[llvm] 2245fb8 - [SimplifyCFG] FoldBranchToCommonDest: lift use-restriction on bonus instructions
Roman Lebedev via llvm-commits
llvm-commits at lists.llvm.org
Thu Nov 26 11:52:12 PST 2020
Author: Roman Lebedev
Date: 2020-11-26T22:51:22+03:00
New Revision: 2245fb8aaa1c1f85f53f7b19a1ee3ac69b1a1dfe
URL: https://github.com/llvm/llvm-project/commit/2245fb8aaa1c1f85f53f7b19a1ee3ac69b1a1dfe
DIFF: https://github.com/llvm/llvm-project/commit/2245fb8aaa1c1f85f53f7b19a1ee3ac69b1a1dfe.diff
LOG: [SimplifyCFG] FoldBranchToCommonDest: lift use-restriction on bonus instructions
1. It doesn't make sense to enforce that the bonus instruction
is only used once in it's basic block. What matters is
whether those user instructions fit within our budget, sure,
but that is another question.
2. It doesn't make sense to enforce that said bonus instructions
are only used within their basic block. Perhaps the branch
condition isn't using the value computed by said bonus instruction,
and said bonus instruction is simply being calculated
to be used in successors?
So iff we can clone bonus instructions, to lift these restrictions,
we just need to carefully update their external uses
to use the new cloned instructions.
Notably, this transform (even without this change) appears to be
poison-unsafe as per alive2, but is otherwise (including the patch) legal.
We don't introduce any new PHI nodes, but only "move" the instructions
around, i'm not really seeing much potential for extra cost modelling
for the transform, especially since now we allow at most one such
bonus instruction by default.
This causes the fold to fire +11.4% more (13216 -> 14725)
as of vanilla llvm test-suite + RawSpeed.
The motivational pattern is IEEE-754-2008 Binary16->Binary32
extension code:
https://github.com/darktable-org/rawspeed/blob/ca57d77fb2ba81f21fc712cfac26e54f46406473/src/librawspeed/common/FloatingPoint.h#L115-L120
^ that should be a switch, but it is not now: https://godbolt.org/z/bvja5v
That being said, even thought this seemed like this would fix it: https://godbolt.org/z/xGq3TM
apparently that fold is happening somewhere else afterall,
so something else also has a similar 'artificial' restriction.
Added:
Modified:
llvm/lib/Transforms/Utils/SimplifyCFG.cpp
llvm/test/Transforms/LoopUnroll/peel-loop-inner.ll
llvm/test/Transforms/SimplifyCFG/fold-branch-to-common-dest.ll
Removed:
################################################################################
diff --git a/llvm/lib/Transforms/Utils/SimplifyCFG.cpp b/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
index 0f4c1b4038e5..1bb6ea794bb0 100644
--- a/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
@@ -2779,15 +2779,9 @@ bool llvm::FoldBranchToCommonDest(BranchInst *BI, MemorySSAUpdater *MSSAU,
// Ignore dbg intrinsics.
if (isa<DbgInfoIntrinsic>(I))
continue;
- if (!I->hasOneUse() || !isSafeToSpeculativelyExecute(&*I))
+ // I must be safe to execute unconditionally.
+ if (!isSafeToSpeculativelyExecute(&*I))
return Changed;
- // I has only one use and can be executed unconditionally.
- Instruction *User = dyn_cast<Instruction>(I->user_back());
- if (User == nullptr || User->getParent() != BB)
- return Changed;
- // I is used in the same BB. Since BI uses Cond and doesn't have more slots
- // to use any other instruction, User must be an instruction between next(I)
- // and Cond.
// Account for the cost of duplicating this instruction into each
// predecessor.
@@ -2883,6 +2877,13 @@ bool llvm::FoldBranchToCommonDest(BranchInst *BI, MemorySSAUpdater *MSSAU,
PBI->swapSuccessors();
}
+ // Before cloning instructions, notify the successor basic block that it
+ // is about to have a new predecessor. This will update PHI nodes,
+ // which will allow us to update live-out uses of bonus instructions.
