[Patch] GVN fold conditional-branch on-the-fly
Shuxin Yang
shuxin.llvm at gmail.com
Wed Sep 18 10:14:45 PDT 2013
Hi, Daniel:
With the helper of Andy and Nadav, I manage to come up a much simpler
fix.
It no longer delete dead code on the fly, it just proceeds in the presence
of dead code.
It is no longer complex both in terms of concept and LOC. In terms of
LOC,
it add 157 LOC with plenty of comments. In concept, it is almost
"no-brainer",
so to speak :-).
How does this patch look to you?
Thanks
Shxuin
How it work
=======================
Here is baiscially how dead code are "ignored".
1) When a dead branch target, say block B, is identified, all the
blocks dominated by B is dead as well.
2) The phis of those blocks in dominance-frontier(B) is updated such
that the operands corresponding to dead predecessors are replaced
by "UndefVal".
Using lattice's jargon, the "UndefVal" is the "Top" in essence.
Phi node like this "phi(v1 bb1, undef xx)" will be optimized into
"v1" if v1 is constant, or v1 is an instruction which dominate this
phi node.
3) When analyzing the availability of a load L, all dead mem-ops which
L depends on disguise as a load which evaluate exactly same value
as L.
4) The dead mem-ops will be materialized as "UndefVal" during code
motion.
example
=======
Consider the e.g slightly generalized from the original motivating eg.
Fig 1
-----------------------------------------------------
BB1 :
= *p; // s1
if (false) {
BB-then-dead:
puts("Shuxin is a pretty nice guy!lol"). // s2, alias with *p;
} else {
BB-else:
...
}
BB-junk:
another if-then-else construct.
BB-last:
= *p; // s3;
-----------------------------------------------------
The load in S3 deps on s1 and s2; the later is dead. Now that s2
alias with s3,
due to the limit the Ld-PRE, it gives up.
The step-3) will modify the dep-info about s2 to disguise as a load
producing the same value as s3, and hence enable the transformation.
Fig 2:
------------------
BB1 :
v1 = *p; // s1
if (false) {
BB-then-dead:
puts("Shuxin is a pretty nice guy!lol"). // s2, alias with *p;
} else {
BB-else:
...
}
BB-junk:
v3 = phi(v1 at BB-else, undef at BB-then-dead)
another if-then-else construct.
BB-last:
= v3; // s3;
------------------
LD-PRE Limitation
=================
With a closer look of the original problem. I realize GVN fail to
eliminate
the redundant load due to the limit the of LD-PRE.
A typical whatever-PRE would pick up the "observation point" at the place
available and unavailble path just meet (in this case, it would be the entry
of BB-junk). However, GVN LD-PRE choose the point right before the load in
question. In this case, it would be the entry point of BB-last, from that
point of view, the load is unavilable in all the incoming pred. Hence
give up.
That said, I don't want to blame GVN as choosing right "observation
point"
entails expensive data flow analysis. However, llvm users are pretty picky
at compile time.
The limitation of this approach
===============================
The limitation comes from the fact that we are not able to simplify
"v2 = phi(v1, undef)" into v if v neither dominate this phi nor is
constant. By it machinary, val-num(v2) != val-num(v1) and hence obscure
some opportunities.
-------------- next part --------------
Index: include/llvm/Analysis/Dominators.h
===================================================================
--- include/llvm/Analysis/Dominators.h (revision 190918)
+++ include/llvm/Analysis/Dominators.h (working copy)
@@ -346,6 +346,20 @@
DomTreeNodeBase<NodeT> *getRootNode() { return RootNode; }
const DomTreeNodeBase<NodeT> *getRootNode() const { return RootNode; }
+ /// Get all nodes dominated by R, including R itself. Return true on success.
+ void getDescendants(NodeT *R, SmallVectorImpl<NodeT *> &Result) const {
+ const DomTreeNodeBase<NodeT> *RN = getNode(R);
+ SmallVector<const DomTreeNodeBase<NodeT> *, 8> WL;
+ WL.push_back(RN);
+ Result.clear();
+
+ while (!WL.empty()) {
+ const DomTreeNodeBase<NodeT> *N = WL.pop_back_val();
+ Result.push_back(N->getBlock());
+ WL.append(N->begin(), N->end());
+ }
+ }
+
/// properlyDominates - Returns true iff A dominates B and A != B.
