mirror of
https://github.com/RPCS3/llvm-mirror.git
synced 2024-10-31 16:02:52 +01:00
8cb4abbc0e
insertp[sd] $0, which is a noop. Before: _f32: ## @f32 pshufd $1, %xmm1, %xmm2 pshufd $1, %xmm0, %xmm3 addss %xmm2, %xmm3 addss %xmm1, %xmm0 ## kill: XMM0<def> XMM0<kill> XMM0<def> insertps $0, %xmm0, %xmm0 insertps $16, %xmm3, %xmm0 ret after: _f32: ## @f32 movdqa %xmm0, %xmm2 addss %xmm1, %xmm2 pshufd $1, %xmm1, %xmm1 pshufd $1, %xmm0, %xmm3 addss %xmm1, %xmm3 movdqa %xmm2, %xmm0 insertps $16, %xmm3, %xmm0 ret The extra movs are due to a random (poor) scheduling decision. llvm-svn: 112379
10 lines
347 B
LLVM
10 lines
347 B
LLVM
; RUN: llc < %s -march=x86 -mattr=+sse41 > %t
|
|
; RUN: grep pinsrd %t | count 1
|
|
|
|
define <4 x i32> @var_insert2(<4 x i32> %x, i32 %val, i32 %idx) nounwind {
|
|
entry:
|
|
%tmp3 = insertelement <4 x i32> undef, i32 %val, i32 0 ; <<4 x i32>> [#uses=1]
|
|
%tmp4 = insertelement <4 x i32> %tmp3, i32 %idx, i32 3 ; <<4 x i32>> [#uses=1]
|
|
ret <4 x i32> %tmp4
|
|
}
|