6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2024-12-18 10:31:57 +00:00
Code Issues Projects Releases Wiki Activity

Generate AVX/AVX2 shuffles even when there is a memory op somewhere else in the program.

Browse Source 
Starting r155461 we are able to select patterns for vbroadcast even when the load op is used by other users.

Fix PR11900.




git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@156539 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Nadav Rotem 2012-05-10 12:22:05 +00:00
parent 7b521ad6ee
commit 5fc2187a02
2 changed files with 17 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

4
lib/Target/X86/X86ISelLowering.cpp
View File

@ -5029,10 +5029,6 @@ X86TargetLowering::LowerVectorBroadcast(SDValue &Op, SelectionDAG &DAG) const {
if (!ISD::isNormalLoad(Ld.getNode())) if (!ISD::isNormalLoad(Ld.getNode()))
return SDValue(); return SDValue();
// Reject loads that have uses of the chain result
if (Ld->hasAnyUseOfValue(1))
return SDValue();
unsigned ScalarSize = Ld.getValueType().getSizeInBits(); unsigned ScalarSize = Ld.getValueType().getSizeInBits();
if (ScalarSize == 32 || (Is256 && ScalarSize == 64)) if (ScalarSize == 32 || (Is256 && ScalarSize == 64))

17
test/CodeGen/X86/avx-vbroadcast.ll
View File

@ -112,3 +112,20 @@ entry:
%vecinit2.i = insertelement <2 x double> %vecinit.i, double %q, i32 1 %vecinit2.i = insertelement <2 x double> %vecinit.i, double %q, i32 1
ret <2 x double> %vecinit2.i ret <2 x double> %vecinit2.i
} }
; CHECK: _RR
; CHECK: vbroadcastss (%
; CHECK: ret
define <4 x float> @_RR(float* %ptr, i32* %k) nounwind uwtable readnone ssp {
entry:
%q = load float* %ptr, align 4
%vecinit.i = insertelement <4 x float> undef, float %q, i32 0
%vecinit2.i = insertelement <4 x float> %vecinit.i, float %q, i32 1
%vecinit4.i = insertelement <4 x float> %vecinit2.i, float %q, i32 2
%vecinit6.i = insertelement <4 x float> %vecinit4.i, float %q, i32 3
; force a chain
%j = load i32* %k, align 4
store i32 %j, i32* undef
ret <4 x float> %vecinit6.i
}