/// Note that this is not a constant time operation!
///
@@ -755,6 +769,12 @@
return DT->getRootNode();
}
+ /// Get all nodes dominated by R, including R itself. Return true on success.
+ void getDescendants(BasicBlock *R,
+ SmallVectorImpl<BasicBlock *> &Result) const {
+ DT->getDescendants(R, Result);
+ }
+
/// compare - Return false if the other dominator tree matches this
/// dominator tree. Otherwise return true.
inline bool compare(DominatorTree &Other) const {
Index: lib/Transforms/Scalar/GVN.cpp
===================================================================
--- lib/Transforms/Scalar/GVN.cpp (revision 190918)
+++ lib/Transforms/Scalar/GVN.cpp (working copy)
@@ -21,6 +21,7 @@
#include "llvm/ADT/DepthFirstIterator.h"
#include "llvm/ADT/Hashing.h"
#include "llvm/ADT/SmallPtrSet.h"
+#include "llvm/ADT/SetVector.h"
#include "llvm/ADT/Statistic.h"
#include "llvm/Analysis/AliasAnalysis.h"
#include "llvm/Analysis/CFG.h"
@@ -507,7 +508,9 @@
enum ValType {
SimpleVal, // A simple offsetted value that is accessed.
LoadVal, // A value produced by a load.
- MemIntrin // A memory intrinsic which is loaded from.
+ MemIntrin, // A memory intrinsic which is loaded from.
+ UndefVal // A UndefValue representing a value from dead block (which
+ // is not yet physically removed from the CFG).
};
/// V - The value that is live out of the block.
@@ -545,10 +548,20 @@
Res.Offset = Offset;
return Res;
}
-
+
+ static AvailableValueInBlock getUndef(BasicBlock *BB) {
+ AvailableValueInBlock Res;
+ Res.BB = BB;
+ Res.Val.setPointer(0);
+ Res.Val.setInt(UndefVal);
+ Res.Offset = 0;
+ return Res;
+ }
+
bool isSimpleValue() const { return Val.getInt() == SimpleVal; }
bool isCoercedLoadValue() const { return Val.getInt() == LoadVal; }
bool isMemIntrinValue() const { return Val.getInt() == MemIntrin; }
+ bool isUndefValue() const { return Val.getInt() == UndefVal; }
Value *getSimpleValue() const {
assert(isSimpleValue() && "Wrong accessor");
@@ -576,6 +589,7 @@
DominatorTree *DT;
const DataLayout *TD;
const TargetLibraryInfo *TLI;
+ SetVector<BasicBlock *> DeadBlocks;
ValueTable VN;
@@ -698,6 +712,9 @@
unsigned replaceAllDominatedUsesWith(Value *From, Value *To,
const BasicBlockEdge &Root);
bool propagateEquality(Value *LHS, Value *RHS, const BasicBlockEdge &Root);
+ bool processFoldableCondBr(BranchInst *BI);
+ void addDeadBlock(BasicBlock *BB);
+ void assignValNumForDeadCode();
};
char GVN::ID = 0;
@@ -1253,8 +1270,10 @@
// just use the dominating value directly.
if (ValuesPerBlock.size() == 1 &&
gvn.getDominatorTree().properlyDominates(ValuesPerBlock[0].BB,
- LI->getParent()))
+ LI->getParent())) {
+ assert(!ValuesPerBlock[0].isUndefValue() && "Dead BB dominate this block");
return ValuesPerBlock[0].MaterializeAdjustedValue(LI->getType(), gvn);
+ }
// Otherwise, we have to construct SSA form.
SmallVector<PHINode*, 8> NewPHIs;
@@ -1324,7 +1343,7 @@
<< *getCoercedLoadValue() << '\n'
<< *Res << '\n' << "\n\n\n");
}
- } else {
+ } else if (isMemIntrinValue()) {
const DataLayout *TD = gvn.getDataLayout();
assert(TD && "Need target data to handle type mismatch case");
Res = GetMemInstValueForLoad(getMemIntrinValue(), Offset,
@@ -1332,6 +1351,10 @@
DEBUG(dbgs() << "GVN COERCED NONLOCAL MEM INTRIN:\nOffset: " << Offset
<< " " << *getMemIntrinValue() << '\n'
<< *Res << '\n' << "\n\n\n");
+ } else {
+ assert(isUndefValue() && "Should be UndefVal");
+ DEBUG(dbgs() << "GVN COERCED NONLOCAL Undef:\n";);
+ return UndefValue::get(LoadTy);
}
return Res;
}
@@ -1355,6 +1378,13 @@
BasicBlock *DepBB = Deps[i].getBB();
MemDepResult DepInfo = Deps[i].getResult();
+ if (DeadBlocks.count(DepBB)) {
+ // Dead dependent mem-op disguise as a load evaluating the same value
+ // as the load in question.