+ if (BI->isConditional())
+ AddPredecessorToBlock(PBI->getSuccessor(0) == BB ? TrueDest : FalseDest,
+ PredBlock, BB, MSSAU);
+
// If we have bonus instructions, clone them into the predecessor block.
// Note that there may be multiple predecessor blocks, so we cannot move
// bonus instructions to a predecessor block.
@@ -2914,6 +2915,18 @@ bool llvm::FoldBranchToCommonDest(BranchInst *BI, MemorySSAUpdater *MSSAU,
PredBlock->getInstList().insert(PBI->getIterator(), NewBonusInst);
NewBonusInst->takeName(&*BonusInst);
BonusInst->setName(BonusInst->getName() + ".old");
+ BonusInst->replaceUsesWithIf(
+ NewBonusInst, [CurrBB = BonusInst->getParent(), PredBlock](Use &U) {
+ auto *User = cast<Instruction>(U.getUser());
+ // Ignore uses in the same block as the bonus instruction itself.
+ if (User->getParent() == CurrBB)
+ return false;
+ // We can safely update external non-PHI uses.
+ if (!isa<PHINode>(User))
+ return true;
+ // For PHI nodes, only update the uses for the current predecessor.
+ return cast<PHINode>(User)->getIncomingBlock(U) == PredBlock;
+ });
}
// Clone Cond into the predecessor basic block, and or/and the
@@ -2955,7 +2968,6 @@ bool llvm::FoldBranchToCommonDest(BranchInst *BI, MemorySSAUpdater *MSSAU,
(SuccFalseWeight + SuccTrueWeight) +
PredTrueWeight * SuccFalseWeight);
}
- AddPredecessorToBlock(TrueDest, PredBlock, BB, MSSAU);
PBI->setSuccessor(0, TrueDest);
}
if (PBI->getSuccessor(1) == BB) {
@@ -2970,7 +2982,6 @@ bool llvm::FoldBranchToCommonDest(BranchInst *BI, MemorySSAUpdater *MSSAU,
// FalseWeight is FalseWeight for PBI * FalseWeight for BI.
NewWeights.push_back(PredFalseWeight * SuccFalseWeight);
}
- AddPredecessorToBlock(FalseDest, PredBlock, BB, MSSAU);
PBI->setSuccessor(1, FalseDest);
}
if (NewWeights.size() == 2) {
diff --git a/llvm/test/Transforms/LoopUnroll/peel-loop-inner.ll b/llvm/test/Transforms/LoopUnroll/peel-loop-inner.ll
index 7b088b31a54c..664702c53595 100644
--- a/llvm/test/Transforms/LoopUnroll/peel-loop-inner.ll
+++ b/llvm/test/Transforms/LoopUnroll/peel-loop-inner.ll
@@ -1,3 +1,4 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
; RUN: opt < %s -S -passes='require<opt-remark-emit>,loop-unroll<peeling;no-runtime>,simplify-cfg,instcombine' -unroll-force-peel-count=3 -verify-dom-info | FileCheck %s
define void @basic(i32 %K, i32 %N) {
@@ -6,24 +7,19 @@ define void @basic(i32 %K, i32 %N) {
; CHECK-NEXT: br label [[OUTER:%.*]]
; CHECK: outer:
