c3c/test/test_suite/vector/vector_init_regression.c3t

// #target: macos-x64
// #opt: --fp-math=relaxed
module test;
extern fn int printf(char* format, ...);

fn void main() {
    float radians = 3.1415 / 4;
    float[<3>] axis = {0.0, 0.0, 1.0};

    float cosr = (float) $$cos(radians);
    float sinr = (float) $$sin(radians);
    float x = axis[0];
    float y = axis[1];
    float z = axis[2];

    float[<4>][4] a = {};
    a[0] = {
        cosr + (x * x) * (float) (1.0 - cosr),
        (x * y) * (float) (1.0 - cosr) - (z * sinr),
        (x * z) * (float) (1.0 - cosr) + (y * sinr),
        0.0
    };

    a[1] = {
        (y * x) * (float) (1.0 - cosr) + (z * sinr),
        cosr + (y * y) * (float) (1.0 - cosr),
        (y * z) * (float) (1.0 - cosr) - (x * sinr),
        0.0
    };

    a[2] = {
        (z * x) * (float) (1.0 - cosr) - (y * sinr),
        (z * y) * (float) (1.0 - cosr) + (x * sinr),
        cosr + (z * z) * (float) (1.0 - cosr),
        0.0
    };

    a[3] = {
        0.0,
        0.0,
        0.0,
        1.0
    };

    float[<4>][4] b = {
        {
            cosr + (x * x) * (float) (1.0 - cosr),
            (x * y) * (float) (1.0 - cosr) - (z * sinr),
            (x * z) * (float) (1.0 - cosr) + (y * sinr),
            0.0
        },
        {
            (y * x) * (float) (1.0 - cosr) + (z * sinr),
            cosr + (y * y) * (float) (1.0 - cosr),
            (y * z) * (float) (1.0 - cosr) - (x * sinr),
            0.0
        },
        {
            (z * x) * (float) (1.0 - cosr) - (y * sinr),
            (z * y) * (float) (1.0 - cosr) + (x * sinr),
            cosr + (z * z) * (float) (1.0 - cosr),
            0.0
        },
        {
            0.0,
            0.0,
            0.0,
            1.0
        }
    };

    foreach(v : a) {
        printf("A: %f %f %f %f\n", v[0], v[1], v[2], v[3]);
    }

    printf("\n");

    foreach(v : b) {
        printf("B: %f %f %f %f\n", v[0], v[1], v[2], v[3]);
    }
}