+ ValuesPerBlock.push_back(AvailableValueInBlock::getUndef(DepBB));
+ continue;
+ }
+
if (!DepInfo.isDef() && !DepInfo.isClobber()) {
UnavailableBlocks.push_back(DepBB);
continue;
@@ -2191,11 +2221,13 @@
// For conditional branches, we can perform simple conditional propagation on
// the condition value itself.
if (BranchInst *BI = dyn_cast<BranchInst>(I)) {
- if (!BI->isConditional() || isa<Constant>(BI->getCondition()))
+ if (!BI->isConditional())
return false;
+ if (isa<Constant>(BI->getCondition()))
+ return processFoldableCondBr(BI);
+
Value *BranchCond = BI->getCondition();
-
BasicBlock *TrueSucc = BI->getSuccessor(0);
BasicBlock *FalseSucc = BI->getSuccessor(1);
// Avoid multiple edges early.
@@ -2312,6 +2344,9 @@
}
if (EnablePRE) {
+ // Fabricate val-num for dead-code in order to suppress assertion in
+ // performPRE().
+ assignValNumForDeadCode();
bool PREChanged = true;
while (PREChanged) {
PREChanged = performPRE(F);
@@ -2325,6 +2360,9 @@
// Actually, when this happens, we should just fully integrate PRE into GVN.
cleanupGlobalSets();
+ // Do not cleanup DeadBlocks in cleanupGlobalSets() as it's called for each
+ // iteration.
+ DeadBlocks.clear();
return Changed;
}
@@ -2335,6 +2373,9 @@
// (and incrementing BI before processing an instruction).
assert(InstrsToErase.empty() &&
"We expect InstrsToErase to be empty across iterations");
+ if (DeadBlocks.count(BB))
+ return false;
+
bool ChangedFunction = false;
for (BasicBlock::iterator BI = BB->begin(), BE = BB->end();
@@ -2628,3 +2669,120 @@
}
}
}
+
+// BB is declared dead, which implied other blocks become dead as well. This
+// function is to add all these blocks to "DeadBlocks". For the dead blocks'
+// live successors, update their phi nodes by replacing the operands
+// corresponding to dead blocks with UndefVal.
+//
+void GVN::addDeadBlock(BasicBlock *BB) {
+ SmallVector<BasicBlock *, 4> NewDead;
+ SmallSetVector<BasicBlock *, 4> DF;
+
+ NewDead.push_back(BB);
+ while (!NewDead.empty()) {
+ BasicBlock *D = NewDead.pop_back_val();
+ if (DeadBlocks.count(D))
+ continue;
+
+ // All blocks dominated by D are dead.
+ SmallVector<BasicBlock *, 8> Dom;
+ DT->getDescendants(D, Dom);
+ DeadBlocks.insert(Dom.begin(), Dom.end());
+
+ // Figure out the dominance-frontier(D).
+ for (SmallVectorImpl<BasicBlock *>::iterator I = Dom.begin(),
+ E = Dom.end(); I != E; I++) {
+ BasicBlock *B = *I;
+ for (succ_iterator SI = succ_begin(B), SE = succ_end(B); SI != SE; SI++) {
+ BasicBlock *S = *SI;
+ if (DeadBlocks.count(S))
+ continue;
+
+ bool AllPredDead = true;
+ for (pred_iterator PI = pred_begin(S), PE = pred_end(S); PI != PE; PI++)
+ if (!DeadBlocks.count(*PI)) {
+ AllPredDead = false;
+ break;
+ }
+
+ if (!AllPredDead) {
+ // S could be proved dead later on. That is why we don't update phi
+ // operands at this moment.
