summaryrefslogtreecommitdiffstats
path: root/llvm/lib/Target/X86/X86ISelLowering.cpp
diff options
context:
space:
mode:
authorSimon Pilgrim <llvm-dev@redking.me.uk>2017-02-24 17:46:18 +0000
committerSimon Pilgrim <llvm-dev@redking.me.uk>2017-02-24 17:46:18 +0000
commitbd9fb2ae959dc2bc0a2a6a309b56ea239d41797e (patch)
treef6b76d75fb74cca216bc2f37094c6d97c0ba4d13 /llvm/lib/Target/X86/X86ISelLowering.cpp
parent24a1bedf765fd2dab8384073c9ac930563fa0110 (diff)
downloadbcm5719-llvm-bd9fb2ae959dc2bc0a2a6a309b56ea239d41797e.tar.gz
bcm5719-llvm-bd9fb2ae959dc2bc0a2a6a309b56ea239d41797e.zip
[APInt] Add APInt::extractBits() method to extract APInt subrange
The current pattern for extract bits in range is typically: Mask.lshr(BitOffset).trunc(SubSizeInBits); Which can be particularly slow for large APInts (MaskSizeInBits > 64) as they require the allocation of memory for the temporary variable. This is another of the compile time issues identified in PR32037 (see also D30265). This patch adds the APInt::extractBits() helper method which avoids the temporary memory allocation. Differential Revision: https://reviews.llvm.org/D30336 llvm-svn: 296141
Diffstat (limited to 'llvm/lib/Target/X86/X86ISelLowering.cpp')
-rw-r--r--llvm/lib/Target/X86/X86ISelLowering.cpp8
1 files changed, 4 insertions, 4 deletions
diff --git a/llvm/lib/Target/X86/X86ISelLowering.cpp b/llvm/lib/Target/X86/X86ISelLowering.cpp
index 618d7a8c561..348f8308099 100644
--- a/llvm/lib/Target/X86/X86ISelLowering.cpp
+++ b/llvm/lib/Target/X86/X86ISelLowering.cpp
@@ -5207,8 +5207,8 @@ static bool getTargetConstantBitsFromNode(SDValue Op, unsigned EltSizeInBits,
EltBits.resize(NumElts, APInt(EltSizeInBits, 0));
for (unsigned i = 0; i != NumElts; ++i) {
- APInt UndefEltBits = UndefBits.lshr(i * EltSizeInBits);
- UndefEltBits = UndefEltBits.zextOrTrunc(EltSizeInBits);
+ unsigned BitOffset = i * EltSizeInBits;
+ APInt UndefEltBits = UndefBits.extractBits(EltSizeInBits, BitOffset);
// Only treat an element as UNDEF if all bits are UNDEF.
if (UndefEltBits.isAllOnesValue()) {
@@ -5223,7 +5223,7 @@ static bool getTargetConstantBitsFromNode(SDValue Op, unsigned EltSizeInBits,
if (UndefEltBits.getBoolValue() && !AllowPartialUndefs)
return false;
- APInt Bits = MaskBits.lshr(i * EltSizeInBits).zextOrTrunc(EltSizeInBits);
+ APInt Bits = MaskBits.extractBits(EltSizeInBits, BitOffset);
EltBits[i] = Bits.getZExtValue();
}
return true;
@@ -6421,7 +6421,7 @@ static Constant *getConstantVector(MVT VT, const APInt &SplatValue,
SmallVector<Constant *, 32> ConstantVec;
for (unsigned i = 0; i < NumElm; i++) {
- APInt Val = SplatValue.lshr(ScalarSize * i).trunc(ScalarSize);
+ APInt Val = SplatValue.extractBits(ScalarSize, ScalarSize * i);
Constant *Const;
if (VT.isFloatingPoint()) {
assert((ScalarSize == 32 || ScalarSize == 64) &&
OpenPOWER on IntegriCloud