; CHECK-NEXT: [[I:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[I_INC:%.*]], [[OUTER_BACKEDGE:%.*]] ]
-; CHECK-NEXT: [[CMP_INNER_PEEL:%.*]] = icmp sgt i32 [[K:%.*]], 1
-; CHECK-NEXT: br i1 [[CMP_INNER_PEEL]], label [[INNER_PEEL2:%.*]], label [[OUTER_BACKEDGE]]
-; CHECK: inner.peel2:
-; CHECK-NEXT: [[CMP_INNER_PEEL4:%.*]] = icmp eq i32 [[K]], 2
-; CHECK-NEXT: br i1 [[CMP_INNER_PEEL4]], label [[OUTER_BACKEDGE]], label [[INNER_PEEL6:%.*]]
-; CHECK: inner.peel6:
-; CHECK-NEXT: [[CMP_INNER_PEEL8:%.*]] = icmp sgt i32 [[K]], 3
+; CHECK-NEXT: [[CMP_INNER_PEEL8:%.*]] = icmp sgt i32 [[K:%.*]], 3
; CHECK-NEXT: br i1 [[CMP_INNER_PEEL8]], label [[INNER:%.*]], label [[OUTER_BACKEDGE]]
; CHECK: inner:
-; CHECK-NEXT: [[J:%.*]] = phi i32 [ [[J_INC:%.*]], [[INNER]] ], [ 3, [[INNER_PEEL6]] ]
+; CHECK-NEXT: [[J:%.*]] = phi i32 [ [[J_INC:%.*]], [[INNER]] ], [ undef, [[OUTER]] ]
; CHECK-NEXT: [[J_INC]] = add nuw nsw i32 [[J]], 1
; CHECK-NEXT: [[CMP_INNER:%.*]] = icmp slt i32 [[J_INC]], [[K]]
-; CHECK-NEXT: br i1 [[CMP_INNER]], label [[INNER]], label [[OUTER_BACKEDGE]], !llvm.loop !0
+; CHECK-NEXT: br i1 [[CMP_INNER]], label [[INNER]], label [[OUTER_BACKEDGE]], [[LOOP0:!llvm.loop !.*]]
; CHECK: outer.backedge:
; CHECK-NEXT: [[I_INC]] = add i32 [[I]], 1
; CHECK-NEXT: [[CMP_OUTER:%.*]] = icmp slt i32 [[I_INC]], [[N:%.*]]
-; CHECK-NOT: !llvm.loop
+; CHECK-NEXT: br i1 [[CMP_OUTER]], label [[OUTER]], label [[END:%.*]]
; CHECK: end:
+; CHECK-NEXT: ret void
;
entry:
br label %outer
diff --git a/llvm/test/Transforms/SimplifyCFG/fold-branch-to-common-dest.ll b/llvm/test/Transforms/SimplifyCFG/fold-branch-to-common-dest.ll
index 1a6f7955fd80..e45a88c016dd 100644
--- a/llvm/test/Transforms/SimplifyCFG/fold-branch-to-common-dest.ll
+++ b/llvm/test/Transforms/SimplifyCFG/fold-branch-to-common-dest.ll
@@ -160,12 +160,11 @@ define void @one_pred_with_extra_op_multiuse(i8 %v0, i8 %v1) {
; CHECK-LABEL: @one_pred_with_extra_op_multiuse(
; CHECK-NEXT: pred:
; CHECK-NEXT: [[C0:%.*]] = icmp eq i8 [[V0:%.*]], 0
-; CHECK-NEXT: br i1 [[C0]], label [[DISPATCH:%.*]], label [[FINAL_RIGHT:%.*]]
-; CHECK: dispatch:
; CHECK-NEXT: [[V1_ADJ:%.*]] = add i8 [[V0]], [[V1:%.*]]
; CHECK-NEXT: [[V1_ADJ_ADJ:%.*]] = add i8 [[V1_ADJ]], [[V1_ADJ]]
; CHECK-NEXT: [[C1:%.*]] = icmp eq i8 [[V1_ADJ_ADJ]], 0
-; CHECK-NEXT: br i1 [[C1]], label [[FINAL_LEFT:%.*]], label [[FINAL_RIGHT]]
+; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[C0]], [[C1]]
+; CHECK-NEXT: br i1 [[OR_COND]], label [[FINAL_LEFT:%.*]], label [[FINAL_RIGHT:%.*]]