+ DF.insert(S);
+ } else {
+ // While S is not dominated by D, it is dead by now. This could take
+ // place if S already have a dead predecessor before D is declared
+ // dead.
+ NewDead.push_back(S);
+ }
+ }
+ }
+ }
+
+ // For the dead blocks' live successors, update their phi nodes by replacing
+ // the operands corresponding to dead blocks with UndefVal.
+ for(SmallSetVector<BasicBlock *, 4>::iterator I = DF.begin(), E = DF.end();
+ I != E; I++) {
+ BasicBlock *B = *I;
+ if (DeadBlocks.count(B))
+ continue;
+
+ for (pred_iterator PI = pred_begin(B), PE = pred_end(B); PI != PE; PI++) {
+ BasicBlock *P = *PI;
+ if (!DeadBlocks.count(P))
+ continue;
+ for (BasicBlock::iterator II = B->begin(); isa<PHINode>(II); ++II) {
+ PHINode &Phi = cast<PHINode>(*II);
+ Phi.setIncomingValue(Phi.getBasicBlockIndex(P),
+ UndefValue::get(Phi.getType()));
+ }
+ }
+ }
+}
+
+// If the given branch is recognized as a foldable branch (i.e. conditional
+// branch with constant condition), it will perform following analyses and
+// transformation.
+// 1) If the dead out-coming edge is a critical-edge, split it. Let
+// R be the target of the dead out-coming edge.
+// 1) Identify the set of dead blocks implied by the branch's dead outcoming
+// edge. The result of this step will be {X| X is dominated by R}
+// 2) Identify those blocks which haves at least one dead prodecessor. The
+// result of this step will be dominance-frontier(R).
+// 3) Update the PHIs in DF(R) by replacing the operands corresponding to
+// dead blocks with "UndefVal" in an hope these PHIs will optimized away.
+//
+// Return true iff *NEW* dead code are found.
+bool GVN::processFoldableCondBr(BranchInst *BI) {
+ if (!BI || BI->isUnconditional())
+ return false;
+
+ ConstantInt *Cond = dyn_cast<ConstantInt>(BI->getCondition());
+ if (!Cond)
+ return false;
+
+ BasicBlock *DeadRoot = Cond->getZExtValue() ?
+ BI->getSuccessor(1) : BI->getSuccessor(0);
+ if (DeadBlocks.count(DeadRoot))
+ return false;
+
+ if (!DeadRoot->getSinglePredecessor())
+ DeadRoot = splitCriticalEdges(BI->getParent(), DeadRoot);
+
+ addDeadBlock(DeadRoot);
+ return true;
+}
+
+// performPRE() will trigger assert if it come across an instruciton without
+// associated val-num. As it normally has far more live instructions than dead
+// instructions, it makes more sense just to "fabricate" a val-number for the
+// dead code than checking if instruction involved is dead or not.
+void GVN::assignValNumForDeadCode() {
+ for (SetVector<BasicBlock *>::iterator I = DeadBlocks.begin(),
+ E = DeadBlocks.end(); I != E; I++) {
+ for (BasicBlock::iterator II = (*I)->begin(), EE = (*I)->end();
+ II != EE; II++)
+ VN.lookup_or_add(&*II);
+ }
+}
-------------- next part --------------
Index: test/Transforms/GVN/2007-07-26-InterlockingLoops.ll
===================================================================
--- test/Transforms/GVN/2007-07-26-InterlockingLoops.ll (revision 190918)
