6502/llvm-6502
1
0
Fork 0
mirror of https://github.com/c64scene-ar/llvm-6502.git synced 2025-02-19 06:31:18 +00:00
Code Issues Projects Releases Wiki Activity

On Haswell, perfer storing YMM registers using a single instruction.

Browse Source 
git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@157129 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Nadav Rotem 2012-05-19 20:30:08 +00:00
parent 4fc8a5de44
commit 87d35e8c71
2 changed files with 18 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

9
lib/Target/X86/X86ISelLowering.cpp
View File

@ -14532,13 +14532,12 @@ static SDValue PerformSTORECombine(SDNode *N, SelectionDAG &DAG,
const TargetLowering &TLI = DAG.getTargetLoweringInfo(); const TargetLowering &TLI = DAG.getTargetLoweringInfo();
// If we are saving a concatenation of two XMM registers, perform two stores. // If we are saving a concatenation of two XMM registers, perform two stores.
// This is better in Sandy Bridge cause one 256-bit mem op is done via two // On Sandy Bridge, 256-bit memory operations are executed by two
// 128-bit ones. If in the future the cost becomes only one memory access the // 128-bit ports. However, on Haswell it is better to issue a single 256-bit
// first version would be better. // memory operation.
if (VT.getSizeInBits() == 256 && if (VT.getSizeInBits() == 256 && !Subtarget->hasAVX2() &&
StoredVal.getNode()->getOpcode() == ISD::CONCAT_VECTORS && StoredVal.getNode()->getOpcode() == ISD::CONCAT_VECTORS &&
StoredVal.getNumOperands() == 2) { StoredVal.getNumOperands() == 2) {
SDValue Value0 = StoredVal.getOperand(0); SDValue Value0 = StoredVal.getOperand(0);
SDValue Value1 = StoredVal.getOperand(1); SDValue Value1 = StoredVal.getOperand(1);

14
test/CodeGen/X86/2012-05-19-avx2-store.ll Normal file
View File

@ -0,0 +1,14 @@
; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=corei7-avx -mattr=+avx2 | FileCheck %s
define void @double_save(<4 x i32>* %Ap, <4 x i32>* %Bp, <8 x i32>* %P) nounwind ssp {
entry:
; CHECK: vmovaps
; CHECK: vmovaps
; CHECK: vinsertf128
; CHECK: vmovups
%A = load <4 x i32>* %Ap
%B = load <4 x i32>* %Bp
%Z = shufflevector <4 x i32>%A, <4 x i32>%B, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
store <8 x i32> %Z, <8 x i32>* %P, align 16
ret void
}