; CHECK: final_left:
; CHECK-NEXT: call void @sideeffect0()
; CHECK-NEXT: ret void
@@ -196,15 +195,18 @@ define void @two_preds_with_extra_op_multiuse(i8 %v0, i8 %v1, i8 %v2, i8 %v3) {
; CHECK-NEXT: br i1 [[C0]], label [[PRED0:%.*]], label [[PRED1:%.*]]
; CHECK: pred0:
; CHECK-NEXT: [[C1:%.*]] = icmp eq i8 [[V1:%.*]], 0
-; CHECK-NEXT: br i1 [[C1]], label [[FINAL_LEFT:%.*]], label [[DISPATCH:%.*]]
+; CHECK-NEXT: [[DOTOLD:%.*]] = add i8 [[V1]], [[V2:%.*]]
+; CHECK-NEXT: [[DOTOLD1:%.*]] = add i8 [[DOTOLD]], [[DOTOLD]]
+; CHECK-NEXT: [[C3_OLD:%.*]] = icmp eq i8 [[DOTOLD1]], 0
+; CHECK-NEXT: [[OR_COND4:%.*]] = or i1 [[C1]], [[C3_OLD]]
+; CHECK-NEXT: br i1 [[OR_COND4]], label [[FINAL_LEFT:%.*]], label [[FINAL_RIGHT:%.*]]
; CHECK: pred1:
-; CHECK-NEXT: [[C2:%.*]] = icmp eq i8 [[V2:%.*]], 0
-; CHECK-NEXT: br i1 [[C2]], label [[DISPATCH]], label [[FINAL_RIGHT:%.*]]
-; CHECK: dispatch:
+; CHECK-NEXT: [[C2:%.*]] = icmp eq i8 [[V2]], 0
; CHECK-NEXT: [[V3_ADJ:%.*]] = add i8 [[V1]], [[V2]]
; CHECK-NEXT: [[V3_ADJ_ADJ:%.*]] = add i8 [[V3_ADJ]], [[V3_ADJ]]
; CHECK-NEXT: [[C3:%.*]] = icmp eq i8 [[V3_ADJ_ADJ]], 0
-; CHECK-NEXT: br i1 [[C3]], label [[FINAL_LEFT]], label [[FINAL_RIGHT]]
+; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[C2]], [[C3]]
+; CHECK-NEXT: br i1 [[OR_COND]], label [[FINAL_LEFT]], label [[FINAL_RIGHT]]
; CHECK: final_left:
; CHECK-NEXT: call void @sideeffect0()
; CHECK-NEXT: ret void
@@ -241,11 +243,10 @@ define void @one_pred_with_extra_op_liveout(i8 %v0, i8 %v1) {
; CHECK-LABEL: @one_pred_with_extra_op_liveout(
; CHECK-NEXT: pred:
; CHECK-NEXT: [[C0:%.*]] = icmp eq i8 [[V0:%.*]], 0
-; CHECK-NEXT: br i1 [[C0]], label [[DISPATCH:%.*]], label [[FINAL_RIGHT:%.*]]
-; CHECK: dispatch:
; CHECK-NEXT: [[V1_ADJ:%.*]] = add i8 [[V0]], [[V1:%.*]]
; CHECK-NEXT: [[C1:%.*]] = icmp eq i8 [[V1_ADJ]], 0
-; CHECK-NEXT: br i1 [[C1]], label [[FINAL_LEFT:%.*]], label [[FINAL_RIGHT]]
+; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[C0]], [[C1]]
+; CHECK-NEXT: br i1 [[OR_COND]], label [[FINAL_LEFT:%.*]], label [[FINAL_RIGHT:%.*]]
; CHECK: final_left:
; CHECK-NEXT: call void @sideeffect0()
; CHECK-NEXT: call void @use8(i8 [[V1_ADJ]])
@@ -273,11 +274,10 @@ define void @one_pred_with_extra_op_liveout_multiuse(i8 %v0, i8 %v1) {
; CHECK-LABEL: @one_pred_with_extra_op_liveout_multiuse(
; CHECK-NEXT: pred:
; CHECK-NEXT: [[C0:%.*]] = icmp eq i8 [[V0:%.*]], 0
-; CHECK-NEXT: br i1 [[C0]], label [[DISPATCH:%.*]], label [[FINAL_RIGHT:%.*]]
