// instructions. TODO: We can further optimize this scan if we split the
// instructions into different buckets based on the insert lane.
SmallPtrSet<Instruction*, 16> Visited;
+ SmallPtrSet<Instruction*, 16> ToRemove;
ReversePostOrderTraversal<Function*> RPOT(F);
for (ReversePostOrderTraversal<Function*>::rpo_iterator I = RPOT.begin(),
E = RPOT.end(); I != E; ++I) {
if (Insert->isIdenticalTo(*v) &&
DT->dominates((*v)->getParent(), Insert->getParent())) {
Insert->replaceAllUsesWith(*v);
+ ToRemove.insert(Insert);
Insert = 0;
break;
}
Visited.insert(Insert);
}
}
+
+ // Erase all of the instructions that we RAUWed.
+ for (SmallPtrSet<Instruction*, 16>::iterator v = ToRemove.begin(),
+ ve = ToRemove.end(); v != ve; ++v) {
+ assert((*v)->getNumUses() == 0 && "Can't remove instructions with uses");
+ (*v)->eraseFromParent();
+ }
}
/// The SLPVectorizer Pass.
-; RUN: opt < %s -basicaa -slp-vectorizer -dce -S -mtriple=i386-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
+; RUN: opt < %s -basicaa -slp-vectorizer -S -mtriple=i386-apple-macosx10.8.0 -mcpu=corei7-avx | FileCheck %s
target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128-n8:16:32-S128"
target triple = "i386-apple-macosx10.8.0"