1; RUN: llc < %s -march=x86 -mattr=+sse42 | FileCheck %s 2 3; Widen a v3i8 to v16i8 to use a vector add 4 5define void @update(<3 x i8>* %dst, <3 x i8>* %src, i32 %n) nounwind { 6entry: 7; CHECK-NOT: pextrw 8; CHECK: paddb 9; CHECK: pextrb 10 %dst.addr = alloca <3 x i8>* ; <<3 x i8>**> [#uses=2] 11 %src.addr = alloca <3 x i8>* ; <<3 x i8>**> [#uses=2] 12 %n.addr = alloca i32 ; <i32*> [#uses=2] 13 %i = alloca i32, align 4 ; <i32*> [#uses=6] 14 store <3 x i8>* %dst, <3 x i8>** %dst.addr 15 store <3 x i8>* %src, <3 x i8>** %src.addr 16 store i32 %n, i32* %n.addr 17 store i32 0, i32* %i 18 br label %forcond 19 20forcond: ; preds = %forinc, %entry 21 %tmp = load i32* %i ; <i32> [#uses=1] 22 %tmp1 = load i32* %n.addr ; <i32> [#uses=1] 23 %cmp = icmp slt i32 %tmp, %tmp1 ; <i1> [#uses=1] 24 br i1 %cmp, label %forbody, label %afterfor 25 26forbody: ; preds = %forcond 27 %tmp2 = load i32* %i ; <i32> [#uses=1] 28 %tmp3 = load <3 x i8>** %dst.addr ; <<3 x i8>*> [#uses=1] 29 %arrayidx = getelementptr <3 x i8>* %tmp3, i32 %tmp2 ; <<3 x i8>*> [#uses=1] 30 %tmp4 = load i32* %i ; <i32> [#uses=1] 31 %tmp5 = load <3 x i8>** %src.addr ; <<3 x i8>*> [#uses=1] 32 %arrayidx6 = getelementptr <3 x i8>* %tmp5, i32 %tmp4 ; <<3 x i8>*> [#uses=1] 33 %tmp7 = load <3 x i8>* %arrayidx6 ; <<3 x i8>> [#uses=1] 34 %add = add <3 x i8> %tmp7, < i8 1, i8 1, i8 1 > ; <<3 x i8>> [#uses=1] 35 store <3 x i8> %add, <3 x i8>* %arrayidx 36 br label %forinc 37 38forinc: ; preds = %forbody 39 %tmp8 = load i32* %i ; <i32> [#uses=1] 40 %inc = add i32 %tmp8, 1 ; <i32> [#uses=1] 41 store i32 %inc, i32* %i 42 br label %forcond 43 44afterfor: ; preds = %forcond 45 ret void 46} 47