-; CHECK: dispatch:
; CHECK-NEXT: [[V1_ADJ:%.*]] = add i8 [[V0]], [[V1:%.*]]
; CHECK-NEXT: [[C1:%.*]] = icmp eq i8 [[V1_ADJ]], 0
-; CHECK-NEXT: br i1 [[C1]], label [[FINAL_LEFT:%.*]], label [[FINAL_RIGHT]]
+; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[C0]], [[C1]]
+; CHECK-NEXT: br i1 [[OR_COND]], label [[FINAL_LEFT:%.*]], label [[FINAL_RIGHT:%.*]]
; CHECK: final_left:
; CHECK-NEXT: call void @sideeffect0()
; CHECK-NEXT: call void @use8(i8 [[V1_ADJ]])
@@ -314,13 +314,16 @@ define void @two_preds_with_extra_op_liveout(i8 %v0, i8 %v1, i8 %v2, i8 %v3) {
; CHECK-NEXT: br i1 [[C1]], label [[FINAL_LEFT:%.*]], label [[DISPATCH:%.*]]
; CHECK: pred1:
; CHECK-NEXT: [[C2:%.*]] = icmp eq i8 [[V2:%.*]], 0
-; CHECK-NEXT: br i1 [[C2]], label [[DISPATCH]], label [[FINAL_RIGHT:%.*]]
-; CHECK: dispatch:
; CHECK-NEXT: [[V3_ADJ:%.*]] = add i8 [[V1]], [[V2]]
; CHECK-NEXT: [[C3:%.*]] = icmp eq i8 [[V3_ADJ]], 0
-; CHECK-NEXT: br i1 [[C3]], label [[FINAL_LEFT]], label [[FINAL_RIGHT]]
+; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[C2]], [[C3]]
+; CHECK-NEXT: br i1 [[OR_COND]], label [[FINAL_LEFT]], label [[FINAL_RIGHT:%.*]]
+; CHECK: dispatch:
+; CHECK-NEXT: [[DOTOLD:%.*]] = add i8 [[V1]], [[V2]]
+; CHECK-NEXT: [[C3_OLD:%.*]] = icmp eq i8 [[DOTOLD]], 0
+; CHECK-NEXT: br i1 [[C3_OLD]], label [[FINAL_LEFT]], label [[FINAL_RIGHT]]
; CHECK: final_left:
-; CHECK-NEXT: [[MERGE_LEFT:%.*]] = phi i8 [ [[V3_ADJ]], [[DISPATCH]] ], [ 0, [[PRED0]] ]
+; CHECK-NEXT: [[MERGE_LEFT:%.*]] = phi i8 [ [[DOTOLD]], [[DISPATCH]] ], [ 0, [[PRED0]] ], [ [[V3_ADJ]], [[PRED1]] ]
; CHECK-NEXT: call void @use8(i8 [[MERGE_LEFT]])
; CHECK-NEXT: call void @sideeffect0()
; CHECK-NEXT: ret void
@@ -361,14 +364,17 @@ define void @two_preds_with_extra_op_liveout_multiuse(i8 %v0, i8 %v1, i8 %v2, i8
; CHECK-NEXT: br i1 [[C1]], label [[FINAL_LEFT:%.*]], label [[DISPATCH:%.*]]
; CHECK: pred1:
; CHECK-NEXT: [[C2:%.*]] = icmp eq i8 [[V2:%.*]], 0
-; CHECK-NEXT: br i1 [[C2]], label [[DISPATCH]], label [[FINAL_RIGHT:%.*]]
-; CHECK: dispatch:
; CHECK-NEXT: [[V3_ADJ:%.*]] = add i8 [[V1]], [[V2]]
; CHECK-NEXT: [[C3:%.*]] = icmp eq i8 [[V3_ADJ]], 0
-; CHECK-NEXT: br i1 [[C3]], label [[FINAL_LEFT]], label [[FINAL_RIGHT]]
+; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[C2]], [[C3]]
+; CHECK-NEXT: br i1 [[OR_COND]], label [[FINAL_LEFT]], label [[FINAL_RIGHT:%.*]]
+; CHECK: dispatch:
+; CHECK-NEXT: [[DOTOLD:%.*]] = add i8 [[V1]], [[V2]]
+; CHECK-NEXT: [[C3_OLD:%.*]] = icmp eq i8 [[DOTOLD]], 0
