1 ; RUN: llc < %s -march=x86 -mattr=+sse4.2 | FileCheck %s 2 ; CHECK: padd 3 ; CHECK: pand 4 5 ; widen v8i8 to v16i8 (checks even power of 2 widening with add & and) 6 7 define void @update(i64* %dst_i, i64* %src_i, i32 %n) nounwind { 8 entry: 9 %dst_i.addr = alloca i64* ; <i64**> [#uses=2] 10 %src_i.addr = alloca i64* ; <i64**> [#uses=2] 11 %n.addr = alloca i32 ; <i32*> [#uses=2] 12 %i = alloca i32, align 4 ; <i32*> [#uses=8] 13 %dst = alloca <8 x i8>*, align 4 ; <<8 x i8>**> [#uses=2] 14 %src = alloca <8 x i8>*, align 4 ; <<8 x i8>**> [#uses=2] 15 store i64* %dst_i, i64** %dst_i.addr 16 store i64* %src_i, i64** %src_i.addr 17 store i32 %n, i32* %n.addr 18 store i32 0, i32* %i 19 br label %forcond 20 21 forcond: ; preds = %forinc, %entry 22 %tmp = load i32, i32* %i ; <i32> [#uses=1] 23 %tmp1 = load i32, i32* %n.addr ; <i32> [#uses=1] 24 %cmp = icmp slt i32 %tmp, %tmp1 ; <i1> [#uses=1] 25 br i1 %cmp, label %forbody, label %afterfor 26 27 forbody: ; preds = %forcond 28 %tmp2 = load i32, i32* %i ; <i32> [#uses=1] 29 %tmp3 = load i64*, i64** %dst_i.addr ; <i64*> [#uses=1] 30 %arrayidx = getelementptr i64, i64* %tmp3, i32 %tmp2 ; <i64*> [#uses=1] 31 %conv = bitcast i64* %arrayidx to <8 x i8>* ; <<8 x i8>*> [#uses=1] 32 store <8 x i8>* %conv, <8 x i8>** %dst 33 %tmp4 = load i32, i32* %i ; <i32> [#uses=1] 34 %tmp5 = load i64*, i64** %src_i.addr ; <i64*> [#uses=1] 35 %arrayidx6 = getelementptr i64, i64* %tmp5, i32 %tmp4 ; <i64*> [#uses=1] 36 %conv7 = bitcast i64* %arrayidx6 to <8 x i8>* ; <<8 x i8>*> [#uses=1] 37 store <8 x i8>* %conv7, <8 x i8>** %src 38 %tmp8 = load i32, i32* %i ; <i32> [#uses=1] 39 %tmp9 = load <8 x i8>*, <8 x i8>** %dst ; <<8 x i8>*> [#uses=1] 40 %arrayidx10 = getelementptr <8 x i8>, <8 x i8>* %tmp9, i32 %tmp8 ; <<8 x i8>*> [#uses=1] 41 %tmp11 = load i32, i32* %i ; <i32> [#uses=1] 42 %tmp12 = load <8 x i8>*, <8 x i8>** %src ; <<8 x i8>*> [#uses=1] 43 %arrayidx13 = getelementptr <8 x i8>, <8 x i8>* %tmp12, i32 %tmp11 ; <<8 x i8>*> [#uses=1] 44 %tmp14 = load <8 x i8>, <8 x i8>* %arrayidx13 ; <<8 x i8>> [#uses=1] 45 %add = add <8 x i8> %tmp14, < i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1 > ; <<8 x i8>> [#uses=1] 46 %and = and <8 x i8> %add, < i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4 > ; <<8 x i8>> [#uses=1] 47 store <8 x i8> %and, <8 x i8>* %arrayidx10 48 br label %forinc 49 50 forinc: ; preds = %forbody 51 %tmp15 = load i32, i32* %i ; <i32> [#uses=1] 52 %inc = add i32 %tmp15, 1 ; <i32> [#uses=1] 53 store i32 %inc, i32* %i 54 br label %forcond 55 56 afterfor: ; preds = %forcond 57 ret void 58 } 59 60
