RPCSX/llvm
1
0
Fork 0
mirror of https://github.com/RPCSX/llvm.git synced 2024-12-22 20:18:38 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

[DAGCombiner] Add general constant vector support to (shl (add x, c1), c2) -> (add (shl x, c2), c1 << c2)

Browse Source 
We already supported scalar constant / splatted constant vector - now accepts any (non opaque) constant scalar / vector

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@284613 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Simon Pilgrim 2016-10-19 17:12:22 +00:00
parent 0c4885106a
commit 9f2ab76dca
2 changed files with 6 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
lib/CodeGen/SelectionDAG/DAGCombiner.cpp
View File

@ -4667,12 +4667,13 @@ SDValue DAGCombiner::visitSHL(SDNode *N) {
// fold (shl (add x, c1), c2) -> (add (shl x, c2), c1 << c2)
// Variant of version done on multiply, except mul by a power of 2 is turned
// into a shift.
APInt Val;
if (N1C && N0.getOpcode() == ISD::ADD && N0.getNode()->hasOneUse() &&
(isa<ConstantSDNode>(N0.getOperand(1)) ||
ISD::isConstantSplatVector(N0.getOperand(1).getNode(), Val))) {
if (N0.getOpcode() == ISD::ADD && N0.getNode()->hasOneUse() &&
isConstantOrConstantVector(N1, /* No Opaques */ true) &&
isConstantOrConstantVector(N0.getOperand(1), /* No Opaques */ true)) {
SDValue Shl0 = DAG.getNode(ISD::SHL, SDLoc(N0), VT, N0.getOperand(0), N1);
SDValue Shl1 = DAG.getNode(ISD::SHL, SDLoc(N1), VT, N0.getOperand(1), N1);
AddToWorklist(Shl0.getNode());
AddToWorklist(Shl1.getNode());
return DAG.getNode(ISD::ADD, SDLoc(N), VT, Shl0, Shl1);
}

2
test/CodeGen/X86/combine-shl.ll
View File

@ -526,8 +526,8 @@ define <4 x i32> @combine_vec_shl_add1(<4 x i32> %x) {
;
; AVX-LABEL: combine_vec_shl_add1:
; AVX: # BB#0:
; AVX-NEXT: vpaddd {{.*}}(%rip), %xmm0, %xmm0
; AVX-NEXT: vpsllvd {{.*}}(%rip), %xmm0, %xmm0
; AVX-NEXT: vpaddd {{.*}}(%rip), %xmm0, %xmm0
; AVX-NEXT: retq
%1 = add <4 x i32> %x, <i32 5, i32 6, i32 7, i32 8>
%2 = shl <4 x i32> %1, <i32 1, i32 2, i32 3, i32 4>