1 ; RUN: llc < %s -march=x86 -mattr=+sse42 -disable-mmx -o %t
2 ; RUN: grep paddb %t | count 1
3 ; RUN: grep pextrb %t | count 1
4 ; RUN: not grep pextrw %t
6 ; Widen a v3i8 to v16i8 to use a vector add
8 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:128:128"
10 define void @update(<3 x i8>* %dst, <3 x i8>* %src, i32 %n) nounwind {
12 %dst.addr = alloca <3 x i8>* ; <<3 x i8>**> [#uses=2]
13 %src.addr = alloca <3 x i8>* ; <<3 x i8>**> [#uses=2]
14 %n.addr = alloca i32 ; <i32*> [#uses=2]
15 %i = alloca i32, align 4 ; <i32*> [#uses=6]
16 store <3 x i8>* %dst, <3 x i8>** %dst.addr
17 store <3 x i8>* %src, <3 x i8>** %src.addr
18 store i32 %n, i32* %n.addr
22 forcond: ; preds = %forinc, %entry
23 %tmp = load i32* %i ; <i32> [#uses=1]
24 %tmp1 = load i32* %n.addr ; <i32> [#uses=1]
25 %cmp = icmp slt i32 %tmp, %tmp1 ; <i1> [#uses=1]
26 br i1 %cmp, label %forbody, label %afterfor
28 forbody: ; preds = %forcond
29 %tmp2 = load i32* %i ; <i32> [#uses=1]
30 %tmp3 = load <3 x i8>** %dst.addr ; <<3 x i8>*> [#uses=1]
31 %arrayidx = getelementptr <3 x i8>* %tmp3, i32 %tmp2 ; <<3 x i8>*> [#uses=1]
32 %tmp4 = load i32* %i ; <i32> [#uses=1]
33 %tmp5 = load <3 x i8>** %src.addr ; <<3 x i8>*> [#uses=1]
34 %arrayidx6 = getelementptr <3 x i8>* %tmp5, i32 %tmp4 ; <<3 x i8>*> [#uses=1]
35 %tmp7 = load <3 x i8>* %arrayidx6 ; <<3 x i8>> [#uses=1]
36 %add = add <3 x i8> %tmp7, < i8 1, i8 1, i8 1 > ; <<3 x i8>> [#uses=1]
37 store <3 x i8> %add, <3 x i8>* %arrayidx
40 forinc: ; preds = %forbody
41 %tmp8 = load i32* %i ; <i32> [#uses=1]
42 %inc = add i32 %tmp8, 1 ; <i32> [#uses=1]
43 store i32 %inc, i32* %i
46 afterfor: ; preds = %forcond