1 ; RUN: opt < %s -O3 | \ 2 ; RUN: llc -mtriple=thumbv7-apple-darwin10 -mattr=+neon | FileCheck %s 3 4 target datalayout = "e-p:32:32:32-i1:8:32-i8:8:32-i16:16:32-i32:32:32-i64:32:32-f32:32:32-f64:32:32-v64:64:64-v128:128:128-a0:0:32" 5 6 define void @fred(i32 %three_by_three, i8* %in, double %dt1, i32 %x_size, i32 %y_size, i8* %bp) nounwind { 7 entry: 8 ; -- The loop following the load should only use a single add-literation 9 ; instruction. 10 ; CHECK: vldr 11 ; CHECK-NOT: adds 12 ; CHECK: subsections_via_symbols 13 14 15 %three_by_three_addr = alloca i32 ; <i32*> [#uses=2] 16 %in_addr = alloca i8* ; <i8**> [#uses=2] 17 %dt_addr = alloca float ; <float*> [#uses=4] 18 %x_size_addr = alloca i32 ; <i32*> [#uses=2] 19 %y_size_addr = alloca i32 ; <i32*> [#uses=1] 20 %bp_addr = alloca i8* ; <i8**> [#uses=1] 21 %tmp_image = alloca i8* ; <i8**> [#uses=0] 22 %out = alloca i8* ; <i8**> [#uses=1] 23 %cp = alloca i8* ; <i8**> [#uses=0] 24 %dpt = alloca i8* ; <i8**> [#uses=4] 25 %dp = alloca i8* ; <i8**> [#uses=2] 26 %ip = alloca i8* ; <i8**> [#uses=0] 27 %centre = alloca i32 ; <i32*> [#uses=0] 28 %tmp = alloca i32 ; <i32*> [#uses=0] 29 %brightness = alloca i32 ; <i32*> [#uses=0] 30 %area = alloca i32 ; <i32*> [#uses=0] 31 %y = alloca i32 ; <i32*> [#uses=0] 32 %x = alloca i32 ; <i32*> [#uses=2] 33 %j = alloca i32 ; <i32*> [#uses=6] 34 %i = alloca i32 ; <i32*> [#uses=1] 35 %mask_size = alloca i32 ; <i32*> [#uses=5] 36 %increment = alloca i32 ; <i32*> [#uses=1] 37 %n_max = alloca i32 ; <i32*> [#uses=4] 38 %temp = alloca float ; <float*> [#uses=1] 39 %"alloca point" = bitcast i32 0 to i32 ; <i32> [#uses=0] 40 store i32 %three_by_three, i32* %three_by_three_addr 41 store i8* %in, i8** %in_addr 42 %dt = fptrunc double %dt1 to float ; <float> [#uses=1] 43 store float %dt, float* %dt_addr 44 store i32 %x_size, i32* %x_size_addr 45 store i32 %y_size, i32* %y_size_addr 46 store i8* %bp, i8** %bp_addr 47 %0 = load i8*, i8** %in_addr, align 4 ; <i8*> [#uses=1] 48 store i8* %0, i8** %out, align 4 49 %1 = call i32 (...) @foo() nounwind ; <i32> [#uses=1] 50 store i32 %1, i32* %i, align 4 51 %2 = load i32, i32* %three_by_three_addr, align 4 ; <i32> [#uses=1] 52 %3 = icmp eq i32 %2, 0 ; <i1> [#uses=1] 53 br i1 %3, label %bb, label %bb2 54 55 bb: ; preds = %entry 56 %4 = load float, float* %dt_addr, align 4 ; <float> [#uses=1] 57 %5 = fpext float %4 to double ; <double> [#uses=1] 58 %6 = fmul double %5, 1.500000e+00 ; <double> [#uses=1] 59 %7 = fptosi double %6 to i32 ; <i32> [#uses=1] 60 %8 = add nsw i32 %7, 1 ; <i32> [#uses=1] 61 store i32 %8, i32* %mask_size, align 4 62 br label %bb3 63 64 bb2: ; preds = %entry 65 store i32 1, i32* %mask_size, align 4 66 br label %bb3 67 68 bb3: ; preds = %bb2, %bb 69 %9 = load i32, i32* %mask_size, align 4 ; <i32> [#uses=1] 70 %10 = mul i32 %9, 2 ; <i32> [#uses=1] 71 %11 = add nsw i32 %10, 1 ; <i32> [#uses=1] 72 store i32 %11, i32* %n_max, align 4 73 %12 = load i32, i32* %x_size_addr, align 4 ; <i32> [#uses=1] 74 %13 = load i32, i32* %n_max, align 4 ; <i32> [#uses=1] 75 %14 = sub i32 %12, %13 ; <i32> [#uses=1] 76 store i32 %14, i32* %increment, align 4 77 %15 = load i32, i32* %n_max, align 4 ; <i32> [#uses=1] 78 %16 = load i32, i32* %n_max, align 4 ; <i32> [#uses=1] 79 %17 = mul i32 %15, %16 ; <i32> [#uses=1] 80 %18 = call noalias i8* @malloc(i32 %17) nounwind ; <i8*> [#uses=1] 81 store i8* %18, i8** %dp, align 4 82 %19 = load i8*, i8** %dp, align 4 ; <i8*> [#uses=1] 83 store i8* %19, i8** %dpt, align 4 84 %20 = load float, float* %dt_addr, align 4 ; <float> [#uses=1] 85 %21 = load float, float* %dt_addr, align 4 ; <float> [#uses=1] 86 %22 = fmul float %20, %21 ; <float> [#uses=1] 87 %23 = fsub float -0.000000e+00, %22 ; <float> [#uses=1] 88 store float %23, float* %temp, align 4 89 %24 = load i32, i32* %mask_size, align 4 ; <i32> [#uses=1] 90 %25 = sub i32 0, %24 ; <i32> [#uses=1] 91 store i32 %25, i32* %j, align 4 92 br label %bb5 93 94 bb4: ; preds = %bb5 95 %26 = load i32, i32* %j, align 4 ; <i32> [#uses=1] 96 %27 = load i32, i32* %j, align 4 ; <i32> [#uses=1] 97 %28 = mul i32 %26, %27 ; <i32> [#uses=1] 98 %29 = sitofp i32 %28 to double ; <double> [#uses=1] 99 %30 = fmul double %29, 1.234000e+00 ; <double> [#uses=1] 100 %31 = fptosi double %30 to i32 ; <i32> [#uses=1] 101 store i32 %31, i32* %x, align 4 102 %32 = load i32, i32* %x, align 4 ; <i32> [#uses=1] 103 %33 = trunc i32 %32 to i8 ; <i8> [#uses=1] 104 %34 = load i8*, i8** %dpt, align 4 ; <i8*> [#uses=1] 105 store i8 %33, i8* %34, align 1 106 %35 = load i8*, i8** %dpt, align 4 ; <i8*> [#uses=1] 107 %36 = getelementptr inbounds i8, i8* %35, i64 1 ; <i8*> [#uses=1] 108 store i8* %36, i8** %dpt, align 4 109 %37 = load i32, i32* %j, align 4 ; <i32> [#uses=1] 110 %38 = add nsw i32 %37, 1 ; <i32> [#uses=1] 111 store i32 %38, i32* %j, align 4 112 br label %bb5 113 114 bb5: ; preds = %bb4, %bb3 115 %39 = load i32, i32* %j, align 4 ; <i32> [#uses=1] 116 %40 = load i32, i32* %mask_size, align 4 ; <i32> [#uses=1] 117 %41 = icmp sle i32 %39, %40 ; <i1> [#uses=1] 118 br i1 %41, label %bb4, label %bb6 119 120 bb6: ; preds = %bb5 121 br label %return 122 123 return: ; preds = %bb6 124 ret void 125 } 126 127 declare i32 @foo(...) 128 129 declare noalias i8* @malloc(i32) nounwind 130
