[SimplifyCFG] Don't block sinking for allocas if no phi created (#104579)

SimplifyCFG sinking currently does not sink loads/stores of allocas,
because historically SROA was unable to handle the resulting IR. Since
then, SROA both learned to speculate loads/stores over selects and phis,
*and* SimplifyCFG sinking has been deferred to the end of the function
simplification pipeline, which means that SROA happens before it.

As such, I believe that this workaround should no longer be necessary.
Given how sensitive SimplifyCFG sinking seems to be, this patch takes a
very conservative step towards removing this, by allowing sinking if we
don't actually need to form a phi over the pointer argument.

This fixes #104567, where
sinking a store to an escaped alloca allows converting a switch into
arithmetic.

Author: nikic

llvm/lib/Transforms/Utils/SimplifyCFG.cpp

@@ -1994,28 +1994,6 @@ static bool canSinkInstructions(
       return false;
   }
 
-  // Because SROA can't handle speculating stores of selects, try not to sink
-  // loads, stores or lifetime markers of allocas when we'd have to create a
-  // PHI for the address operand. Also, because it is likely that loads or
-  // stores of allocas will disappear when Mem2Reg/SROA is run, don't sink
-  // them.
-  // This can cause code churn which can have unintended consequences down
-  // the line - see https://llvm.org/bugs/show_bug.cgi?id=30244.
-  // FIXME: This is a workaround for a deficiency in SROA - see
-  // https://llvm.org/bugs/show_bug.cgi?id=30188
-  if (isa<StoreInst>(I0) && any_of(Insts, [](const Instruction *I) {
-        return isa<AllocaInst>(I->getOperand(1)->stripPointerCasts());
-      }))
-    return false;
-  if (isa<LoadInst>(I0) && any_of(Insts, [](const Instruction *I) {
-        return isa<AllocaInst>(I->getOperand(0)->stripPointerCasts());
-      }))
-    return false;
-  if (isLifeTimeMarker(I0) && any_of(Insts, [](const Instruction *I) {
-        return isa<AllocaInst>(I->getOperand(1)->stripPointerCasts());
-      }))
-    return false;
-
   // For calls to be sinkable, they must all be indirect, or have same callee.
   // I.e. if we have two direct calls to different callees, we don't want to
   // turn that into an indirect call. Likewise, if we have an indirect call,
@@ -2053,6 +2031,27 @@ static bool canSinkInstructions(
       return I->getOperand(OI) == I0->getOperand(OI);
     };
     if (!all_of(Insts, SameAsI0)) {
+      // Because SROA historically couldn't handle speculating stores of
+      // selects, we try not to sink loads, stores or lifetime markers of
+      // allocas when we'd have to create a PHI for the address operand.
+      // TODO: SROA supports speculation for loads and stores now -- remove
+      // this hack?
+      if (isa<StoreInst>(I0) && OI == 1 &&
+          any_of(Insts, [](const Instruction *I) {
+            return isa<AllocaInst>(I->getOperand(1)->stripPointerCasts());
+          }))
+        return false;
+      if (isa<LoadInst>(I0) && OI == 0 &&
+          any_of(Insts, [](const Instruction *I) {
+            return isa<AllocaInst>(I->getOperand(0)->stripPointerCasts());
+          }))
+        return false;
+      if (isLifeTimeMarker(I0) && OI == 1 &&
+          any_of(Insts, [](const Instruction *I) {
+            return isa<AllocaInst>(I->getOperand(1)->stripPointerCasts());
+          }))
+        return false;
+
       if ((isa<Constant>(Op) && !replacingOperandWithVariableIsCheap(I0, OI)) ||
           !canReplaceOperandWithVariable(I0, OI))
         // We can't create a PHI from this GEP.

llvm/test/CodeGen/Hexagon/block-addr.ll

@@ -1,4 +1,4 @@
-; RUN: llc -march=hexagon < %s | FileCheck %s
+; RUN: llc -march=hexagon -hexagon-initial-cfg-cleanup=0 < %s | FileCheck %s
 
 ; CHECK-DAG: r[[REG:[0-9]+]] = memw(r{{[0-9]+<<#[0-9]+}}+##.LJTI{{.*}})
 ; CHECK-DAG: jumpr r[[REG]]

llvm/test/DebugInfo/ARM/single-constant-use-preserves-dbgloc.ll

@@ -1,4 +1,4 @@
-; RUN: llc -filetype=asm -asm-verbose=0 < %s | FileCheck %s
+; RUN: llc -filetype=asm -asm-verbose=0 -arm-atomic-cfg-tidy=0 < %s | FileCheck %s
 