+++ test/Transforms/GVN/2007-07-26-InterlockingLoops.ll (working copy)
@@ -2,7 +2,7 @@
@last = external global [65 x i32*]
-define i32 @NextRootMove(i32 %wtm) {
+define i32 @NextRootMove(i32 %wtm, i32 %x, i32 %y, i32 %z) {
entry:
%A = alloca i32*
%tmp17618 = load i32** getelementptr ([65 x i32*]* @last, i32 0, i32 1), align 4
@@ -15,12 +15,14 @@
br label %cond_true116
cond_true116:
- br i1 false, label %cond_true128, label %cond_true145
+ %cmp = icmp eq i32 %x, %y
+ br i1 %cmp, label %cond_true128, label %cond_true145
cond_true128:
%tmp17625 = load i32** getelementptr ([65 x i32*]* @last, i32 0, i32 1), align 4
store i32* %tmp17625, i32** %A
- br i1 false, label %bb98.backedge, label %return.loopexit
+ %cmp1 = icmp eq i32 %x, %z
+ br i1 %cmp1 , label %bb98.backedge, label %return.loopexit
bb98.backedge:
br label %cond_true116
Index: test/Transforms/GVN/2008-07-02-Unreachable.ll
===================================================================
--- test/Transforms/GVN/2008-07-02-Unreachable.ll (revision 190918)
+++ test/Transforms/GVN/2008-07-02-Unreachable.ll (working copy)
@@ -3,10 +3,11 @@
@g_3 = external global i8 ; <i8*> [#uses=2]
-define i8 @func_1() nounwind {
+define i8 @func_1(i32 %x, i32 %y) nounwind {
entry:
%A = alloca i8
- br i1 false, label %ifelse, label %ifthen
+ %cmp = icmp eq i32 %x, %y
+ br i1 %cmp, label %ifelse, label %ifthen
ifthen: ; preds = %entry
br label %ifend
@@ -14,9 +15,6 @@
ifelse: ; preds = %entry
%tmp3 = load i8* @g_3 ; <i8> [#uses=0]
store i8 %tmp3, i8* %A
- br label %forcond.thread
-
-forcond.thread: ; preds = %ifelse
br label %afterfor
forcond: ; preds = %forinc
Index: test/Transforms/GVN/cond_br.ll
===================================================================
--- test/Transforms/GVN/cond_br.ll (revision 0)
+++ test/Transforms/GVN/cond_br.ll (working copy)
@@ -0,0 +1,55 @@
+; RUN: opt -basicaa -gvn -S < %s | FileCheck %s
+ at y = external global i32
+ at z = external global i32
+
+; Function Attrs: nounwind ssp uwtable
+define void @foo(i32 %x) {
+; CHECK: @foo(i32 %x)
+; CHECK: %.pre = load i32* @y
+; CHECK: call void @bar(i32 %.pre)
+
+ %t = sub i32 %x, %x
+ %.pre = load i32* @y, align 4
+ %cmp = icmp sgt i32 %t, 2
+ br i1 %cmp, label %if.then, label %entry.if.end_crit_edge
+
+entry.if.end_crit_edge: ; preds = %entry
+ br label %if.end
+
+if.then: ; preds = %entry
+ %add = add nsw i32 %x, 3
+ store i32 %add, i32* @y, align 4
+ br label %if.end
+
+if.end: ; preds = %entry.if.end_crit_edge, %if.then
+ %1 = phi i32 [ %.pre, %entry.if.end_crit_edge ], [ %add, %if.then ]
+ tail call void @bar(i32 %1)
+ ret void
+}
+
+define void @foo2(i32 %x) {
+; CHECK: @foo2(i32 %x)
+; CHECK: %.pre = load i32* @y
+; CHECK: tail call void @bar(i32 %.pre)
+entry:
+ %t = sub i32 %x, %x
+ %.pre = load i32* @y, align 4
+ %cmp = icmp sgt i32 %t, 2
+ br i1 %cmp, label %if.then, label %if.else
+
+if.then: ; preds = %entry
+ %add = add nsw i32 %x, 3
+ store i32 %add, i32* @y, align 4
+ br label %if.end
+
+if.else: ; preds = %entry
