6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2025-06-26 07:24:25 +00:00
Code Issues Projects Releases Wiki Activity

Change SelectionDAG type legalization to allow BUILD_VECTOR operands to be

Browse Source 
promoted to legal types without changing the type of the vector.  This is
following a suggestion from Duncan
(http://lists.cs.uiuc.edu/pipermail/llvmdev/2009-February/019923.html).
The transformation that used to be done during type legalization is now
postponed to DAG legalization.  This allows the BUILD_VECTORs to be optimized
and potentially handled specially by target-specific code.

It turns out that this is also consistent with an optimization done by the
DAG combiner: a BUILD_VECTOR and INSERT_VECTOR_ELT may be combined by
replacing one of the BUILD_VECTOR operands with the newly inserted element;
but INSERT_VECTOR_ELT allows its scalar operand to be larger than the
element type, with any extra high bits being implicitly truncated.  The
result is a BUILD_VECTOR where one of the operands has a type larger the
the vector element type.

Any code that operates on BUILD_VECTORs may now need to be aware of the
potential type discrepancy between the vector element type and the
BUILD_VECTOR operands.  This patch updates all of the places that I could
find to handle that case.


git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@68996 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Bob Wilson
2009-04-13 22:05:19 +00:00
parent 88c7af096b
commit b1303d05a8
6 changed files with 83 additions and 32 deletions
Download Patch File Download Diff File Expand all files Collapse all files

35
lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
View File

@ -5483,6 +5483,41 @@ SDValue SelectionDAGLegalize::ExpandBUILD_VECTOR(SDNode *Node) {
MVT OpVT = SplatValue.getValueType();
MVT EltVT = VT.getVectorElementType();
// Check if the BUILD_VECTOR operands were promoted to legalize their types.
if (OpVT != EltVT) {
// Now that the DAG combiner and target-specific lowering have had a
// chance to optimize/recognize the BUILD_VECTOR with promoted operands,
// transform it so the operand types match the vector. Build a vector of
// half the length out of elements of twice the bitwidth.
// For example <4 x i16> -> <2 x i32>.
MVT NewVT = MVT::getIntegerVT(2 * EltVT.getSizeInBits());
assert(OpVT.isSimple() && NewVT.isSimple());
SmallVector<SDValue, 16> NewElts;
for (unsigned i = 0; i < NumElems; i += 2) {
// Combine two successive elements into one promoted element.
SDValue Lo = Node->getOperand(i);
SDValue Hi = Node->getOperand(i+1);
if (TLI.isBigEndian())
std::swap(Lo, Hi);
Lo = DAG.getZeroExtendInReg(Lo, dl, EltVT);
Hi = DAG.getNode(ISD::SHL, dl, OpVT, Hi,
DAG.getConstant(EltVT.getSizeInBits(),
TLI.getPointerTy()));
NewElts.push_back(DAG.getNode(ISD::OR, dl, OpVT, Lo, Hi));
}
SDValue NewVec = DAG.getNode(ISD::BUILD_VECTOR, dl,
MVT::getVectorVT(NewVT, NewElts.size()),
&NewElts[0], NewElts.size());
// Recurse
NewVec = ExpandBUILD_VECTOR(NewVec.getNode());
// Convert the new vector to the old vector type.
return DAG.getNode(ISD::BIT_CONVERT, dl, VT, NewVec);
}
// If the only non-undef value is the low element, turn this into a
// SCALAR_TO_VECTOR node. If this is { X, X, X, X }, determine X.
bool isOnlyLowElement = true;