RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-25 12:12:47 +01:00
Code Issues Projects Releases Wiki Activity

[X86][XOP] Use default AVX2 lowering for v4i64 ashr by splat constants

Browse Source 
XOP shifts only support 128-bit vectors, so we were ending up with less optimal codegen requiring constants

llvm-svn: 308430
This commit is contained in:
Simon Pilgrim 2017-07-19 10:29:31 +00:00
parent 816413b87e
commit 1871c33bf5
2 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
lib/Target/X86/X86ISelLowering.cpp
View File

@ -22022,8 +22022,9 @@ static SDValue LowerScalarImmediateShift(SDValue Op, SelectionDAG &DAG,
return getTargetVShiftByConstNode(X86Opc, dl, VT, R, ShiftAmt, DAG);
// i64 SRA needs to be performed as partial shifts.
if ((VT == MVT::v2i64 || (Subtarget.hasInt256() && VT == MVT::v4i64)) &&
Op.getOpcode() == ISD::SRA && !Subtarget.hasXOP())
if (((!Subtarget.hasXOP() && VT == MVT::v2i64) ||
(Subtarget.hasInt256() && VT == MVT::v4i64)) &&
Op.getOpcode() == ISD::SRA)
return ArithmeticShiftRight64(ShiftAmt);
if (VT == MVT::v16i8 ||

5
test/CodeGen/X86/vector-shift-ashr-256.ll
View File

@ -1699,10 +1699,9 @@ define <4 x i64> @splatconstant_shift_v4i64(<4 x i64> %a) nounwind {
;
; XOPAVX2-LABEL: splatconstant_shift_v4i64:
; XOPAVX2: # BB#0:
; XOPAVX2-NEXT: vpsrad $7, %ymm0, %ymm1
; XOPAVX2-NEXT: vpsrlq $7, %ymm0, %ymm0
; XOPAVX2-NEXT: vpbroadcastq {{.*#+}} ymm1 = [72057594037927936,72057594037927936,72057594037927936,72057594037927936]
; XOPAVX2-NEXT: vpxor %ymm1, %ymm0, %ymm0
; XOPAVX2-NEXT: vpsubq %ymm1, %ymm0, %ymm0
; XOPAVX2-NEXT: vpblendd {{.*#+}} ymm0 = ymm0[0],ymm1[1],ymm0[2],ymm1[3],ymm0[4],ymm1[5],ymm0[6],ymm1[7]
; XOPAVX2-NEXT: retq
;
; AVX512-LABEL: splatconstant_shift_v4i64: