1 ; RUN: llc < %s -march=x86 -mattr=+sse2 -mcpu=yonah | FileCheck %s
2 ; RUN: llc < %s -march=x86-64 -mattr=+sse2 -mcpu=core2 | FileCheck %s
4 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
6 define i32 @t(<2 x i64>* %val) nounwind {
11 %tmp2 = load <2 x i64>* %val, align 16 ; <<2 x i64>> [#uses=1]
12 %tmp3 = bitcast <2 x i64> %tmp2 to <4 x i32> ; <<4 x i32>> [#uses=1]
13 %tmp4 = extractelement <4 x i32> %tmp3, i32 2 ; <i32> [#uses=1]
17 ; Case where extractelement of load ends up as undef.
18 ; (Making sure this doesn't crash.)
19 define i32 @t2(<8 x i32>* %xp) {
22 %x = load <8 x i32>* %xp
23 %Shuff68 = shufflevector <8 x i32> %x, <8 x i32> undef, <8 x i32> <i32
24 undef, i32 7, i32 9, i32 undef, i32 13, i32 15, i32 1, i32 3>
25 %y = extractelement <8 x i32> %Shuff68, i32 0
29 ; This case could easily end up inf-looping in the DAG combiner due to an
30 ; low alignment load of the vector which prevents us from reliably forming a
32 ; FIXME: It would be nice to detect whether the target has fast and legal
33 ; unaligned loads and use them here.
37 ; This movs the entire vector, shuffling the high double down. If we fixed the
38 ; FIXME above it would just move the high double directly.
44 %tmp13 = load <2 x double>* undef, align 1
45 %.sroa.3.24.vec.extract = extractelement <2 x double> %tmp13, i32 1
46 store double %.sroa.3.24.vec.extract, double* undef, align 8