+ store i32 1, i32* @z, align 4
+ br label %if.end
+
+if.end: ; preds = %if.else, %if.then
+ %0 = phi i32 [ %.pre, %if.else ], [ %add, %if.then ]
+ tail call void @bar(i32 %0)
+ ret void
+}
+
+declare void @bar(i32)
Index: test/Transforms/GVN/cond_br2.ll
===================================================================
--- test/Transforms/GVN/cond_br2.ll (revision 0)
+++ test/Transforms/GVN/cond_br2.ll (working copy)
@@ -0,0 +1,139 @@
+; RUN: opt -basicaa -gvn -S < %s | FileCheck %s
+target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64-S128"
+
+%"class.llvm::SmallVector" = type { %"class.llvm::SmallVectorImpl", [1 x %"union.llvm::SmallVectorBase::U"] }
+%"class.llvm::SmallVectorImpl" = type { %"class.llvm::SmallVectorTemplateBase" }
+%"class.llvm::SmallVectorTemplateBase" = type { %"class.llvm::SmallVectorTemplateCommon" }
+%"class.llvm::SmallVectorTemplateCommon" = type { %"class.llvm::SmallVectorBase" }
+%"class.llvm::SmallVectorBase" = type { i8*, i8*, i8*, %"union.llvm::SmallVectorBase::U" }
+%"union.llvm::SmallVectorBase::U" = type { x86_fp80 }
+
+; Function Attrs: ssp uwtable
+define void @_Z4testv() #0 {
+; CHECK: @_Z4testv()
+; CHECK: invoke.cont:
+; CHECK: br i1 true, label %new.notnull.i11, label %if.end.i14
+; CHECK: Retry.i10:
+
+entry:
+ %sv = alloca %"class.llvm::SmallVector", align 16
+ %0 = bitcast %"class.llvm::SmallVector"* %sv to i8*
+ call void @llvm.lifetime.start(i64 64, i8* %0) #1
+ %BeginX.i.i.i.i.i.i = getelementptr inbounds %"class.llvm::SmallVector"* %sv, i64 0, i32 0, i32 0, i32 0, i32 0, i32 0
+ %FirstEl.i.i.i.i.i.i = getelementptr inbounds %"class.llvm::SmallVector"* %sv, i64 0, i32 0, i32 0, i32 0, i32 0, i32 3
+ %1 = bitcast %"union.llvm::SmallVectorBase::U"* %FirstEl.i.i.i.i.i.i to i8*
+ store i8* %1, i8** %BeginX.i.i.i.i.i.i, align 16, !tbaa !0
+ %EndX.i.i.i.i.i.i = getelementptr inbounds %"class.llvm::SmallVector"* %sv, i64 0, i32 0, i32 0, i32 0, i32 0, i32 1
+ store i8* %1, i8** %EndX.i.i.i.i.i.i, align 8, !tbaa !0
+ %CapacityX.i.i.i.i.i.i = getelementptr inbounds %"class.llvm::SmallVector"* %sv, i64 0, i32 0, i32 0, i32 0, i32 0, i32 2
+ %add.ptr.i.i.i.i2.i.i = getelementptr inbounds %"union.llvm::SmallVectorBase::U"* %FirstEl.i.i.i.i.i.i, i64 2
+ %add.ptr.i.i.i.i.i.i = bitcast %"union.llvm::SmallVectorBase::U"* %add.ptr.i.i.i.i2.i.i to i8*
+ store i8* %add.ptr.i.i.i.i.i.i, i8** %CapacityX.i.i.i.i.i.i, align 16, !tbaa !0
+ %EndX.i = getelementptr inbounds %"class.llvm::SmallVector"* %sv, i64 0, i32 0, i32 0, i32 0, i32 0, i32 1
+ %2 = load i8** %EndX.i, align 8, !tbaa !0
+ %CapacityX.i = getelementptr inbounds %"class.llvm::SmallVector"* %sv, i64 0, i32 0, i32 0, i32 0, i32 0, i32 2
+ %cmp.i = icmp ult i8* %2, %add.ptr.i.i.i.i.i.i
+ br i1 %cmp.i, label %Retry.i, label %if.end.i
+
+Retry.i: ; preds = %.noexc, %entry
+ %3 = phi i8* [ %2, %entry ], [ %.pre.i, %.noexc ]
+ %new.isnull.i = icmp eq i8* %3, null
+ br i1 %new.isnull.i, label %invoke.cont, label %new.notnull.i
+