/* #expect: test.ll

define void @test.main() #0 {
entry:
  %radians = alloca float, align 4
  %axis = alloca <3 x float>, align 16
  %cosr = alloca float, align 4
  %sinr = alloca float, align 4
  %x = alloca float, align 4
  %y = alloca float, align 4
  %z = alloca float, align 4
  %a = alloca [4 x <4 x float>], align 16
  %b = alloca [4 x <4 x float>], align 16
  %.anon = alloca i64, align 8
  %v = alloca <4 x float>, align 16
  %.anon90 = alloca i64, align 8
  %v94 = alloca <4 x float>, align 16
  store float 0x3FE921CAC0000000, ptr %radians, align 4
  store <3 x float> <float 0.000000e+00, float 0.000000e+00, float 1.000000e+00>, ptr %axis, align 16
  %0 = load float, ptr %radians, align 4
  %1 = call reassoc arcp contract float @llvm.cos.f32(float %0)
  store float %1, ptr %cosr, align 4
  %2 = load float, ptr %radians, align 4
  %3 = call reassoc arcp contract float @llvm.sin.f32(float %2)
  store float %3, ptr %sinr, align 4
  %4 = load <3 x float>, ptr %axis, align 16
  %5 = extractelement <3 x float> %4, i64 0
  store float %5, ptr %x, align 4
  %6 = load <3 x float>, ptr %axis, align 16
  %7 = extractelement <3 x float> %6, i64 1
  store float %7, ptr %y, align 4
  %8 = load <3 x float>, ptr %axis, align 16
  %9 = extractelement <3 x float> %8, i64 2
  store float %9, ptr %z, align 4
  call void @llvm.memset.p0.i64(ptr align 16 %a, i8 0, i64 64, i1 false)
  %10 = load float, ptr %cosr, align 4
  %11 = load float, ptr %x, align 4
  %12 = load float, ptr %x, align 4
  %fmul = fmul reassoc arcp contract float %11, %12
  %13 = load float, ptr %cosr, align 4
  %fpfpext = fpext float %13 to double
  %fsub = fsub reassoc arcp contract double 1.000000e+00, %fpfpext
  %fpfptrunc = fptrunc double %fsub to float
  %14 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul, float %fpfptrunc, float %10)
  %15 = insertelement <4 x float> undef, float %14, i64 0
  %16 = load float, ptr %x, align 4
  %17 = load float, ptr %y, align 4
  %fmul1 = fmul reassoc arcp contract float %16, %17
  %18 = load float, ptr %cosr, align 4
  %fpfpext2 = fpext float %18 to double
  %fsub3 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext2
  %fpfptrunc4 = fptrunc double %fsub3 to float
  %19 = load float, ptr %z, align 4
  %20 = load float, ptr %sinr, align 4
  %fmul5 = fmul reassoc arcp contract float %19, %20
  %21 = fneg reassoc arcp contract float %fmul5
  %22 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul1, float %fpfptrunc4, float %21)
  %23 = insertelement <4 x float> %15, float %22, i64 1
  %24 = load float, ptr %x, align 4
  %25 = load float, ptr %z, align 4
  %fmul6 = fmul reassoc arcp contract float %24, %25
  %26 = load float, ptr %cosr, align 4
  %fpfpext7 = fpext float %26 to double
  %fsub8 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext7
  %fpfptrunc9 = fptrunc double %fsub8 to float
  %27 = load float, ptr %y, align 4
  %28 = load float, ptr %sinr, align 4
  %fmul10 = fmul reassoc arcp contract float %27, %28
  %29 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul6, float %fpfptrunc9, float %fmul10)
  %30 = insertelement <4 x float> %23, float %29, i64 2
  %31 = insertelement <4 x float> %30, float 0.000000e+00, i64 3
  store <4 x float> %31, ptr %a, align 16
  %ptradd = getelementptr inbounds i8, ptr %a, i64 16
  %32 = load float, ptr %y, align 4
  %33 = load float, ptr %x, align 4
  %fmul11 = fmul reassoc arcp contract float %32, %33
  %34 = load float, ptr %cosr, align 4
  %fpfpext12 = fpext float %34 to double
  %fsub13 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext12
  %fpfptrunc14 = fptrunc double %fsub13 to float
  %35 = load float, ptr %z, align 4
  %36 = load float, ptr %sinr, align 4
  %fmul15 = fmul reassoc arcp contract float %35, %36
  %37 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul11, float %fpfptrunc14, float %fmul15)
  %38 = insertelement <4 x float> undef, float %37, i64 0
  %39 = load float, ptr %cosr, align 4
  %40 = load float, ptr %y, align 4
  %41 = load float, ptr %y, align 4
  %fmul16 = fmul reassoc arcp contract float %40, %41
  %42 = load float, ptr %cosr, align 4
  %fpfpext17 = fpext float %42 to double
  %fsub18 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext17
  %fpfptrunc19 = fptrunc double %fsub18 to float
  %43 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul16, float %fpfptrunc19, float %39)
  %44 = insertelement <4 x float> %38, float %43, i64 1
  %45 = load float, ptr %y, align 4
  %46 = load float, ptr %z, align 4
  %fmul20 = fmul reassoc arcp contract float %45, %46
  %47 = load float, ptr %cosr, align 4
  %fpfpext21 = fpext float %47 to double
