1; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py 2; RUN: llc < %s -mtriple=i686-unknown-unknown -mattr=+sse4.2 | FileCheck %s 3 4; widen v8i8 to v16i8 (checks even power of 2 widening with add & and) 5 6define void @update(i64* %dst_i, i64* %src_i, i32 %n) nounwind { 7; CHECK-LABEL: update: 8; CHECK: # %bb.0: # %entry 9; CHECK-NEXT: subl $12, %esp 10; CHECK-NEXT: movl $0, (%esp) 11; CHECK-NEXT: pcmpeqd %xmm0, %xmm0 12; CHECK-NEXT: movdqa {{.*#+}} xmm1 = [4,4,4,4,4,4,4,4] 13; CHECK-NEXT: jmp .LBB0_1 14; CHECK-NEXT: .p2align 4, 0x90 15; CHECK-NEXT: .LBB0_2: # %forbody 16; CHECK-NEXT: # in Loop: Header=BB0_1 Depth=1 17; CHECK-NEXT: movl (%esp), %eax 18; CHECK-NEXT: leal (,%eax,8), %ecx 19; CHECK-NEXT: movl {{[0-9]+}}(%esp), %edx 20; CHECK-NEXT: addl %ecx, %edx 21; CHECK-NEXT: movl %edx, {{[0-9]+}}(%esp) 22; CHECK-NEXT: addl {{[0-9]+}}(%esp), %ecx 23; CHECK-NEXT: movl %ecx, {{[0-9]+}}(%esp) 24; CHECK-NEXT: pmovzxbw {{.*#+}} xmm2 = mem[0],zero,mem[1],zero,mem[2],zero,mem[3],zero,mem[4],zero,mem[5],zero,mem[6],zero,mem[7],zero 25; CHECK-NEXT: psubw %xmm0, %xmm2 26; CHECK-NEXT: pand %xmm1, %xmm2 27; CHECK-NEXT: packuswb %xmm0, %xmm2 28; CHECK-NEXT: movq %xmm2, (%edx,%eax,8) 29; CHECK-NEXT: incl (%esp) 30; CHECK-NEXT: .LBB0_1: # %forcond 31; CHECK-NEXT: # =>This Inner Loop Header: Depth=1 32; CHECK-NEXT: movl (%esp), %eax 33; CHECK-NEXT: cmpl {{[0-9]+}}(%esp), %eax 34; CHECK-NEXT: jl .LBB0_2 35; CHECK-NEXT: # %bb.3: # %afterfor 36; CHECK-NEXT: addl $12, %esp 37; CHECK-NEXT: retl 38entry: 39 %dst_i.addr = alloca i64* 40 %src_i.addr = alloca i64* 41 %n.addr = alloca i32 42 %i = alloca i32, align 4 43 %dst = alloca <8 x i8>*, align 4 44 %src = alloca <8 x i8>*, align 4 45 store i64* %dst_i, i64** %dst_i.addr 46 store i64* %src_i, i64** %src_i.addr 47 store i32 %n, i32* %n.addr 48 store i32 0, i32* %i 49 br label %forcond 50 51forcond: 52 %tmp = load i32, i32* %i 53 %tmp1 = load i32, i32* %n.addr 54 %cmp = icmp slt i32 %tmp, %tmp1 55 br i1 %cmp, label %forbody, label %afterfor 56 57forbody: 58 %tmp2 = load i32, i32* %i 59 %tmp3 = load i64*, i64** %dst_i.addr 60 %arrayidx = getelementptr i64, i64* %tmp3, i32 %tmp2 61 %conv = bitcast i64* %arrayidx to <8 x i8>* 62 store <8 x i8>* %conv, <8 x i8>** %dst 63 %tmp4 = load i32, i32* %i 64 %tmp5 = load i64*, i64** %src_i.addr 65 %arrayidx6 = getelementptr i64, i64* %tmp5, i32 %tmp4 66 %conv7 = bitcast i64* %arrayidx6 to <8 x i8>* 67 store <8 x i8>* %conv7, <8 x i8>** %src 68 %tmp8 = load i32, i32* %i 69 %tmp9 = load <8 x i8>*, <8 x i8>** %dst 70 %arrayidx10 = getelementptr <8 x i8>, <8 x i8>* %tmp9, i32 %tmp8 71 %tmp11 = load i32, i32* %i 72 %tmp12 = load <8 x i8>*, <8 x i8>** %src 73 %arrayidx13 = getelementptr <8 x i8>, <8 x i8>* %tmp12, i32 %tmp11 74 %tmp14 = load <8 x i8>, <8 x i8>* %arrayidx13 75 %add = add <8 x i8> %tmp14, < i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1 > 76 %and = and <8 x i8> %add, < i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4, i8 4 > 77 store <8 x i8> %and, <8 x i8>* %arrayidx10 78 br label %forinc 79 80forinc: 81 %tmp15 = load i32, i32* %i 82 %inc = add i32 %tmp15, 1 83 store i32 %inc, i32* %i 84 br label %forcond 85 86afterfor: 87 ret void 88} 89 90