+new.notnull.i: ; preds = %Retry.i
+ %4 = bitcast i8* %3 to i32*
+ store i32 1, i32* %4, align 4, !tbaa !3
+ br label %invoke.cont
+
+if.end.i: ; preds = %entry
+ %5 = getelementptr inbounds %"class.llvm::SmallVector"* %sv, i64 0, i32 0, i32 0, i32 0, i32 0
+ invoke void @_ZN4llvm15SmallVectorBase8grow_podEmm(%"class.llvm::SmallVectorBase"* %5, i64 0, i64 4)
+ to label %.noexc unwind label %lpad
+
+.noexc: ; preds = %if.end.i
+ %.pre.i = load i8** %EndX.i, align 8, !tbaa !0
+ br label %Retry.i
+
+invoke.cont: ; preds = %new.notnull.i, %Retry.i
+ %add.ptr.i = getelementptr inbounds i8* %3, i64 4
+ store i8* %add.ptr.i, i8** %EndX.i, align 8, !tbaa !0
+ %6 = load i8** %CapacityX.i, align 16, !tbaa !0
+ %cmp.i8 = icmp ult i8* %add.ptr.i, %6
+ br i1 %cmp.i8, label %new.notnull.i11, label %if.end.i14
+
+Retry.i10: ; preds = %if.end.i14
+ %.pre.i13 = load i8** %EndX.i, align 8, !tbaa !0
+ %new.isnull.i9 = icmp eq i8* %.pre.i13, null
+ br i1 %new.isnull.i9, label %invoke.cont2, label %new.notnull.i11
+
+new.notnull.i11: ; preds = %invoke.cont, %Retry.i10
+ %7 = phi i8* [ %.pre.i13, %Retry.i10 ], [ %add.ptr.i, %invoke.cont ]
+ %8 = bitcast i8* %7 to i32*
+ store i32 2, i32* %8, align 4, !tbaa !3
+ br label %invoke.cont2
+
+if.end.i14: ; preds = %invoke.cont
+ %9 = getelementptr inbounds %"class.llvm::SmallVector"* %sv, i64 0, i32 0, i32 0, i32 0, i32 0
+ invoke void @_ZN4llvm15SmallVectorBase8grow_podEmm(%"class.llvm::SmallVectorBase"* %9, i64 0, i64 4)
+ to label %Retry.i10 unwind label %lpad
+
+invoke.cont2: ; preds = %new.notnull.i11, %Retry.i10
+ %10 = phi i8* [ null, %Retry.i10 ], [ %7, %new.notnull.i11 ]
+ %add.ptr.i12 = getelementptr inbounds i8* %10, i64 4
+ store i8* %add.ptr.i12, i8** %EndX.i, align 8, !tbaa !0
+ invoke void @_Z1gRN4llvm11SmallVectorIiLj8EEE(%"class.llvm::SmallVector"* %sv)
+ to label %invoke.cont3 unwind label %lpad
+
+invoke.cont3: ; preds = %invoke.cont2
+ %11 = load i8** %BeginX.i.i.i.i.i.i, align 16, !tbaa !0
+ %cmp.i.i.i.i19 = icmp eq i8* %11, %1
+ br i1 %cmp.i.i.i.i19, label %_ZN4llvm11SmallVectorIiLj8EED1Ev.exit21, label %if.then.i.i.i20
+
+if.then.i.i.i20: ; preds = %invoke.cont3
+ call void @free(i8* %11) #1
+ br label %_ZN4llvm11SmallVectorIiLj8EED1Ev.exit21
+
+_ZN4llvm11SmallVectorIiLj8EED1Ev.exit21: ; preds = %invoke.cont3, %if.then.i.i.i20
+ call void @llvm.lifetime.end(i64 64, i8* %0) #1
+ ret void
+
+lpad: ; preds = %if.end.i14, %if.end.i, %invoke.cont2
+ %12 = landingpad { i8*, i32 } personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*)
+ cleanup
+ %13 = load i8** %BeginX.i.i.i.i.i.i, align 16, !tbaa !0
+ %cmp.i.i.i.i = icmp eq i8* %13, %1
+ br i1 %cmp.i.i.i.i, label %eh.resume, label %if.then.i.i.i
+
+if.then.i.i.i: ; preds = %lpad
+ call void @free(i8* %13) #1
+ br label %eh.resume
+
+eh.resume: ; preds = %if.then.i.i.i, %lpad
+ resume { i8*, i32 } %12
+}
+
+; Function Attrs: nounwind
+declare void @llvm.lifetime.start(i64, i8* nocapture) #1
+
+declare i32 @__gxx_personality_v0(...)
