[MemorySSA] Cleanup trivial phis.

Summary:
This is unfortunately needed for correctness, if we are to extend the tolerance of the update API to the way simple loop unswitch is doing cloning.

In simple loop unswitch (as opposed to loop unswitch), not all blocks are cloned. This can create unreachable cloned blocks (no predecessor), which are later cleaned up.

In MemorySSA, the  APIs for supporting these kind of updates (clone + update exit blocks), make certain assumption on the integrity of the CFG. When cloning, if something was not cloned, it's values in MemorySSA default to LiveOnEntry. When updating exit blocks, it is safe to assume that we can first insert phis in the blocks merging two clones, then add additional phis in the IDF of the blocks that received phis. This no longer holds true if one of the clones being merged comes from an unreachable block. We'd conservatively need to add all phis before filling in their incoming definitions. In practice this restriction can be relaxed if we clean up trivial phis after the first round of insertion.

Reviewers: george.burgess.iv

Subscribers: jlebar, Prazek, llvm-commits

Tags: #llvm

Differential Revision: https://reviews.llvm.org/D63354

llvm-svn: 363880
This commit is contained in:
Alina Sbirlea 2019-06-19 21:33:09 +00:00
parent 238b8e62b6
commit 109d2ea153
2 changed files with 121 additions and 5 deletions

View File

@ -866,7 +866,6 @@ void MemorySSAUpdater::applyInsertUpdates(ArrayRef<CFGUpdate> Updates,
for (auto *BB : NewBlocks)
PredMap.erase(BB);
SmallVector<BasicBlock *, 8> BlocksToProcess;
SmallVector<BasicBlock *, 16> BlocksWithDefsToReplace;
SmallVector<WeakVH, 8> InsertedPhis;
@ -942,10 +941,6 @@ void MemorySSAUpdater::applyInsertUpdates(ArrayRef<CFGUpdate> Updates,
for (auto *Pred : PrevBlockSet)
for (int I = 0, E = EdgeCountMap[{Pred, BB}]; I < E; ++I)
NewPhi->addIncoming(DefP1, Pred);
// Insert BB in the set of blocks that now have definition. We'll use this
// to compute IDF and add Phis there next.
BlocksToProcess.push_back(BB);
}
// Get all blocks that used to dominate BB and no longer do after adding
@ -960,6 +955,14 @@ void MemorySSAUpdater::applyInsertUpdates(ArrayRef<CFGUpdate> Updates,
GetNoLongerDomBlocks(PrevIDom, NewIDom, BlocksWithDefsToReplace);
}
tryRemoveTrivialPhis(InsertedPhis);
// Create the set of blocks that now have a definition. We'll use this to
// compute IDF and add Phis there next.
SmallVector<BasicBlock *, 8> BlocksToProcess;
for (auto &VH : InsertedPhis)
if (auto *MPhi = cast_or_null<MemoryPhi>(VH))
BlocksToProcess.push_back(MPhi->getBlock());
// Compute IDF and add Phis in all IDF blocks that do not have one.
SmallVector<BasicBlock *, 32> IDFBlocks;
if (!BlocksToProcess.empty()) {

View File

@ -0,0 +1,113 @@
; RUN: opt -simple-loop-unswitch -enable-nontrivial-unswitch -verify-memoryssa -S %s | FileCheck %s
; REQUIRES: asserts
target triple = "x86_64-unknown-linux-gnu"
declare void @foo()
; In Test1, there are no definitions. MemorySSA updates insert trivial phis and remove them.
; Verify all are removed, considering the SLU pass leaves unreachable blocks hanging when the MSSA updates are done.
; CHECK-LABEL: @Test1
define void @Test1(i32) {
header:
br label %outer
outer.loopexit.split: ; preds = %continue
br label %outer.loopexit
outer.loopexit: ; preds = %outer.loopexit.split.us, %outer.loopexit.split
br label %outer
outer: ; preds = %outer.loopexit, %header
br i1 false, label %outer.split.us, label %outer.split
outer.split.us: ; preds = %outer
br label %inner.us
inner.us: ; preds = %continue.us, %outer.split.us
br label %overflow.us
overflow.us: ; preds = %inner.us
br label %continue.us
continue.us: ; preds = %overflow.us
br i1 true, label %outer.loopexit.split.us, label %inner.us
outer.loopexit.split.us: ; preds = %continue.us
br label %outer.loopexit
outer.split: ; preds = %outer
br label %inner
inner: ; preds = %continue, %outer.split
br label %switchme
switchme: ; preds = %inner
switch i32 %0, label %continue [
i32 88, label %go_out
i32 99, label %case2
]
case2: ; preds = %switchme
br label %continue
continue: ; preds = %case2, %switchme
br i1 true, label %outer.loopexit.split, label %inner
go_out: ; preds = %switchme
unreachable
}
; In Test2 there is a single def (call to foo). There are already Phis in place that are cloned when unswitching.
; Ensure MemorySSA remains correct. Due to SLU's pruned cloning, continue.us2 becomes unreachable, with an empty Phi that is later cleaned.
; CHECK-LABEL: @Test2
define void @Test2(i32) {
header:
br label %outer
outer.loopexit.split: ; preds = %continue
br label %outer.loopexit
outer.loopexit: ; preds = %outer.loopexit.split.us, %outer.loopexit.split
br label %outer
outer: ; preds = %outer.loopexit, %header
br i1 false, label %outer.split.us, label %outer.split
outer.split.us: ; preds = %outer
br label %inner.us
inner.us: ; preds = %continue.us, %outer.split.us
br label %overflow.us
overflow.us: ; preds = %inner.us
br label %continue.us
continue.us: ; preds = %overflow.us
br i1 true, label %outer.loopexit.split.us, label %inner.us
outer.loopexit.split.us: ; preds = %continue.us
br label %outer.loopexit
outer.split: ; preds = %outer
br label %inner
inner: ; preds = %continue, %outer.split
br label %switchme
switchme: ; preds = %inner
switch i32 %0, label %continue [
i32 88, label %go_out
i32 99, label %case2
]
case2: ; preds = %switchme
call void @foo()
br label %continue
continue: ; preds = %case2, %switchme
br i1 true, label %outer.loopexit.split, label %inner
go_out: ; preds = %switchme
unreachable
}