Merging r266088:

------------------------------------------------------------------------
r266088 | nhaehnle | 2016-04-12 09:10:38 -0700 (Tue, 12 Apr 2016) | 16 lines

AMDGPU/SI: Fix a mis-compilation of multi-level breaks

Summary:
Under certain circumstances, multi-level breaks (or what is understood by
the control flow passes as such) could be miscompiled in a way that causes
infinite loops, by emitting incorrect control flow intrinsics.

This fixes a hang in
dEQP-GLES3.functional.shaders.loops.while_dynamic_iterations.conditional_continue_vertex

Reviewers: arsenm, tstellarAMD

Subscribers: arsenm, llvm-commits

Differential Revision: http://reviews.llvm.org/D18967

------------------------------------------------------------------------

llvm-svn: 271730

llvm/lib/Target/AMDGPU/SIAnnotateControlFlow.cpp

@@ -252,7 +252,23 @@ Value *SIAnnotateControlFlow::handleLoopCondition(Value *Cond, PHINode *Broken,
 
       BasicBlock *From = Phi->getIncomingBlock(i);
       if (From == IDom) {
+        // We're in the following situation:
+        //   IDom/From
+        //      |   \
+        //      |   If-block
+        //      |   /
+        //     Parent
+        // where we want to break out of the loop if the If-block is not taken.
+        // Due to the depth-first traversal, there should be an end.cf
+        // intrinsic in Parent, and we insert an else.break before it.
+        //
+        // Note that the end.cf need not be the first non-phi instruction
+        // of parent, particularly when we're dealing with a multi-level
+        // break, but it should occur within a group of intrinsic calls
+        // at the beginning of the block.
         CallInst *OldEnd = dyn_cast<CallInst>(Parent->getFirstInsertionPt());
+        while (OldEnd && OldEnd->getCalledFunction() != EndCf)
+          OldEnd = dyn_cast<CallInst>(OldEnd->getNextNode());
         if (OldEnd && OldEnd->getCalledFunction() == EndCf) {
           Value *Args[] = { OldEnd->getArgOperand(0), NewPhi };
           Ret = CallInst::Create(ElseBreak, Args, "", OldEnd);

llvm/test/CodeGen/AMDGPU/multilevel-break.ll

@@ -0,0 +1,41 @@
+; RUN: opt -S -mtriple=amdgcn-- -structurizecfg -si-annotate-control-flow < %s | FileCheck %s
+
+; CHECK-LABEL: {{^}}define void @main
+; CHECK: main_body:
+; CHECK: LOOP.outer:
+; CHECK: LOOP:
+; CHECK:     [[if:%[0-9]+]] = call { i1, i64 } @llvm.SI.if(
+; CHECK:     [[if_exec:%[0-9]+]] = extractvalue { i1, i64 } [[if]], 1
+;
+; CHECK: Flow:
+;
+; Ensure two else.break calls, for both the inner and outer loops
+;
+; CHECK:        call i64 @llvm.SI.else.break(i64 [[if_exec]],
+; CHECK-NEXT:   call i64 @llvm.SI.else.break(i64 [[if_exec]],
+; CHECK-NEXT:   call void @llvm.SI.end.cf
+;
+; CHECK: Flow1:
+define void @main(<4 x float> %vec, i32 %ub, i32 %cont) {
+main_body:
+  br label %LOOP.outer
+
+LOOP.outer:                                       ; preds = %ENDIF, %main_body
+  %tmp43 = phi i32 [ 0, %main_body ], [ %tmp47, %ENDIF ]
+  br label %LOOP
+
+LOOP:                                             ; preds = %ENDIF, %LOOP.outer
+  %tmp45 = phi i32 [ %tmp43, %LOOP.outer ], [ %tmp47, %ENDIF ]
+  %tmp47 = add i32 %tmp45, 1
+  %tmp48 = icmp slt i32 %tmp45, %ub
+  br i1 %tmp48, label %ENDIF, label %IF
+
+IF:                                               ; preds = %LOOP
+  ret void
+
+ENDIF:                                            ; preds = %LOOP
+  %tmp51 = icmp eq i32 %tmp47, %cont
+  br i1 %tmp51, label %LOOP, label %LOOP.outer
+}
+
+attributes #0 = { nounwind readnone }