+
+declare void @_Z1gRN4llvm11SmallVectorIiLj8EEE(%"class.llvm::SmallVector"*) #2
+
+; Function Attrs: nounwind
+declare void @llvm.lifetime.end(i64, i8* nocapture) #1
+
+declare void @_ZN4llvm15SmallVectorBase8grow_podEmm(%"class.llvm::SmallVectorBase"*, i64, i64) #2
+
+; Function Attrs: nounwind
+declare void @free(i8* nocapture) #3
+
+attributes #0 = { ssp uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nounwind }
+attributes #2 = { "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #3 = { nounwind "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
+
+!0 = metadata !{metadata !"any pointer", metadata !1}
+!1 = metadata !{metadata !"omnipotent char", metadata !2}
+!2 = metadata !{metadata !"Simple C/C++ TBAA"}
+!3 = metadata !{metadata !"int", metadata !1}
+
Index: test/Transforms/GVN/local-pre.ll
===================================================================
--- test/Transforms/GVN/local-pre.ll (revision 190918)
+++ test/Transforms/GVN/local-pre.ll (working copy)
@@ -1,9 +1,9 @@
; RUN: opt < %s -gvn -enable-pre -S | grep "b.pre"
-define i32 @main(i32 %p) {
+define i32 @main(i32 %p, i32 %q) {
block1:
-
- br i1 true, label %block2, label %block3
+ %cmp = icmp eq i32 %p, %q
+ br i1 %cmp, label %block2, label %block3
block2:
%a = add i32 %p, 1
Index: test/Transforms/GVN/rle-nonlocal.ll
===================================================================
--- test/Transforms/GVN/rle-nonlocal.ll (revision 190918)
+++ test/Transforms/GVN/rle-nonlocal.ll (working copy)
@@ -1,8 +1,9 @@
; RUN: opt < %s -basicaa -gvn -S | FileCheck %s
-define i32 @main(i32** %p) {
+define i32 @main(i32** %p, i32 %x, i32 %y) {
block1:
- br i1 true, label %block2, label %block3
+ %cmp = icmp eq i32 %x, %y
+ br i1 %cmp , label %block2, label %block3
block2:
%a = load i32** %p
Index: test/Transforms/GVN/rle-semidominated.ll
===================================================================
--- test/Transforms/GVN/rle-semidominated.ll (revision 190918)
+++ test/Transforms/GVN/rle-semidominated.ll (working copy)
@@ -1,9 +1,10 @@
; RUN: opt < %s -basicaa -gvn -S | grep "DEAD = phi i32 "
-define i32 @main(i32* %p) {
+define i32 @main(i32* %p, i32 %x, i32 %y) {
block1:
%z = load i32* %p
- br i1 true, label %block2, label %block3
+ %cmp = icmp eq i32 %x, %y
+ br i1 %cmp, label %block2, label %block3
block2:
br label %block4
Index: test/Transforms/GVN/rle.ll
===================================================================
--- test/Transforms/GVN/rle.ll (revision 190918)
+++ test/Transforms/GVN/rle.ll (working copy)
@@ -357,13 +357,14 @@
; CHECK: ret i8 %A
}
-define i32 @chained_load(i32** %p) {
+define i32 @chained_load(i32** %p, i32 %x, i32 %y) {
block1:
%A = alloca i32*
%z = load i32** %p
store i32* %z, i32** %A
- br i1 true, label %block2, label %block3
+ %cmp = icmp eq i32 %x, %y
+ br i1 %cmp, label %block2, label %block3
block2:
%a = load i32** %p
@@ -427,10 +428,11 @@
ret i32 0
}
-define i32 @phi_trans3(i32* %p) {
+define i32 @phi_trans3(i32* %p, i32 %x, i32 %y, i32 %z) {
; CHECK-LABEL: @phi_trans3(
block1:
- br i1 true, label %block2, label %block3
+ %cmpxy = icmp eq i32 %x, %y
+ br i1 %cmpxy, label %block2, label %block3
block2:
store i32 87, i32* %p
@@ -443,7 +445,7 @@
block4:
%A = phi i32 [-1, %block2], [42, %block3]
- br i1 true, label %block5, label %exit
+ br i1 %cmpxy, label %block5, label %exit
; CHECK: block4:
; CHECK-NEXT: %D = phi i32 [ 87, %block2 ], [ 97, %block3 ]
@@ -451,11 +453,11 @@
block5:
%B = add i32 %A, 1
- br i1 true, label %block6, label %exit
+ br i1 %cmpxy, label %block6, label %exit
block6:
%C = getelementptr i32* %p, i32 %B
- br i1 true, label %block7, label %exit
+ br i1 %cmpxy, label %block7, label %exit
block7:
%D = load i32* %C
More information about the llvm-commits
mailing list