1; RUN: llc < %s -mcpu=generic -march=x86 -mattr=+sse4.2 -post-RA-scheduler=true | FileCheck %s 2; CHECK: paddd 3 4; Widen a v3i16 to v8i16 to do a vector add 5 6@.str = internal constant [4 x i8] c"%d \00" ; <[4 x i8]*> [#uses=1] 7@.str1 = internal constant [2 x i8] c"\0A\00" ; <[2 x i8]*> [#uses=1] 8 9define void @update(<3 x i16>* %dst, <3 x i16>* %src, i32 %n) nounwind { 10entry: 11 %dst.addr = alloca <3 x i16>* ; <<3 x i16>**> [#uses=2] 12 %src.addr = alloca <3 x i16>* ; <<3 x i16>**> [#uses=2] 13 %n.addr = alloca i32 ; <i32*> [#uses=2] 14 %v = alloca <3 x i16>, align 8 ; <<3 x i16>*> [#uses=1] 15 %i = alloca i32, align 4 ; <i32*> [#uses=6] 16 store <3 x i16>* %dst, <3 x i16>** %dst.addr 17 store <3 x i16>* %src, <3 x i16>** %src.addr 18 store i32 %n, i32* %n.addr 19 store <3 x i16> < i16 1, i16 1, i16 1 >, <3 x i16>* %v 20 store i32 0, i32* %i 21 br label %forcond 22 23forcond: ; preds = %forinc, %entry 24 %tmp = load i32, i32* %i ; <i32> [#uses=1] 25 %tmp1 = load i32, i32* %n.addr ; <i32> [#uses=1] 26 %cmp = icmp slt i32 %tmp, %tmp1 ; <i1> [#uses=1] 27 br i1 %cmp, label %forbody, label %afterfor 28 29forbody: ; preds = %forcond 30 %tmp2 = load i32, i32* %i ; <i32> [#uses=1] 31 %tmp3 = load <3 x i16>*, <3 x i16>** %dst.addr ; <<3 x i16>*> [#uses=1] 32 %arrayidx = getelementptr <3 x i16>, <3 x i16>* %tmp3, i32 %tmp2 ; <<3 x i16>*> [#uses=1] 33 %tmp4 = load i32, i32* %i ; <i32> [#uses=1] 34 %tmp5 = load <3 x i16>*, <3 x i16>** %src.addr ; <<3 x i16>*> [#uses=1] 35 %arrayidx6 = getelementptr <3 x i16>, <3 x i16>* %tmp5, i32 %tmp4 ; <<3 x i16>*> [#uses=1] 36 %tmp7 = load <3 x i16>, <3 x i16>* %arrayidx6 ; <<3 x i16>> [#uses=1] 37 %add = add <3 x i16> %tmp7, < i16 1, i16 1, i16 1 > ; <<3 x i16>> [#uses=1] 38 store <3 x i16> %add, <3 x i16>* %arrayidx 39 br label %forinc 40 41forinc: ; preds = %forbody 42 %tmp8 = load i32, i32* %i ; <i32> [#uses=1] 43 %inc = add i32 %tmp8, 1 ; <i32> [#uses=1] 44 store i32 %inc, i32* %i 45 br label %forcond 46 47afterfor: ; preds = %forcond 48 ret void 49} 50 51