aboutsummaryrefslogtreecommitdiffstats
diff options
context:
space:
mode:
authorEli Friedman <eli.friedman@gmail.com>2009-07-18 09:07:47 +0000
committerEli Friedman <eli.friedman@gmail.com>2009-07-18 09:07:47 +0000
commitfc21f8ff1404f26ceeee3202714cbc307d579997 (patch)
tree8cda1e6dd30545fa7532064a286b996d451d427c
parent62bb4134358d0d1f8b892cf0df42d758ac8e19e7 (diff)
downloadexternal_llvm-fc21f8ff1404f26ceeee3202714cbc307d579997.zip
external_llvm-fc21f8ff1404f26ceeee3202714cbc307d579997.tar.gz
external_llvm-fc21f8ff1404f26ceeee3202714cbc307d579997.tar.bz2
Canonicalize insert/extractelement from single-element vectors into
bitcasts. It would also be possible to canonicalize the other way; does anyone have a preference? git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@76300 91177308-0d34-0410-b5e6-96231b3b80d8
-rw-r--r--lib/Transforms/Scalar/InstructionCombining.cpp22
-rw-r--r--test/Transforms/InstCombine/vec-single-elt.ll12
2 files changed, 28 insertions, 6 deletions
diff --git a/lib/Transforms/Scalar/InstructionCombining.cpp b/lib/Transforms/Scalar/InstructionCombining.cpp
index 060abc5..0a5fc80 100644
--- a/lib/Transforms/Scalar/InstructionCombining.cpp
+++ b/lib/Transforms/Scalar/InstructionCombining.cpp
@@ -12366,13 +12366,18 @@ Instruction *InstCombiner::visitExtractElementInst(ExtractElementInst &EI) {
if (op0)
return ReplaceInstUsesWith(EI, op0);
}
-
+
+ unsigned VectorWidth =
+ cast<VectorType>(EI.getOperand(0)->getType())->getNumElements();
+
+ // Canonicalize extractelement from a vector of width 1 to a bitcast
+ if (VectorWidth == 1)
+ return new BitCastInst(EI.getOperand(0), EI.getType());
+
// If extracting a specified index from the vector, see if we can recursively
// find a previously computed scalar that was inserted into the vector.
if (ConstantInt *IdxC = dyn_cast<ConstantInt>(EI.getOperand(1))) {
unsigned IndexVal = IdxC->getZExtValue();
- unsigned VectorWidth =
- cast<VectorType>(EI.getOperand(0)->getType())->getNumElements();
// If this is extracting an invalid index, turn this into undef, to avoid
// crashing the code below.
@@ -12382,7 +12387,7 @@ Instruction *InstCombiner::visitExtractElementInst(ExtractElementInst &EI) {
// This instruction only demands the single element from the input vector.
// If the input vector has a single use, simplify it based on this use
// property.
- if (EI.getOperand(0)->hasOneUse() && VectorWidth != 1) {
+ if (EI.getOperand(0)->hasOneUse()) {
APInt UndefElts(VectorWidth, 0);
APInt DemandedMask(VectorWidth, 1 << IndexVal);
if (Value *V = SimplifyDemandedVectorElts(EI.getOperand(0),
@@ -12619,13 +12624,18 @@ Instruction *InstCombiner::visitInsertElementInst(InsertElementInst &IE) {
// Inserting an undef or into an undefined place, remove this.
if (isa<UndefValue>(ScalarOp) || isa<UndefValue>(IdxOp))
ReplaceInstUsesWith(IE, VecOp);
-
+
+ unsigned NumVectorElts = IE.getType()->getNumElements();
+
+ // Canonicalize insertelement into vector of width 1 to a bitcast
+ if (NumVectorElts == 1)
+ return new BitCastInst(IE.getOperand(1), IE.getType());
+
// If the inserted element was extracted from some other vector, and if the
// indexes are constant, try to turn this into a shufflevector operation.
if (ExtractElementInst *EI = dyn_cast<ExtractElementInst>(ScalarOp)) {
if (isa<ConstantInt>(EI->getOperand(1)) && isa<ConstantInt>(IdxOp) &&
EI->getOperand(0)->getType() == IE.getType()) {
- unsigned NumVectorElts = IE.getType()->getNumElements();
unsigned ExtractedIdx =
cast<ConstantInt>(EI->getOperand(1))->getZExtValue();
unsigned InsertedIdx = cast<ConstantInt>(IdxOp)->getZExtValue();
diff --git a/test/Transforms/InstCombine/vec-single-elt.ll b/test/Transforms/InstCombine/vec-single-elt.ll
new file mode 100644
index 0000000..45bca42
--- /dev/null
+++ b/test/Transforms/InstCombine/vec-single-elt.ll
@@ -0,0 +1,12 @@
+; RUN: llvm-as < %s | opt -instcombine | llvm-dis | not grep element
+; insert/extractelement should canonicalize to bitcast
+
+define i64 @a(<1 x i64> %x) {
+ %r = extractelement <1 x i64> %x, i32 0
+ ret i64 %r
+}
+
+define <1 x i64> @b(i64 %x) {
+ %r = insertelement <1 x i64> undef, i64 %x, i32 0
+ ret <1 x i64> %r
+}