1 ; RUN: llc < %s -mcpu=generic -march=x86 -mattr=+sse42 -post-RA-scheduler=true | FileCheck %s
7 ; Widen a v3i16 to v8i16 to do a vector add
9 @.str = internal constant [4 x i8] c"%d \00" ; <[4 x i8]*> [#uses=1]
10 @.str1 = internal constant [2 x i8] c"\0A\00" ; <[2 x i8]*> [#uses=1]
12 define void @update(<3 x i16>* %dst, <3 x i16>* %src, i32 %n) nounwind {
14 %dst.addr = alloca <3 x i16>* ; <<3 x i16>**> [#uses=2]
15 %src.addr = alloca <3 x i16>* ; <<3 x i16>**> [#uses=2]
16 %n.addr = alloca i32 ; <i32*> [#uses=2]
17 %v = alloca <3 x i16>, align 8 ; <<3 x i16>*> [#uses=1]
18 %i = alloca i32, align 4 ; <i32*> [#uses=6]
19 store <3 x i16>* %dst, <3 x i16>** %dst.addr
20 store <3 x i16>* %src, <3 x i16>** %src.addr
21 store i32 %n, i32* %n.addr
22 store <3 x i16> < i16 1, i16 1, i16 1 >, <3 x i16>* %v
26 forcond: ; preds = %forinc, %entry
27 %tmp = load i32* %i ; <i32> [#uses=1]
28 %tmp1 = load i32* %n.addr ; <i32> [#uses=1]
29 %cmp = icmp slt i32 %tmp, %tmp1 ; <i1> [#uses=1]
30 br i1 %cmp, label %forbody, label %afterfor
32 forbody: ; preds = %forcond
33 %tmp2 = load i32* %i ; <i32> [#uses=1]
34 %tmp3 = load <3 x i16>** %dst.addr ; <<3 x i16>*> [#uses=1]
35 %arrayidx = getelementptr <3 x i16>* %tmp3, i32 %tmp2 ; <<3 x i16>*> [#uses=1]
36 %tmp4 = load i32* %i ; <i32> [#uses=1]
37 %tmp5 = load <3 x i16>** %src.addr ; <<3 x i16>*> [#uses=1]
38 %arrayidx6 = getelementptr <3 x i16>* %tmp5, i32 %tmp4 ; <<3 x i16>*> [#uses=1]
39 %tmp7 = load <3 x i16>* %arrayidx6 ; <<3 x i16>> [#uses=1]
40 %add = add <3 x i16> %tmp7, < i16 1, i16 1, i16 1 > ; <<3 x i16>> [#uses=1]
41 store <3 x i16> %add, <3 x i16>* %arrayidx
44 forinc: ; preds = %forbody
45 %tmp8 = load i32* %i ; <i32> [#uses=1]
46 %inc = add i32 %tmp8, 1 ; <i32> [#uses=1]
47 store i32 %inc, i32* %i
50 afterfor: ; preds = %forcond