Require break-critical-edges to run before reg2mem_hlsl (#5180)

* Require break-critical-edges to run before reg2mem_hlsl

A DXIL shader containing a loop, where the `phi` node at the
beginning of the loop is re-used outside the loop, does not get
translated correctly by this version of reg2mem. Running critical
edge splitting beforehand resolves the issue.

* Add a test which verifies the fix

lib/Transforms/Scalar/Reg2MemHLSL.cpp

@@ -33,6 +33,8 @@ namespace {
     }
 
     void getAnalysisUsage(AnalysisUsage &AU) const override {
+        AU.addRequiredID(BreakCriticalEdgesID);
+        AU.addPreservedID(BreakCriticalEdgesID);
     }
 
     bool valueEscapes(const Instruction *Inst) const {
@@ -198,6 +200,7 @@ namespace {
 char RegToMemHlsl::ID = 0;
 INITIALIZE_PASS_BEGIN(RegToMemHlsl, "reg2mem_hlsl", "Demote values with phi-node usage to stack slots",
                 false, false)
+INITIALIZE_PASS_DEPENDENCY(BreakCriticalEdges)
 INITIALIZE_PASS_END(RegToMemHlsl,   "reg2mem_hlsl", "Demote values with phi-node usage to stack slots",
                 false, false)
 

tools/clang/test/HLSLFileCheck/passes/hl/mem2reg_hlsl/mem2reg_hlsl.ll

@@ -3,8 +3,8 @@
 ; Make sure store is after load.
 ; CHECK: while.body:
 ; CHECK: load i32
+; CHECK: while.body.while.cond_crit_edge:
 ; CHECK: store i32 -1
-; CHECK: switch
 
 ; ModuleID = 'MyModule'
 target triple = "dxil-ms-dx"

