[llvm] r205179 - Look at shuffles of build_vectors in DAGCombiner::visitEXTRACT_VECTOR_ELT
Hal Finkel
hfinkel at anl.gov
Mon Mar 31 04:43:19 PDT 2014
Author: hfinkel
Date: Mon Mar 31 06:43:19 2014
New Revision: 205179
URL: http://llvm.org/viewvc/llvm-project?rev=205179&view=rev
Log:
Look at shuffles of build_vectors in DAGCombiner::visitEXTRACT_VECTOR_ELT
When the loop vectorizer vectorizes code that uses the loop induction variable,
we often end up with IR like this:
%b1 = insertelement <2 x i32> undef, i32 %v, i32 0
%b2 = shufflevector <2 x i32> %b1, <2 x i32> undef, <2 x i32> zeroinitializer
%i = add <2 x i32> %b2, <i32 2, i32 3>
If the add in this example is not legal (as is the case on PPC with VSX), it
will be scalarized, and we'll end up with a number of extract_vector_elt nodes
with the vector shuffle as the input operand, and that vector shuffle is fed by
one or more build_vector nodes. By the time that vector operations are
expanded, visitEXTRACT_VECTOR_ELT will not create new extract_vector_elt by
looking through the vector shuffle (to make sure that no illegal operations are
created), and so the extract_vector_elt -> vector shuffle -> build_vector is
never simplified to an operand of the build vector.
By looking at build_vectors through a shuffle we fix this particular situation,
preventing a vector from being built, only to be deconstructed again (for the
scalarized add) -- an expensive proposition when this all needs to be done via
the stack. We probably want a more comprehensive fix here where we look back
recursively through any shuffles to any build_vectors or scalar_to_vectors,
etc. but that can come later.
Modified:
llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
llvm/trunk/test/CodeGen/PowerPC/vsx.ll
Modified: llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp?rev=205179&r1=205178&r2=205179&view=diff
==============================================================================
--- llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp (original)
+++ llvm/trunk/lib/CodeGen/SelectionDAG/DAGCombiner.cpp Mon Mar 31 06:43:19 2014
@@ -9745,9 +9745,10 @@ SDValue DAGCombiner::visitEXTRACT_VECTOR
// We only perform this optimization before the op legalization phase because
// we may introduce new vector instructions which are not backed by TD
// patterns. For example on AVX, extracting elements from a wide vector
- // without using extract_subvector.
+ // without using extract_subvector. However, if we can find an underlying
+ // scalar value, then we can always use that.
if (InVec.getOpcode() == ISD::VECTOR_SHUFFLE
- && ConstEltNo && !LegalOperations) {
+ && ConstEltNo) {
int Elt = cast<ConstantSDNode>(EltNo)->getZExtValue();
int NumElem = VT.getVectorNumElements();
ShuffleVectorSDNode *SVOp = cast<ShuffleVectorSDNode>(InVec);
@@ -9759,16 +9760,32 @@ SDValue DAGCombiner::visitEXTRACT_VECTOR
return DAG.getUNDEF(NVT);
// Select the right vector half to extract from.
+ SDValue SVInVec;
if (OrigElt < NumElem) {
- InVec = InVec->getOperand(0);
+ SVInVec = InVec->getOperand(0);
} else {
- InVec = InVec->getOperand(1);
+ SVInVec = InVec->getOperand(1);
OrigElt -= NumElem;
}
- EVT IndexTy = TLI.getVectorIdxTy();
- return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SDLoc(N), NVT,
- InVec, DAG.getConstant(OrigElt, IndexTy));
+ if (SVInVec.getOpcode() == ISD::BUILD_VECTOR) {
+ SDValue InOp = SVInVec.getOperand(OrigElt);
+ if (InOp.getValueType() != NVT) {
+ assert(InOp.getValueType().isInteger() && NVT.isInteger());
+ InOp = DAG.getSExtOrTrunc(InOp, SDLoc(SVInVec), NVT);
+ }
+
+ return InOp;
+ }
+
+ // FIXME: We should handle recursing on other vector shuffles and
+ // scalar_to_vector here as well.
+
+ if (!LegalOperations) {
+ EVT IndexTy = TLI.getVectorIdxTy();
+ return DAG.getNode(ISD::EXTRACT_VECTOR_ELT, SDLoc(N), NVT,
+ SVInVec, DAG.getConstant(OrigElt, IndexTy));
+ }
}
// Perform only after legalization to ensure build_vector / vector_shuffle
Modified: llvm/trunk/test/CodeGen/PowerPC/vsx.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/trunk/test/CodeGen/PowerPC/vsx.ll?rev=205179&r1=205178&r2=205179&view=diff
==============================================================================
--- llvm/trunk/test/CodeGen/PowerPC/vsx.ll (original)
+++ llvm/trunk/test/CodeGen/PowerPC/vsx.ll Mon Mar 31 06:43:19 2014
@@ -624,3 +624,19 @@ define <2 x double> @test70(<2 x i8> %a)
; CHECK: blr
}
+define <2 x i32> @test80(i32 %v) {
+ %b1 = insertelement <2 x i32> undef, i32 %v, i32 0
+ %b2 = shufflevector <2 x i32> %b1, <2 x i32> undef, <2 x i32> zeroinitializer
+ %i = add <2 x i32> %b2, <i32 2, i32 3>
+ ret <2 x i32> %i
+
+; CHECK-LABEL: @test80
+; CHECK: addi
+; CHECK: addi
+; CHECK: lxvd2x
+; CHECK-NOT: stxvd2x
+; FIXME: We still make one vector for each vector element and this shuffle them
+; together instead of just composing one vector on the stack.
+; CHECK: blr
+}
+
More information about the llvm-commits
mailing list