http://llvm.org/bugs/show_bug.cgi?id=21138
Bug ID: 21138
Summary: New vector shuffle lowering produces an insertf128 +
blend instead of a single insertf128
Product: libraries
Version: trunk
Hardware: PC
OS: All
Status: NEW
Severity: normal
Priority: P
Component: Backend: X86
Assignee: [email protected]
Reporter: [email protected]
CC: [email protected]
Classification: Unclassified
Created attachment 13122
--> http://llvm.org/bugs/attachment.cgi?id=13122&action=edit
IR to reproduce the problem.
To reproduce:
llc -mattr=+avx -x86-experimental-vector-shuffle-lowering=true
avx_insertf_blendps.ll -o new.s
llc -mattr=+avx -x86-experimental-vector-shuffle-lowering=false
avx_insertf_blendps.ll -o old.s
diff -U 10 old.s new.s
_foo: ## @foo
.cfi_startproc
## BB#0: ## %for.body545
vextractf128 $1, %ymm1, %xmm2
vshufps $-35, %xmm2, %xmm1, %xmm1 ## xmm1 = xmm1[1,3],xmm2[1,3]
+ vinsertf128 $1, %xmm1, %ymm0, %ymm1
vextractf128 $1, %ymm0, %xmm2
vshufps $-35, %xmm2, %xmm0, %xmm0 ## xmm0 = xmm0[1,3],xmm2[1,3]
- vinsertf128 $1, %xmm1, %ymm0, %ymm0
+ vblendps $-16, %ymm1, %ymm0, %ymm0 ## ymm0 =
ymm0[0,1,2,3],ymm1[4,5,6,7]
retq
--
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
LLVMbugs mailing list
[email protected]
http://lists.cs.uiuc.edu/mailman/listinfo/llvmbugs
