RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2025-01-29 16:12:44 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

Add support for additional in-reg vbroadcast patterns

Browse Source 
llvm-svn: 157127
This commit is contained in:
Nadav Rotem 2012-05-19 19:57:37 +00:00
parent 386e9a3f75
commit 441318ee29
2 changed files with 47 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

14
lib/Target/X86/X86ISelLowering.cpp
View File

@ -5026,12 +5026,18 @@ X86TargetLowering::LowerVectorBroadcast(SDValue &Op, SelectionDAG &DAG) const {
}
}
// The scalar source must be a normal load.
if (!ISD::isNormalLoad(Ld.getNode()))
return SDValue();
bool IsLoad = ISD::isNormalLoad(Ld.getNode());
unsigned ScalarSize = Ld.getValueType().getSizeInBits();
// Handle AVX2 in-register broadcasts.
if (!IsLoad && Subtarget->hasAVX2() &&
(ScalarSize == 32 || (Is256 && ScalarSize == 64)))
return DAG.getNode(X86ISD::VBROADCAST, dl, VT, Ld);
// The scalar source must be a normal load.
if (!IsLoad)
return SDValue();
if (ScalarSize == 32 || (Is256 && ScalarSize == 64))
return DAG.getNode(X86ISD::VBROADCAST, dl, VT, Ld);

37
test/CodeGen/X86/avx2-vbroadcast.ll
View File

@ -222,3 +222,40 @@ footer349VF:
ret:
ret void
}
; CHECK: _inreg0
; CHECK: broadcastss
; CHECK: ret
define <8 x i32> @_inreg0(i32 %scalar) nounwind uwtable readnone ssp {
%in = insertelement <8 x i32> undef, i32 %scalar, i32 0
%wide = shufflevector <8 x i32> %in, <8 x i32> undef, <8 x i32> zeroinitializer
ret <8 x i32> %wide
}
; CHECK: _inreg1
; CHECK: broadcastss
; CHECK: ret
define <8 x float> @_inreg1(float %scalar) nounwind uwtable readnone ssp {
%in = insertelement <8 x float> undef, float %scalar, i32 0
%wide = shufflevector <8 x float> %in, <8 x float> undef, <8 x i32> zeroinitializer
ret <8 x float> %wide
}
; CHECK: _inreg2
; CHECK: broadcastss
; CHECK: ret
define <4 x float> @_inreg2(float %scalar) nounwind uwtable readnone ssp {
%in = insertelement <4 x float> undef, float %scalar, i32 0
%wide = shufflevector <4 x float> %in, <4 x float> undef, <4 x i32> zeroinitializer
ret <4 x float> %wide
}
; CHECK: _inreg3
; CHECK: broadcastsd
; CHECK: ret
define <4 x double> @_inreg3(double %scalar) nounwind uwtable readnone ssp {
%in = insertelement <4 x double> undef, double %scalar, i32 0
%wide = shufflevector <4 x double> %in, <4 x double> undef, <4 x i32> zeroinitializer
ret <4 x double> %wide
}