1 ; RUN: opt < %s -std-compile-opts | \ 2 ; RUN: llc -mtriple=thumbv7-apple-darwin10 -mattr=+neon | FileCheck %s 3 4 define void @fred(i32 %three_by_three, i8* %in, double %dt1, i32 %x_size, i32 %y_size, i8* %bp) nounwind { 5 entry: 6 ; -- The loop following the load should only use a single add-literation 7 ; instruction. 8 ; CHECK: ldr.64 9 ; CHECK: adds r{{[0-9]+.*}}#1 10 ; CHECK-NOT: adds 11 ; CHECK: subsections_via_symbols 12 13 14 %three_by_three_addr = alloca i32 ; <i32*> [#uses=2] 15 %in_addr = alloca i8* ; <i8**> [#uses=2] 16 %dt_addr = alloca float ; <float*> [#uses=4] 17 %x_size_addr = alloca i32 ; <i32*> [#uses=2] 18 %y_size_addr = alloca i32 ; <i32*> [#uses=1] 19 %bp_addr = alloca i8* ; <i8**> [#uses=1] 20 %tmp_image = alloca i8* ; <i8**> [#uses=0] 21 %out = alloca i8* ; <i8**> [#uses=1] 22 %cp = alloca i8* ; <i8**> [#uses=0] 23 %dpt = alloca i8* ; <i8**> [#uses=4] 24 %dp = alloca i8* ; <i8**> [#uses=2] 25 %ip = alloca i8* ; <i8**> [#uses=0] 26 %centre = alloca i32 ; <i32*> [#uses=0] 27 %tmp = alloca i32 ; <i32*> [#uses=0] 28 %brightness = alloca i32 ; <i32*> [#uses=0] 29 %area = alloca i32 ; <i32*> [#uses=0] 30 %y = alloca i32 ; <i32*> [#uses=0] 31 %x = alloca i32 ; <i32*> [#uses=2] 32 %j = alloca i32 ; <i32*> [#uses=6] 33 %i = alloca i32 ; <i32*> [#uses=1] 34 %mask_size = alloca i32 ; <i32*> [#uses=5] 35 %increment = alloca i32 ; <i32*> [#uses=1] 36 %n_max = alloca i32 ; <i32*> [#uses=4] 37 %temp = alloca float ; <float*> [#uses=1] 38 %"alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0] 39 store i32 %three_by_three, i32* %three_by_three_addr 40 store i8* %in, i8** %in_addr 41 %dt = fptrunc double %dt1 to float ; <float> [#uses=1] 42 store float %dt, float* %dt_addr 43 store i32 %x_size, i32* %x_size_addr 44 store i32 %y_size, i32* %y_size_addr 45 store i8* %bp, i8** %bp_addr 46 %0 = load i8** %in_addr, align 4 ; <i8*> [#uses=1] 47 store i8* %0, i8** %out, align 4 48 %1 = call i32 (...)* @foo() nounwind ; <i32> [#uses=1] 49 store i32 %1, i32* %i, align 4 50 %2 = load i32* %three_by_three_addr, align 4 ; <i32> [#uses=1] 51 %3 = icmp eq i32 %2, 0 ; <i1> [#uses=1] 52 br i1 %3, label %bb, label %bb2 53 54 bb: ; preds = %entry 55 %4 = load float* %dt_addr, align 4 ; <float> [#uses=1] 56 %5 = fpext float %4 to double ; <double> [#uses=1] 57 %6 = fmul double %5, 1.500000e+00 ; <double> [#uses=1] 58 %7 = fptosi double %6 to i32 ; <i32> [#uses=1] 59 %8 = add nsw i32 %7, 1 ; <i32> [#uses=1] 60 store i32 %8, i32* %mask_size, align 4 61 br label %bb3 62 63 bb2: ; preds = %entry 64 store i32 1, i32* %mask_size, align 4 65 br label %bb3 66 67 bb3: ; preds = %bb2, %bb 68 %9 = load i32* %mask_size, align 4 ; <i32> [#uses=1] 69 %10 = mul i32 %9, 2 ; <i32> [#uses=1] 70 %11 = add nsw i32 %10, 1 ; <i32> [#uses=1] 71 store i32 %11, i32* %n_max, align 4 72 %12 = load i32* %x_size_addr, align 4 ; <i32> [#uses=1] 73 %13 = load i32* %n_max, align 4 ; <i32> [#uses=1] 74 %14 = sub i32 %12, %13 ; <i32> [#uses=1] 75 store i32 %14, i32* %increment, align 4 76 %15 = load i32* %n_max, align 4 ; <i32> [#uses=1] 77 %16 = load i32* %n_max, align 4 ; <i32> [#uses=1] 78 %17 = mul i32 %15, %16 ; <i32> [#uses=1] 79 %18 = call noalias i8* @malloc(i32 %17) nounwind ; <i8*> [#uses=1] 80 store i8* %18, i8** %dp, align 4 81 %19 = load i8** %dp, align 4 ; <i8*> [#uses=1] 82 store i8* %19, i8** %dpt, align 4 83 %20 = load float* %dt_addr, align 4 ; <float> [#uses=1] 84 %21 = load float* %dt_addr, align 4 ; <float> [#uses=1] 85 %22 = fmul float %20, %21 ; <float> [#uses=1] 86 %23 = fsub float -0.000000e+00, %22 ; <float> [#uses=1] 87 store float %23, float* %temp, align 4 88 %24 = load i32* %mask_size, align 4 ; <i32> [#uses=1] 89 %25 = sub i32 0, %24 ; <i32> [#uses=1] 90 store i32 %25, i32* %j, align 4 91 br label %bb5 92 93 bb4: ; preds = %bb5 94 %26 = load i32* %j, align 4 ; <i32> [#uses=1] 95 %27 = load i32* %j, align 4 ; <i32> [#uses=1] 96 %28 = mul i32 %26, %27 ; <i32> [#uses=1] 97 %29 = sitofp i32 %28 to double ; <double> [#uses=1] 98 %30 = fmul double %29, 1.234000e+00 ; <double> [#uses=1] 99 %31 = fptosi double %30 to i32 ; <i32> [#uses=1] 100 store i32 %31, i32* %x, align 4 101 %32 = load i32* %x, align 4 ; <i32> [#uses=1] 102 %33 = trunc i32 %32 to i8 ; <i8> [#uses=1] 103 %34 = load i8** %dpt, align 4 ; <i8*> [#uses=1] 104 store i8 %33, i8* %34, align 1 105 %35 = load i8** %dpt, align 4 ; <i8*> [#uses=1] 106 %36 = getelementptr inbounds i8* %35, i64 1 ; <i8*> [#uses=1] 107 store i8* %36, i8** %dpt, align 4 108 %37 = load i32* %j, align 4 ; <i32> [#uses=1] 109 %38 = add nsw i32 %37, 1 ; <i32> [#uses=1] 110 store i32 %38, i32* %j, align 4 111 br label %bb5 112 113 bb5: ; preds = %bb4, %bb3 114 %39 = load i32* %j, align 4 ; <i32> [#uses=1] 115 %40 = load i32* %mask_size, align 4 ; <i32> [#uses=1] 116 %41 = icmp sle i32 %39, %40 ; <i1> [#uses=1] 117 br i1 %41, label %bb4, label %bb6 118 119 bb6: ; preds = %bb5 120 br label %return 121 122 return: ; preds = %bb6 123 ret void 124 } 125 126 declare i32 @foo(...) 127 128 declare noalias i8* @malloc(i32) nounwind 129