+; CHECK-NEXT: br i1 [[C3_OLD]], label [[FINAL_LEFT]], label [[FINAL_RIGHT]]
; CHECK: final_left:
-; CHECK-NEXT: [[MERGE_LEFT:%.*]] = phi i8 [ [[V3_ADJ]], [[DISPATCH]] ], [ 0, [[PRED0]] ]
-; CHECK-NEXT: [[MERGE_LEFT_2:%.*]] = phi i8 [ [[V3_ADJ]], [[DISPATCH]] ], [ 42, [[PRED0]] ]
+; CHECK-NEXT: [[MERGE_LEFT:%.*]] = phi i8 [ [[DOTOLD]], [[DISPATCH]] ], [ 0, [[PRED0]] ], [ [[V3_ADJ]], [[PRED1]] ]
+; CHECK-NEXT: [[MERGE_LEFT_2:%.*]] = phi i8 [ [[DOTOLD]], [[DISPATCH]] ], [ 42, [[PRED0]] ], [ [[V3_ADJ]], [[PRED1]] ]
; CHECK-NEXT: call void @use8(i8 [[MERGE_LEFT]])
; CHECK-NEXT: call void @use8(i8 [[MERGE_LEFT_2]])
; CHECK-NEXT: call void @sideeffect0()
@@ -409,11 +415,10 @@ define void @one_pred_with_extra_op_eexternally_used_only(i8 %v0, i8 %v1) {
; CHECK-LABEL: @one_pred_with_extra_op_eexternally_used_only(
; CHECK-NEXT: pred:
; CHECK-NEXT: [[C0:%.*]] = icmp eq i8 [[V0:%.*]], 0
-; CHECK-NEXT: br i1 [[C0]], label [[DISPATCH:%.*]], label [[FINAL_RIGHT:%.*]]
-; CHECK: dispatch:
; CHECK-NEXT: [[V1_ADJ:%.*]] = add i8 [[V0]], [[V1:%.*]]
; CHECK-NEXT: [[C1:%.*]] = icmp eq i8 [[V1]], 0
-; CHECK-NEXT: br i1 [[C1]], label [[FINAL_LEFT:%.*]], label [[FINAL_RIGHT]]
+; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[C0]], [[C1]]
+; CHECK-NEXT: br i1 [[OR_COND]], label [[FINAL_LEFT:%.*]], label [[FINAL_RIGHT:%.*]]
; CHECK: final_left:
; CHECK-NEXT: call void @sideeffect0()
; CHECK-NEXT: call void @use8(i8 [[V1_ADJ]])
@@ -441,11 +446,10 @@ define void @one_pred_with_extra_op_externally_used_only_multiuse(i8 %v0, i8 %v1
; CHECK-LABEL: @one_pred_with_extra_op_externally_used_only_multiuse(
; CHECK-NEXT: pred:
; CHECK-NEXT: [[C0:%.*]] = icmp eq i8 [[V0:%.*]], 0
-; CHECK-NEXT: br i1 [[C0]], label [[DISPATCH:%.*]], label [[FINAL_RIGHT:%.*]]
-; CHECK: dispatch:
; CHECK-NEXT: [[V1_ADJ:%.*]] = add i8 [[V0]], [[V1:%.*]]
; CHECK-NEXT: [[C1:%.*]] = icmp eq i8 [[V1]], 0
-; CHECK-NEXT: br i1 [[C1]], label [[FINAL_LEFT:%.*]], label [[FINAL_RIGHT]]
+; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[C0]], [[C1]]
+; CHECK-NEXT: br i1 [[OR_COND]], label [[FINAL_LEFT:%.*]], label [[FINAL_RIGHT:%.*]]
; CHECK: final_left:
; CHECK-NEXT: call void @sideeffect0()
; CHECK-NEXT: call void @use8(i8 [[V1_ADJ]])
@@ -482,13 +486,16 @@ define void @two_preds_with_extra_op_externally_used_only(i8 %v0, i8 %v1, i8 %v2
; CHECK-NEXT: br i1 [[C1]], label [[FINAL_LEFT:%.*]], label [[DISPATCH:%.*]]
; CHECK: pred1:
; CHECK-NEXT: [[C2:%.*]] = icmp eq i8 [[V2:%.*]], 0