tools/clang/test/HLSLFileCheck/passes/hl/mem2reg_hlsl/mem2reg_loop_crit_edge.ll

@@ -0,0 +1,144 @@
+; RUN: %opt %s -simplifycfg -reg2mem_hlsl -S | FileCheck %s
+
+; Make sure loop variable isn't rewritten in the loop basic block
+; CHECK-LABEL: loop.inner
+; CHECK: load i32, i32* [[Addr:%.*]]
+; CHECK-NOT: store i32 {{.*}}, i32* [[Addr]]
+; CHECK-NOT: br{{.*}}%loop.inner
+; CHECK: br
+; CHECK: store i32 {{.*}}, i32* [[Addr]]
+; CHECK: br label %loop.inner
+
+; ModuleID = 'MyModule'
+target triple = "dxil-ms-dx"
+
+define void @main() {
+  %1 = call float @dx.op.loadInput.f32(i32 4, i32 0, i32 0, i8 0, i32 undef)
+  %2 = call float @dx.op.loadInput.f32(i32 4, i32 0, i32 0, i8 1, i32 undef)
+  %3 = call float @dx.op.loadInput.f32(i32 4, i32 0, i32 0, i8 2, i32 undef)
+  %4 = call float @dx.op.loadInput.f32(i32 4, i32 0, i32 0, i8 3, i32 undef)
+  %5 = call float @dx.op.loadInput.f32(i32 4, i32 1, i32 0, i8 0, i32 undef)
+  %6 = call float @dx.op.loadInput.f32(i32 4, i32 1, i32 0, i8 1, i32 undef)
+  %7 = call float @dx.op.loadInput.f32(i32 4, i32 1, i32 0, i8 2, i32 undef)
+  %8 = call float @dx.op.loadInput.f32(i32 4, i32 1, i32 0, i8 3, i32 undef)
+  %9 = bitcast float %5 to i32
+  %10 = bitcast float %6 to i32
+  %11 = bitcast float %7 to i32
+  %12 = bitcast float %8 to i32
+  %13 = bitcast i32 0 to float
+  br label %loop.outer
+
+loop.outer:                                      ; preds = %49, %0
+  %14 = phi i32 [ 0, %0 ], [ %45, %49 ]
+  %15 = phi i32 [ %9, %0 ], [ %36, %49 ]
+  %16 = phi i32 [ %10, %0 ], [ %37, %49 ]
+  %17 = phi i32 [ %11, %0 ], [ %38, %49 ]
+  %18 = phi i32 [ %12, %0 ], [ %39, %49 ]
+  %19 = bitcast i32 0 to float
+  br label %loop.inner
+
+loop.inner:                                      ; preds = %43, %loop.outer
+  %20 = phi i32 [ %15, %loop.outer ], [ %36, %43 ]
+  %21 = phi i32 [ %16, %loop.outer ], [ %37, %43 ]
+  %22 = phi i32 [ %17, %loop.outer ], [ %38, %43 ]
+  %23 = phi i32 [ %18, %loop.outer ], [ %39, %43 ]
+  %24 = bitcast i32 %21 to float
+  %25 = bitcast i32 1065353216 to float
+  %26 = fadd fast float %24, %25
+  %27 = bitcast i32 %22 to float
+  %28 = bitcast i32 1065353216 to float
+  %29 = fadd fast float %27, %28
+  %30 = bitcast i32 %23 to float
+  %31 = bitcast i32 1065353216 to float
+  %32 = fadd fast float %30, %31
+  %33 = bitcast i32 %20 to float
+  %34 = bitcast i32 1065353216 to float
+  %35 = fadd fast float %33, %34
+  %36 = bitcast float %26 to i32
+  %37 = bitcast float %29 to i32
+  %38 = bitcast float %32 to i32
+  %39 = bitcast float %35 to i32
+  %40 = icmp slt i32 %14, 3
+  br i1 %40, label %41, label %42
+
+; <label>:41                                      ; preds = %loop.inner
+  br label %44
+
+; <label>:42                                      ; preds = %loop.inner
+  br label %43
+
+; <label>:43                                      ; preds = %42
+  br label %loop.inner
+
+; <label>:44                                      ; preds = %41
+  %45 = add i32 %14, 1
+  %46 = icmp sge i32 %45, 3
+  br i1 %46, label %47, label %48
+
+; <label>:47                                      ; preds = %44
+  br label %50
+
+; <label>:48                                      ; preds = %44
+  br label %49
+
+; <label>:49                                      ; preds = %48
+  br label %loop.outer
+
+; <label>:50                                      ; preds = %47
+  %51 = bitcast i32 -1073741824 to float
+  %52 = bitcast i32 %21 to float
+  %53 = fadd fast float %51, %52
+  %54 = bitcast i32 -1073741824 to float
+  %55 = bitcast i32 %22 to float
+  %56 = fadd fast float %54, %55
+  %57 = bitcast i32 -1073741824 to float
+  %58 = bitcast i32 %23 to float
+  %59 = fadd fast float %57, %58
+  call void @dx.op.storeOutput.f32(i32 5, i32 0, i32 0, i8 0, float %53)
+  call void @dx.op.storeOutput.f32(i32 5, i32 0, i32 0, i8 1, float %56)
+  call void @dx.op.storeOutput.f32(i32 5, i32 0, i32 0, i8 2, float %59)
+  %60 = bitcast i32 -1082130432 to float
+  %61 = fmul fast float %2, %60
+  call void @dx.op.storeOutput.f32(i32 5, i32 1, i32 0, i8 0, float %1)
+  call void @dx.op.storeOutput.f32(i32 5, i32 1, i32 0, i8 1, float %61)
+  call void @dx.op.storeOutput.f32(i32 5, i32 1, i32 0, i8 2, float %3)
+  call void @dx.op.storeOutput.f32(i32 5, i32 1, i32 0, i8 3, float %4)
+  ret void
+}
+
+; Function Attrs: nounwind readnone
+declare float @dx.op.loadInput.f32(i32, i32, i32, i8, i32) #0
+
+; Function Attrs: nounwind
+declare void @dx.op.storeOutput.f32(i32, i32, i32, i8, float) #1
+
+attributes #0 = { nounwind readnone }
+attributes #1 = { nounwind }
+
+!llvm.ident = !{!0}
+!dx.version = !{!1}
+!dx.valver = !{!1}
+!dx.shaderModel = !{!2}
+!dx.typeAnnotations = !{!3}
+!dx.entryPoints = !{!7}
+
+!0 = !{!"Mesa version 23.2.0-devel (git-e2603275dc)"}
+!1 = !{i32 1, i32 7}
+!2 = !{!"vs", i32 6, i32 7}
+!3 = !{i32 1, void ()* @main, !4}
+!4 = !{!5}
+!5 = !{i32 0, !6, !6}
+!6 = !{}
+!7 = !{void ()* @main, !"main", !8, null, null}
+!8 = !{!9, !15, null}
+!9 = !{!10, !13}
+!10 = !{i32 0, !"TEXCOORD", i8 9, i8 0, !11, i8 0, i32 1, i8 4, i32 0, i8 0, !12}
+!11 = !{i32 0}
+!12 = !{i32 3, i8 1}
+!13 = !{i32 1, !"TEXCOORD", i8 9, i8 0, !14, i8 0, i32 1, i8 4, i32 1, i8 0, !12}
+!14 = !{i32 1}
+!15 = !{!16, !18}
+!16 = !{i32 0, !"TEXCOORD", i8 9, i8 0, !11, i8 2, i32 1, i8 3, i32 0, i8 0, !17}
+!17 = !{i32 3, i8 7}
+!18 = !{i32 1, !"SV_Position", i8 9, i8 3, !11, i8 4, i32 1, i8 4, i32 1, i8 0, !19}
+!19 = !{i32 3, i8 15}