[llvm] 62cdfda - [DAG] canCreateUndefOrPoison - add freeze(insert_subvector(x,y,c)) -> insert_subvector(freeze(x),freeze(y),c) support
Simon Pilgrim via llvm-commits
llvm-commits at lists.llvm.org
Sat Sep 3 05:41:51 PDT 2022
Author: Simon Pilgrim
Date: 2022-09-03T13:41:33+01:00
New Revision: 62cdfdab4d717e775a002349f3f65bee3037d426
URL: https://github.com/llvm/llvm-project/commit/62cdfdab4d717e775a002349f3f65bee3037d426
DIFF: https://github.com/llvm/llvm-project/commit/62cdfdab4d717e775a002349f3f65bee3037d426.diff
LOG: [DAG] canCreateUndefOrPoison - add freeze(insert_subvector(x,y,c)) -> insert_subvector(freeze(x),freeze(y),c) support
We already have plenty of assertions in place to ensure that the insertion index is constant and inrange
Added:
Modified:
llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
llvm/test/CodeGen/X86/freeze-vector.ll
Removed:
################################################################################
diff --git a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
index 711bd92a51af7..cffdc3ace5985 100644
--- a/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/SelectionDAG.cpp
@@ -4603,6 +4603,7 @@ bool SelectionDAG::canCreateUndefOrPoison(SDValue Op, const APInt &DemandedElts,
case ISD::AssertSext:
case ISD::AssertZext:
case ISD::FREEZE:
+ case ISD::INSERT_SUBVECTOR:
case ISD::AND:
case ISD::OR:
case ISD::XOR:
diff --git a/llvm/test/CodeGen/X86/freeze-vector.ll b/llvm/test/CodeGen/X86/freeze-vector.ll
index 0275d25cebcd5..af34e5667fa9d 100644
--- a/llvm/test/CodeGen/X86/freeze-vector.ll
+++ b/llvm/test/CodeGen/X86/freeze-vector.ll
@@ -5,9 +5,7 @@
define <4 x i32> @freeze_insert_subvector(<8 x i32> %a0) nounwind {
; CHECK-LABEL: freeze_insert_subvector:
; CHECK: # %bb.0:
-; CHECK-NEXT: vmovaps %xmm0, %xmm0
-; CHECK-NEXT: vextractf128 $1, %ymm0, %xmm0
-; CHECK-NEXT: vzeroupper
+; CHECK-NEXT: vxorps %xmm0, %xmm0, %xmm0
; CHECK-NEXT: ret{{[l|q]}}
%x = shufflevector <8 x i32> %a0, <8 x i32> zeroinitializer, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
%y = freeze <8 x i32> %x
More information about the llvm-commits
mailing list