bool DivergenceAnalysis::propagateJoinDivergence(const BasicBlock &JoinBlock,
   // push non-divergent phi nodes in JoinBlock to the worklist
   pushPHINodes(JoinBlock);
 
-  // JoinBlock is a divergent loop exit
-  if (BranchLoop && !BranchLoop->contains(&JoinBlock)) {
-    return true;
-  }
-
   // disjoint-paths divergent at JoinBlock
   markBlockJoinDivergent(JoinBlock);
-  return false;
+
+  // JoinBlock is a divergent loop exit
+  return BranchLoop && !BranchLoop->contains(&JoinBlock);
 }
 
 void DivergenceAnalysis::propagateBranchDivergence(const Instruction &Term) {

test/Analysis/DivergenceAnalysis/AMDGPU/join-at-loop-exit.ll

@@ -0,0 +1,36 @@
+; RUN: opt -mtriple amdgcn-unknown-amdhsa -analyze -divergence -use-gpu-divergence-analysis %s | FileCheck %s
+
+; CHECK: bb3:
+; CHECK: DIVERGENT:       %Guard.bb4 = phi i1 [ true, %bb1 ], [ false, %bb2 ]
+; CHECK: DIVERGENT:       br i1 %Guard.bb4, label %bb4, label %bb5
+
+; Function Attrs: nounwind readnone speculatable
+declare i32 @llvm.amdgcn.workitem.id.x() #0
+
+define protected amdgpu_kernel void @test() {
+bb0:
+  %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
+  %i5 = icmp eq i32 %tid.x, -1
+  br label %bb1
+
+bb1:                                              ; preds = %bb2, %bb0
+  %lsr.iv = phi i32 [ 7, %bb0 ], [ %lsr.iv.next, %bb2 ]
+  br i1 %i5, label %bb2, label %bb3
+
+bb2:                                              ; preds = %bb1
+  %lsr.iv.next = add nsw i32 %lsr.iv, -1
+  %i14 = icmp eq i32 %lsr.iv.next, 0
+  br i1 %i14, label %bb3, label %bb1
+
+bb3:                                              ; preds = %bb2, %bb1
+  %Guard.bb4 = phi i1 [ true, %bb1 ], [ false, %bb2 ]
+  br i1 %Guard.bb4, label %bb4, label %bb5
+
+bb4:                                              ; preds = %bb3
+  br label %bb5
+
+bb5:                                              ; preds = %bb3, %bb4
+  ret void
+}
+
+attributes #0 = { nounwind readnone speculatable }

test/Analysis/DivergenceAnalysis/AMDGPU/trivial-join-at-loop-exit.ll

@@ -0,0 +1,28 @@
+; RUN: opt -mtriple amdgcn-unknown-amdhsa -analyze -divergence -use-gpu-divergence-analysis %s | FileCheck %s
+; XFAIL: *
+
+; https://bugs.llvm.org/show_bug.cgi?id=46372
+
+; CHECK: bb2:
+; CHECK-NOT: DIVERGENT:       %Guard.bb2 = phi i1 [ true, %bb1 ], [ false, %bb0 ]
+
+; Function Attrs: nounwind readnone speculatable
+declare i32 @llvm.amdgcn.workitem.id.x() #0
+
+define protected amdgpu_kernel void @test2(i1 %uni) {
+bb0:
+  %tid.x = call i32 @llvm.amdgcn.workitem.id.x()
+  %i5 = icmp eq i32 %tid.x, -1
+  br i1 %uni, label %bb1, label %bb2
+
+bb1:                                              ; preds = %bb2, %bb0
+  %lsr.iv = phi i32 [ 7, %bb0 ], [ %lsr.iv.next, %bb1 ]
+  %lsr.iv.next = add nsw i32 %lsr.iv, -1
+  br i1 %i5, label %bb2, label %bb1
+
+bb2:                                              ; preds = %bb2, %bb1
+  %Guard.bb2 = phi i1 [ true, %bb1 ], [ false, %bb0 ]
+  ret void
+}
+
+attributes #0 = { nounwind readnone speculatable }