maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity

[X86][AVX] ComputeNumSignBitsForTargetNode - add support for X86ISD::VTRUNC 

llvm-svn: 316462
This commit is contained in:
Simon Pilgrim 2017-10-24 17:04:57 +00:00
parent 8aa32ffbad
commit 5e8c3f328f
3 changed files with 15 additions and 15 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
llvm/lib/Target/X86/X86ISelLowering.cpp
View File

@ -27239,6 +27239,16 @@ unsigned X86TargetLowering::ComputeNumSignBitsForTargetNode(
return Tmp;
}
case X86ISD::VTRUNC: {
SDValue Src = Op.getOperand(0);
unsigned NumSrcBits = Src.getScalarValueSizeInBits();
assert(VTBits < NumSrcBits && "Illegal truncation input type");
unsigned Tmp = DAG.ComputeNumSignBits(Src, Depth + 1);
if (Tmp > (NumSrcBits - VTBits))
return Tmp - (NumSrcBits - VTBits);
return 1;
}
case X86ISD::PACKSS: {
// PACKSS is just a truncation if the sign bits extend to the packed size.
// TODO: Add DemandedElts support.

8
llvm/test/CodeGen/X86/avg-mask.ll
View File

@ -254,9 +254,7 @@ define <8 x i16> @avg_v8i16_mask(<8 x i16> %a, <8 x i16> %b, <8 x i16> %src, i8
; AVX512F-NEXT: kmovw %edi, %k1
; AVX512F-NEXT: vpternlogq $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
; AVX512F-NEXT: vpmovqw %zmm1, %xmm1
; AVX512F-NEXT: vpand %xmm1, %xmm0, %xmm0
; AVX512F-NEXT: vpandn %xmm2, %xmm1, %xmm1
; AVX512F-NEXT: vpor %xmm1, %xmm0, %xmm0
; AVX512F-NEXT: vpblendvb %xmm1, %xmm0, %xmm2, %xmm0
; AVX512F-NEXT: vzeroupper
; AVX512F-NEXT: retq
;
@ -311,9 +309,7 @@ define <16 x i16> @avg_v16i16_mask(<16 x i16> %a, <16 x i16> %b, <16 x i16> %src
; AVX512F-NEXT: kmovw %edi, %k1
; AVX512F-NEXT: vpternlogd $255, %zmm1, %zmm1, %zmm1 {%k1} {z}
; AVX512F-NEXT: vpmovdw %zmm1, %ymm1
; AVX512F-NEXT: vpand %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpandn %ymm2, %ymm1, %ymm1
; AVX512F-NEXT: vpor %ymm1, %ymm0, %ymm0
; AVX512F-NEXT: vpblendvb %ymm1, %ymm0, %ymm2, %ymm0
; AVX512F-NEXT: retq
;
; AVX512BWVL-LABEL: avg_v16i16_mask:

12
llvm/test/CodeGen/X86/vselect-packss.ll
View File

@ -54,9 +54,7 @@ define <16 x i8> @vselect_packss_v16i16(<16 x i16> %a0, <16 x i16> %a1, <16 x i8
; AVX512-NEXT: vpcmpeqw %ymm1, %ymm0, %ymm0
; AVX512-NEXT: vpmovsxwd %ymm0, %zmm0
; AVX512-NEXT: vpmovdb %zmm0, %xmm0
; AVX512-NEXT: vpand %xmm2, %xmm0, %xmm1
; AVX512-NEXT: vpandn %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vpblendvb %xmm0, %xmm2, %xmm3, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = icmp eq <16 x i16> %a0, %a1
@ -134,9 +132,7 @@ define <16 x i8> @vselect_packss_v16i32(<16 x i32> %a0, <16 x i32> %a1, <16 x i8
; AVX512-NEXT: vpcmpeqd %zmm1, %zmm0, %k1
; AVX512-NEXT: vpternlogd $255, %zmm0, %zmm0, %zmm0 {%k1} {z}
; AVX512-NEXT: vpmovdb %zmm0, %xmm0
; AVX512-NEXT: vpand %xmm2, %xmm0, %xmm1
; AVX512-NEXT: vpandn %xmm3, %xmm0, %xmm0
; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vpblendvb %xmm0, %xmm2, %xmm3, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = icmp eq <16 x i32> %a0, %a1
@ -392,9 +388,7 @@ define <16 x i8> @vselect_packss_v16i64(<16 x i64> %a0, <16 x i64> %a1, <16 x i8
; AVX512-NEXT: vpmovqd %zmm0, %ymm0
; AVX512-NEXT: vinserti64x4 $1, %ymm0, %zmm8, %zmm0
; AVX512-NEXT: vpmovdb %zmm0, %xmm0
; AVX512-NEXT: vpand %xmm4, %xmm0, %xmm1
; AVX512-NEXT: vpandn %xmm5, %xmm0, %xmm0
; AVX512-NEXT: vpor %xmm0, %xmm1, %xmm0
; AVX512-NEXT: vpblendvb %xmm0, %xmm4, %xmm5, %xmm0
; AVX512-NEXT: vzeroupper
; AVX512-NEXT: retq
%1 = icmp eq <16 x i64> %a0, %a1