1 ; RUN: llc < %s -march=x86 -mcpu=nehalem -mattr=+sse4.2 | FileCheck %s 2 ; CHECK: pextrd 3 ; CHECK: pextrd 4 ; CHECK: movd 5 ; CHECK: movdqa 6 7 8 ; bitcast v14i16 to v7i32 9 10 define void @convert(<7 x i32>* %dst, <14 x i16>* %src) nounwind { 11 entry: 12 %dst.addr = alloca <7 x i32>* ; <<7 x i32>**> [#uses=2] 13 %src.addr = alloca <14 x i16>* ; <<14 x i16>**> [#uses=2] 14 %i = alloca i32, align 4 ; <i32*> [#uses=6] 15 store <7 x i32>* %dst, <7 x i32>** %dst.addr 16 store <14 x i16>* %src, <14 x i16>** %src.addr 17 store i32 0, i32* %i 18 br label %forcond 19 20 forcond: ; preds = %forinc, %entry 21 %tmp = load i32, i32* %i ; <i32> [#uses=1] 22 %cmp = icmp slt i32 %tmp, 4 ; <i1> [#uses=1] 23 br i1 %cmp, label %forbody, label %afterfor 24 25 forbody: ; preds = %forcond 26 %tmp1 = load i32, i32* %i ; <i32> [#uses=1] 27 %tmp2 = load <7 x i32>*, <7 x i32>** %dst.addr ; <<2 x i32>*> [#uses=1] 28 %arrayidx = getelementptr <7 x i32>, <7 x i32>* %tmp2, i32 %tmp1 ; <<7 x i32>*> [#uses=1] 29 %tmp3 = load i32, i32* %i ; <i32> [#uses=1] 30 %tmp4 = load <14 x i16>*, <14 x i16>** %src.addr ; <<4 x i16>*> [#uses=1] 31 %arrayidx5 = getelementptr <14 x i16>, <14 x i16>* %tmp4, i32 %tmp3 ; <<4 x i16>*> [#uses=1] 32 %tmp6 = load <14 x i16>, <14 x i16>* %arrayidx5 ; <<4 x i16>> [#uses=1] 33 %add = add <14 x i16> %tmp6, < i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1 > ; <<4 x i16>> [#uses=1] 34 %conv = bitcast <14 x i16> %add to <7 x i32> ; <<7 x i32>> [#uses=1] 35 store <7 x i32> %conv, <7 x i32>* %arrayidx 36 br label %forinc 37 38 forinc: ; preds = %forbody 39 %tmp7 = load i32, i32* %i ; <i32> [#uses=1] 40 %inc = add i32 %tmp7, 1 ; <i32> [#uses=1] 41 store i32 %inc, i32* %i 42 br label %forcond 43 44 afterfor: ; preds = %forcond 45 ret void 46 } 47
