1 ; RUN: llc < %s -march=x86 -mattr=+sse42 -post-RA-scheduler=true | FileCheck %s 2 ; CHECK: paddw 3 ; CHECK: pextrw 4 ; CHECK: movd 5 6 ; Widen a v3i16 to v8i16 to do a vector add 7 8 @.str = internal constant [4 x i8] c"%d \00" ; <[4 x i8]*> [#uses=1] 9 @.str1 = internal constant [2 x i8] c"\0A\00" ; <[2 x i8]*> [#uses=1] 10 11 define void @update(<3 x i16>* %dst, <3 x i16>* %src, i32 %n) nounwind { 12 entry: 13 %dst.addr = alloca <3 x i16>* ; <<3 x i16>**> [#uses=2] 14 %src.addr = alloca <3 x i16>* ; <<3 x i16>**> [#uses=2] 15 %n.addr = alloca i32 ; <i32*> [#uses=2] 16 %v = alloca <3 x i16>, align 8 ; <<3 x i16>*> [#uses=1] 17 %i = alloca i32, align 4 ; <i32*> [#uses=6] 18 store <3 x i16>* %dst, <3 x i16>** %dst.addr 19 store <3 x i16>* %src, <3 x i16>** %src.addr 20 store i32 %n, i32* %n.addr 21 store <3 x i16> < i16 1, i16 1, i16 1 >, <3 x i16>* %v 22 store i32 0, i32* %i 23 br label %forcond 24 25 forcond: ; preds = %forinc, %entry 26 %tmp = load i32* %i ; <i32> [#uses=1] 27 %tmp1 = load i32* %n.addr ; <i32> [#uses=1] 28 %cmp = icmp slt i32 %tmp, %tmp1 ; <i1> [#uses=1] 29 br i1 %cmp, label %forbody, label %afterfor 30 31 forbody: ; preds = %forcond 32 %tmp2 = load i32* %i ; <i32> [#uses=1] 33 %tmp3 = load <3 x i16>** %dst.addr ; <<3 x i16>*> [#uses=1] 34 %arrayidx = getelementptr <3 x i16>* %tmp3, i32 %tmp2 ; <<3 x i16>*> [#uses=1] 35 %tmp4 = load i32* %i ; <i32> [#uses=1] 36 %tmp5 = load <3 x i16>** %src.addr ; <<3 x i16>*> [#uses=1] 37 %arrayidx6 = getelementptr <3 x i16>* %tmp5, i32 %tmp4 ; <<3 x i16>*> [#uses=1] 38 %tmp7 = load <3 x i16>* %arrayidx6 ; <<3 x i16>> [#uses=1] 39 %add = add <3 x i16> %tmp7, < i16 1, i16 1, i16 1 > ; <<3 x i16>> [#uses=1] 40 store <3 x i16> %add, <3 x i16>* %arrayidx 41 br label %forinc 42 43 forinc: ; preds = %forbody 44 %tmp8 = load i32* %i ; <i32> [#uses=1] 45 %inc = add i32 %tmp8, 1 ; <i32> [#uses=1] 46 store i32 %inc, i32* %i 47 br label %forcond 48 49 afterfor: ; preds = %forcond 50 ret void 51 } 52 53
