RPCS3/llvm-mirror
1
0
Fork 0
mirror of https://github.com/RPCS3/llvm-mirror.git synced 2024-11-23 03:02:36 +01:00
Code Issues Projects Releases Wiki Activity

[X86] Use EVT::getVectorVT instead of changeVectorElementType in reduceVMULWidth.

Browse Source 
Changing vector element type doesn't work for v6i32->v6i16 now
that v6i32 is an MVT and v6i16 is not.

I would like to fix this in changeVectorElementType, but you
need a LLVMContext to call getVectorVT which we can't get from
an MVT.

Fixes PR50709.
This commit is contained in:
Craig Topper 2021-06-14 22:01:45 -07:00
parent adf785206e
commit e3c98d1dad
2 changed files with 28 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
lib/Target/X86/X86ISelLowering.cpp
View File

@ -42902,7 +42902,7 @@ static SDValue reduceVMULWidth(SDNode *N, SelectionDAG &DAG,
if ((NumElts % 2) != 0) if ((NumElts % 2) != 0)
return SDValue(); return SDValue();
EVT ReducedVT = VT.changeVectorElementType(MVT::i16); EVT ReducedVT = EVT::getVectorVT(*DAG.getContext(), MVT::i16, NumElts);
// Shrink the operands of mul. // Shrink the operands of mul.
SDValue NewN0 = DAG.getNode(ISD::TRUNCATE, DL, ReducedVT, N0); SDValue NewN0 = DAG.getNode(ISD::TRUNCATE, DL, ReducedVT, N0);

27
test/CodeGen/X86/pr50709.ll Normal file
View File

@ -0,0 +1,27 @@
; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
; RUN: llc < %s -mtriple=x86_64-unknown-linux-gnu | FileCheck %s
define <6 x i32> @foo(<6 x i16> %x, <6 x i16> %y) {
; CHECK-LABEL: foo:
; CHECK: # %bb.0:
; CHECK-NEXT: movq %rdi, %rax
; CHECK-NEXT: movdqa %xmm0, %xmm2
; CHECK-NEXT: pmulhuw %xmm1, %xmm2
; CHECK-NEXT: pmullw %xmm1, %xmm0
; CHECK-NEXT: movdqa %xmm0, %xmm1
; CHECK-NEXT: punpcklwd {{.*#+}} xmm1 = xmm1[0],xmm2[0],xmm1[1],xmm2[1],xmm1[2],xmm2[2],xmm1[3],xmm2[3]
; CHECK-NEXT: pslldq {{.*#+}} xmm2 = zero,zero,xmm2[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
; CHECK-NEXT: pslldq {{.*#+}} xmm0 = zero,zero,xmm0[0,1,2,3,4,5,6,7,8,9,10,11,12,13]
; CHECK-NEXT: punpckhwd {{.*#+}} xmm0 = xmm0[4],xmm2[4],xmm0[5],xmm2[5],xmm0[6],xmm2[6],xmm0[7],xmm2[7]
; CHECK-NEXT: movdqa %xmm1, %xmm2
; CHECK-NEXT: movsd {{.*#+}} xmm2 = xmm0[0],xmm2[1]
; CHECK-NEXT: shufps {{.*#+}} xmm1 = xmm1[0,1],xmm2[2,0]
; CHECK-NEXT: pshufd {{.*#+}} xmm0 = xmm0[1,2,2,3]
; CHECK-NEXT: movq %xmm0, 16(%rdi)
; CHECK-NEXT: movaps %xmm1, (%rdi)
; CHECK-NEXT: retq
%a = zext <6 x i16> %x to <6 x i32>
%b = zext <6 x i16> %y to <6 x i32>
%c = mul <6 x i32> %a, %b
ret <6 x i32> %c
}