1 ; RUN: llc -march=x86 -mattr=+sse42 < %s | FileCheck %s 2 ; CHECK: paddw 3 ; CHECK: pextrd 4 ; CHECK: movd 5 6 ; bitcast a v4i16 to v2i32 7 8 define void @convert(<2 x i32>* %dst, <4 x i16>* %src) nounwind { 9 entry: 10 %dst.addr = alloca <2 x i32>* ; <<2 x i32>**> [#uses=2] 11 %src.addr = alloca <4 x i16>* ; <<4 x i16>**> [#uses=2] 12 %i = alloca i32, align 4 ; <i32*> [#uses=6] 13 store <2 x i32>* %dst, <2 x i32>** %dst.addr 14 store <4 x i16>* %src, <4 x i16>** %src.addr 15 store i32 0, i32* %i 16 br label %forcond 17 18 forcond: ; preds = %forinc, %entry 19 %tmp = load i32* %i ; <i32> [#uses=1] 20 %cmp = icmp slt i32 %tmp, 4 ; <i1> [#uses=1] 21 br i1 %cmp, label %forbody, label %afterfor 22 23 forbody: ; preds = %forcond 24 %tmp1 = load i32* %i ; <i32> [#uses=1] 25 %tmp2 = load <2 x i32>** %dst.addr ; <<2 x i32>*> [#uses=1] 26 %arrayidx = getelementptr <2 x i32>* %tmp2, i32 %tmp1 ; <<2 x i32>*> [#uses=1] 27 %tmp3 = load i32* %i ; <i32> [#uses=1] 28 %tmp4 = load <4 x i16>** %src.addr ; <<4 x i16>*> [#uses=1] 29 %arrayidx5 = getelementptr <4 x i16>* %tmp4, i32 %tmp3 ; <<4 x i16>*> [#uses=1] 30 %tmp6 = load <4 x i16>* %arrayidx5 ; <<4 x i16>> [#uses=1] 31 %add = add <4 x i16> %tmp6, < i16 1, i16 1, i16 1, i16 1 > ; <<4 x i16>> [#uses=1] 32 %conv = bitcast <4 x i16> %add to <2 x i32> ; <<2 x i32>> [#uses=1] 33 store <2 x i32> %conv, <2 x i32>* %arrayidx 34 br label %forinc 35 36 forinc: ; preds = %forbody 37 %tmp7 = load i32* %i ; <i32> [#uses=1] 38 %inc = add i32 %tmp7, 1 ; <i32> [#uses=1] 39 store i32 %inc, i32* %i 40 br label %forcond 41 42 afterfor: ; preds = %forcond 43 ret void 44 } 45