RPCS3/llvm
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm.git synced 2024-11-28 14:10:55 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Fixed the DAG combiner to better handle the folding of AND nodes for vector types. The previous code was making the assumption that the length of the bitmask returned by isConstantSplat was equal to the size of the vector type. Now we first make sure that the splat value has at least the length of the vector lane type, then we only use as many fields as we have available in the splat value.

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@163203 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Silviu Baranga 2012-09-05 08:57:21 +00:00
parent 59a4a47a7b
commit 3d5e161fe4
2 changed files with 22 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

12
lib/CodeGen/SelectionDAG/DAGCombiner.cpp
View File

@ -2496,8 +2496,18 @@ SDValue DAGCombiner::visitAND(SDNode *N) {
// lanes of the constant together.
EVT VT = Vector->getValueType(0);
unsigned BitWidth = VT.getVectorElementType().getSizeInBits();
// If the splat value has been compressed to a bitlength lower
// than the size of the vector lane, we need to re-expand it to
// the lane size.
if (BitWidth > SplatBitSize)
for (SplatValue = SplatValue.zextOrTrunc(BitWidth);
SplatBitSize < BitWidth;
SplatBitSize = SplatBitSize * 2)
SplatValue |= SplatValue.shl(SplatBitSize);
Constant = APInt::getAllOnesValue(BitWidth);
for (unsigned i = 0, n = VT.getVectorNumElements(); i < n; ++i)
for (unsigned i = 0, n = SplatBitSize/BitWidth; i < n; ++i)
Constant &= SplatValue.lshr(i*BitWidth).zextOrTrunc(BitWidth);
}
}

11
test/CodeGen/ARM/vector-extend-narrow.ll
View File

@ -62,3 +62,14 @@ define <4 x i8> @i(<4 x i8>* %x) {
%2 = sdiv <4 x i8> zeroinitializer, %1
ret <4 x i8> %2
}
; CHECK: j:
define <4 x i32> @j(<4 x i8>* %in) nounwind {
; CHECK: vld1
; CHECK: vmovl.u8
; CHECK: vmovl.u16
; CHECK-NOT: vand
%1 = load <4 x i8>* %in, align 4
%2 = zext <4 x i8> %1 to <4 x i32>
ret <4 x i32> %2
}