1 ; RUN: llvm-as < %s | llc -march=x86 -mattr=+sse42 -disable-mmx -o %t -f
2 ; RUN: grep pmulld %t | count 1
3 ; RUN: grep psubd %t | count 1
4 ; RUN: grep movaps %t | count 1
6 ; widen a v3i32 to v4i32 to do a vector multiple and a subtraction
7 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"
9 define void @update(<3 x i32>* %dst, <3 x i32>* %src, i32 %n) nounwind {
11 %dst.addr = alloca <3 x i32>* ; <<3 x i32>**> [#uses=2]
12 %src.addr = alloca <3 x i32>* ; <<3 x i32>**> [#uses=2]
13 %n.addr = alloca i32 ; <i32*> [#uses=2]
14 %v = alloca <3 x i32>, align 16 ; <<3 x i32>*> [#uses=1]
15 %i = alloca i32, align 4 ; <i32*> [#uses=6]
16 store <3 x i32>* %dst, <3 x i32>** %dst.addr
17 store <3 x i32>* %src, <3 x i32>** %src.addr
18 store i32 %n, i32* %n.addr
19 store <3 x i32> < i32 1, i32 1, i32 1 >, <3 x i32>* %v
23 forcond: ; preds = %forinc, %entry
24 %tmp = load i32* %i ; <i32> [#uses=1]
25 %tmp1 = load i32* %n.addr ; <i32> [#uses=1]
26 %cmp = icmp slt i32 %tmp, %tmp1 ; <i1> [#uses=1]
27 br i1 %cmp, label %forbody, label %afterfor
29 forbody: ; preds = %forcond
30 %tmp2 = load i32* %i ; <i32> [#uses=1]
31 %tmp3 = load <3 x i32>** %dst.addr ; <<3 x i32>*> [#uses=1]
32 %arrayidx = getelementptr <3 x i32>* %tmp3, i32 %tmp2 ; <<3 x i32>*> [#uses=1]
33 %tmp4 = load i32* %i ; <i32> [#uses=1]
34 %tmp5 = load <3 x i32>** %src.addr ; <<3 x i32>*> [#uses=1]
35 %arrayidx6 = getelementptr <3 x i32>* %tmp5, i32 %tmp4 ; <<3 x i32>*> [#uses=1]
36 %tmp7 = load <3 x i32>* %arrayidx6 ; <<3 x i32>> [#uses=1]
37 %mul = mul <3 x i32> %tmp7, < i32 4, i32 4, i32 4 > ; <<3 x i32>> [#uses=1]
38 %sub = sub <3 x i32> %mul, < i32 3, i32 3, i32 3 > ; <<3 x i32>> [#uses=1]
39 store <3 x i32> %sub, <3 x i32>* %arrayidx
42 forinc: ; preds = %forbody
43 %tmp8 = load i32* %i ; <i32> [#uses=1]
44 %inc = add i32 %tmp8, 1 ; <i32> [#uses=1]
45 store i32 %inc, i32* %i
48 afterfor: ; preds = %forcond