-; CHECK-NEXT: br i1 [[C2]], label [[DISPATCH]], label [[FINAL_RIGHT:%.*]]
-; CHECK: dispatch:
; CHECK-NEXT: [[V3_ADJ:%.*]] = add i8 [[V1]], [[V2]]
; CHECK-NEXT: [[C3:%.*]] = icmp eq i8 [[V3:%.*]], 0
-; CHECK-NEXT: br i1 [[C3]], label [[FINAL_LEFT]], label [[FINAL_RIGHT]]
+; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[C2]], [[C3]]
+; CHECK-NEXT: br i1 [[OR_COND]], label [[FINAL_LEFT]], label [[FINAL_RIGHT:%.*]]
+; CHECK: dispatch:
+; CHECK-NEXT: [[DOTOLD:%.*]] = add i8 [[V1]], [[V2]]
+; CHECK-NEXT: [[C3_OLD:%.*]] = icmp eq i8 [[V3]], 0
+; CHECK-NEXT: br i1 [[C3_OLD]], label [[FINAL_LEFT]], label [[FINAL_RIGHT]]
; CHECK: final_left:
-; CHECK-NEXT: [[MERGE_LEFT:%.*]] = phi i8 [ [[V3_ADJ]], [[DISPATCH]] ], [ 0, [[PRED0]] ]
+; CHECK-NEXT: [[MERGE_LEFT:%.*]] = phi i8 [ [[DOTOLD]], [[DISPATCH]] ], [ 0, [[PRED0]] ], [ [[V3_ADJ]], [[PRED1]] ]
; CHECK-NEXT: call void @use8(i8 [[MERGE_LEFT]])
; CHECK-NEXT: call void @sideeffect0()
; CHECK-NEXT: ret void
@@ -529,14 +536,17 @@ define void @two_preds_with_extra_op_externally_used_only_multiuse(i8 %v0, i8 %v
; CHECK-NEXT: br i1 [[C1]], label [[FINAL_LEFT:%.*]], label [[DISPATCH:%.*]]
; CHECK: pred1:
; CHECK-NEXT: [[C2:%.*]] = icmp eq i8 [[V2:%.*]], 0
-; CHECK-NEXT: br i1 [[C2]], label [[DISPATCH]], label [[FINAL_RIGHT:%.*]]
-; CHECK: dispatch:
; CHECK-NEXT: [[V3_ADJ:%.*]] = add i8 [[V1]], [[V2]]
; CHECK-NEXT: [[C3:%.*]] = icmp eq i8 [[V3:%.*]], 0
-; CHECK-NEXT: br i1 [[C3]], label [[FINAL_LEFT]], label [[FINAL_RIGHT]]
+; CHECK-NEXT: [[OR_COND:%.*]] = and i1 [[C2]], [[C3]]
+; CHECK-NEXT: br i1 [[OR_COND]], label [[FINAL_LEFT]], label [[FINAL_RIGHT:%.*]]
+; CHECK: dispatch:
+; CHECK-NEXT: [[DOTOLD:%.*]] = add i8 [[V1]], [[V2]]
+; CHECK-NEXT: [[C3_OLD:%.*]] = icmp eq i8 [[V3]], 0
+; CHECK-NEXT: br i1 [[C3_OLD]], label [[FINAL_LEFT]], label [[FINAL_RIGHT]]
; CHECK: final_left:
-; CHECK-NEXT: [[MERGE_LEFT:%.*]] = phi i8 [ [[V3_ADJ]], [[DISPATCH]] ], [ 0, [[PRED0]] ]
-; CHECK-NEXT: [[MERGE_LEFT_2:%.*]] = phi i8 [ [[V3_ADJ]], [[DISPATCH]] ], [ 42, [[PRED0]] ]
+; CHECK-NEXT: [[MERGE_LEFT:%.*]] = phi i8 [ [[DOTOLD]], [[DISPATCH]] ], [ 0, [[PRED0]] ], [ [[V3_ADJ]], [[PRED1]] ]
+; CHECK-NEXT: [[MERGE_LEFT_2:%.*]] = phi i8 [ [[DOTOLD]], [[DISPATCH]] ], [ 42, [[PRED0]] ], [ [[V3_ADJ]], [[PRED1]] ]
; CHECK-NEXT: call void @use8(i8 [[MERGE_LEFT]])
; CHECK-NEXT: call void @use8(i8 [[MERGE_LEFT_2]])
; CHECK-NEXT: call void @sideeffect0()
More information about the llvm-commits
mailing list