 ; int main()
 ; {

llvm/test/Transforms/SimplifyCFG/X86/sink-common-code.ll

@@ -838,15 +838,14 @@ define i32 @test_pr30188a(i1 zeroext %flag, i32 %x) {
 ; CHECK:       if.then:
 ; CHECK-NEXT:    call void @g()
 ; CHECK-NEXT:    [[ONE:%.*]] = load i32, ptr [[Y]], align 4
-; CHECK-NEXT:    [[TWO:%.*]] = add i32 [[ONE]], 2
-; CHECK-NEXT:    store i32 [[TWO]], ptr [[Y]], align 4
 ; CHECK-NEXT:    br label [[IF_END:%.*]]
 ; CHECK:       if.else:
 ; CHECK-NEXT:    [[THREE:%.*]] = load i32, ptr [[Z]], align 4
-; CHECK-NEXT:    [[FOUR:%.*]] = add i32 [[THREE]], 2
-; CHECK-NEXT:    store i32 [[FOUR]], ptr [[Y]], align 4
 ; CHECK-NEXT:    br label [[IF_END]]
 ; CHECK:       if.end:
+; CHECK-NEXT:    [[THREE_SINK:%.*]] = phi i32 [ [[THREE]], [[IF_ELSE]] ], [ [[ONE]], [[IF_THEN]] ]
+; CHECK-NEXT:    [[FOUR:%.*]] = add i32 [[THREE_SINK]], 2
+; CHECK-NEXT:    store i32 [[FOUR]], ptr [[Y]], align 4
 ; CHECK-NEXT:    ret i32 1
 ;
 entry:
@@ -914,15 +913,16 @@ define zeroext i1 @test_pr30244(i1 zeroext %flag, i1 zeroext %flag2, i32 %blksA,
 ; CHECK-NEXT:    br i1 [[FLAG:%.*]], label [[IF_THEN:%.*]], label [[IF_ELSE:%.*]]
 ; CHECK:       if.then:
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp uge i32 [[BLKSA:%.*]], [[NBLKS:%.*]]
-; CHECK-NEXT:    [[FROMBOOL1:%.*]] = zext i1 [[CMP]] to i8
-; CHECK-NEXT:    store i8 [[FROMBOOL1]], ptr [[P]], align 1
-; CHECK-NEXT:    br label [[IF_END:%.*]]
+; CHECK-NEXT:    br label [[IF_END_SINK_SPLIT:%.*]]
 ; CHECK:       if.else:
-; CHECK-NEXT:    br i1 [[FLAG2:%.*]], label [[IF_THEN2:%.*]], label [[IF_END]]
+; CHECK-NEXT:    br i1 [[FLAG2:%.*]], label [[IF_THEN2:%.*]], label [[IF_END:%.*]]
 ; CHECK:       if.then2:
 ; CHECK-NEXT:    [[ADD:%.*]] = add i32 [[NBLKS]], [[BLKSB:%.*]]
 ; CHECK-NEXT:    [[CMP2:%.*]] = icmp ule i32 [[ADD]], [[BLKSA]]
-; CHECK-NEXT:    [[FROMBOOL3:%.*]] = zext i1 [[CMP2]] to i8
+; CHECK-NEXT:    br label [[IF_END_SINK_SPLIT]]
+; CHECK:       if.end.sink.split:
+; CHECK-NEXT:    [[CMP2_SINK:%.*]] = phi i1 [ [[CMP2]], [[IF_THEN2]] ], [ [[CMP]], [[IF_THEN]] ]
+; CHECK-NEXT:    [[FROMBOOL3:%.*]] = zext i1 [[CMP2_SINK]] to i8
 ; CHECK-NEXT:    store i8 [[FROMBOOL3]], ptr [[P]], align 1
 ; CHECK-NEXT:    br label [[IF_END]]
 ; CHECK:       if.end:

llvm/test/Transforms/SimplifyCFG/sink-and-convert-switch.ll

@@ -9,23 +9,8 @@ define void @pr104567(i8 %x, ptr %f) {
 ; CHECK-NEXT:  [[START:.*:]]
 ; CHECK-NEXT:    [[Y:%.*]] = alloca [1 x i8], align 1
 ; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 1, ptr nonnull [[Y]])
-; CHECK-NEXT:    switch i8 [[X]], label %[[DEFAULT_UNREACHABLE:.*]] [
-; CHECK-NEXT:      i8 0, label %[[BB4:.*]]
-; CHECK-NEXT:      i8 1, label %[[BB3:.*]]
-; CHECK-NEXT:      i8 2, label %[[BB2:.*]]
-; CHECK-NEXT:    ]
-; CHECK:       [[DEFAULT_UNREACHABLE]]:
-; CHECK-NEXT:    unreachable
-; CHECK:       [[BB4]]:
-; CHECK-NEXT:    store i8 4, ptr [[Y]], align 1
-; CHECK-NEXT:    br label %[[BB5:.*]]
-; CHECK:       [[BB3]]:
-; CHECK-NEXT:    store i8 5, ptr [[Y]], align 1
-; CHECK-NEXT:    br label %[[BB5]]
-; CHECK:       [[BB2]]:
-; CHECK-NEXT:    store i8 6, ptr [[Y]], align 1
-; CHECK-NEXT:    br label %[[BB5]]
-; CHECK:       [[BB5]]:
+; CHECK-NEXT:    [[SWITCH_OFFSET:%.*]] = add nsw i8 [[X]], 4
+; CHECK-NEXT:    store i8 [[SWITCH_OFFSET]], ptr [[Y]], align 1
 ; CHECK-NEXT:    call void [[F]](ptr [[Y]])
 ; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 1, ptr nonnull [[Y]])
 ; CHECK-NEXT:    ret void