Skip to content

Commit ec02635

Browse files
committed
[amdgpu] Skip OR combining on 64-bit integer before legalizing ops.
Reviewers: arsenm, rampitec Subscribers: kzhuravl, jvesely, wdng, nhaehnle, yaxunl, dstuttard, tpr, t-tye, hiraditya, kerbowa, llvm-commits Tags: #llvm Differential Revision: https://reviews.llvm.org/D81710
1 parent e640598 commit ec02635

File tree

2 files changed

+2
-2
lines changed

2 files changed

+2
-2
lines changed

llvm/lib/Target/AMDGPU/SIISelLowering.cpp

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -8958,7 +8958,7 @@ SDValue SITargetLowering::performOrCombine(SDNode *N,
89588958
}
89598959
}
89608960

8961-
if (VT != MVT::i64)
8961+
if (VT != MVT::i64 || DCI.isBeforeLegalizeOps())
89628962
return SDValue();
89638963

89648964
// TODO: This could be a generic combine with a predicate for extracting the

llvm/test/CodeGen/AMDGPU/fshr.ll

Lines changed: 1 addition & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -763,7 +763,7 @@ define <3 x i16> @v_fshr_v3i16(<3 x i16> %src0, <3 x i16> %src1, <3 x i16> %src2
763763
; SI-NEXT: v_cmp_eq_u32_e32 vcc, 0, v1
764764
; SI-NEXT: v_cndmask_b32_e32 v1, v2, v5, vcc
765765
; SI-NEXT: v_and_b32_e32 v2, v9, v1
766-
; SI-NEXT: v_alignbit_b32 v1, v2, v0, 16
766+
; SI-NEXT: v_alignbit_b32 v1, v1, v0, 16
767767
; SI-NEXT: s_setpc_b64 s[30:31]
768768
;
769769
; VI-LABEL: v_fshr_v3i16:

0 commit comments

Comments
 (0)