  %fsub22 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext21
  %fpfptrunc23 = fptrunc double %fsub22 to float
  %48 = load float, ptr %x, align 4
  %49 = load float, ptr %sinr, align 4
  %fmul24 = fmul reassoc arcp contract float %48, %49
  %50 = fneg reassoc arcp contract float %fmul24
  %51 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul20, float %fpfptrunc23, float %50)
  %52 = insertelement <4 x float> %44, float %51, i64 2
  %53 = insertelement <4 x float> %52, float 0.000000e+00, i64 3
  store <4 x float> %53, ptr %ptradd, align 16
  %ptradd25 = getelementptr inbounds i8, ptr %a, i64 32
  %54 = load float, ptr %z, align 4
  %55 = load float, ptr %x, align 4
  %fmul26 = fmul reassoc arcp contract float %54, %55
  %56 = load float, ptr %cosr, align 4
  %fpfpext27 = fpext float %56 to double
  %fsub28 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext27
  %fpfptrunc29 = fptrunc double %fsub28 to float
  %57 = load float, ptr %y, align 4
  %58 = load float, ptr %sinr, align 4
  %fmul30 = fmul reassoc arcp contract float %57, %58
  %59 = fneg reassoc arcp contract float %fmul30
  %60 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul26, float %fpfptrunc29, float %59)
  %61 = insertelement <4 x float> undef, float %60, i64 0
  %62 = load float, ptr %z, align 4
  %63 = load float, ptr %y, align 4
  %fmul31 = fmul reassoc arcp contract float %62, %63
  %64 = load float, ptr %cosr, align 4
  %fpfpext32 = fpext float %64 to double
  %fsub33 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext32
  %fpfptrunc34 = fptrunc double %fsub33 to float
  %65 = load float, ptr %x, align 4
  %66 = load float, ptr %sinr, align 4
  %fmul35 = fmul reassoc arcp contract float %65, %66
  %67 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul31, float %fpfptrunc34, float %fmul35)
  %68 = insertelement <4 x float> %61, float %67, i64 1
  %69 = load float, ptr %cosr, align 4
  %70 = load float, ptr %z, align 4
  %71 = load float, ptr %z, align 4
  %fmul36 = fmul reassoc arcp contract float %70, %71
  %72 = load float, ptr %cosr, align 4
  %fpfpext37 = fpext float %72 to double
  %fsub38 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext37
  %fpfptrunc39 = fptrunc double %fsub38 to float
  %73 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul36, float %fpfptrunc39, float %69)
  %74 = insertelement <4 x float> %68, float %73, i64 2
  %75 = insertelement <4 x float> %74, float 0.000000e+00, i64 3
  store <4 x float> %75, ptr %ptradd25, align 16
  %ptradd40 = getelementptr inbounds i8, ptr %a, i64 48
  store <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 1.000000e+00>, ptr %ptradd40, align 16
  %76 = load float, ptr %cosr, align 4
  %77 = load float, ptr %x, align 4
  %78 = load float, ptr %x, align 4
  %fmul41 = fmul reassoc arcp contract float %77, %78
  %79 = load float, ptr %cosr, align 4
  %fpfpext42 = fpext float %79 to double
  %fsub43 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext42
  %fpfptrunc44 = fptrunc double %fsub43 to float
  %80 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul41, float %fpfptrunc44, float %76)
  %81 = insertelement <4 x float> undef, float %80, i64 0
  %82 = load float, ptr %x, align 4
  %83 = load float, ptr %y, align 4
  %fmul45 = fmul reassoc arcp contract float %82, %83
  %84 = load float, ptr %cosr, align 4
  %fpfpext46 = fpext float %84 to double
  %fsub47 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext46
  %fpfptrunc48 = fptrunc double %fsub47 to float
  %85 = load float, ptr %z, align 4
  %86 = load float, ptr %sinr, align 4
  %fmul49 = fmul reassoc arcp contract float %85, %86
  %87 = fneg reassoc arcp contract float %fmul49
  %88 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul45, float %fpfptrunc48, float %87)
  %89 = insertelement <4 x float> %81, float %88, i64 1
  %90 = load float, ptr %x, align 4
  %91 = load float, ptr %z, align 4
  %fmul50 = fmul reassoc arcp contract float %90, %91
  %92 = load float, ptr %cosr, align 4
  %fpfpext51 = fpext float %92 to double
  %fsub52 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext51
  %fpfptrunc53 = fptrunc double %fsub52 to float
  %93 = load float, ptr %y, align 4
  %94 = load float, ptr %sinr, align 4
  %fmul54 = fmul reassoc arcp contract float %93, %94
  %95 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul50, float %fpfptrunc53, float %fmul54)
  %96 = insertelement <4 x float> %89, float %95, i64 2
  %97 = insertelement <4 x float> %96, float 0.000000e+00, i64 3
  store <4 x float> %97, ptr %b, align 16
  %ptradd55 = getelementptr inbounds i8, ptr %b, i64 16
  %98 = load float, ptr %y, align 4
  %99 = load float, ptr %x, align 4
  %fmul56 = fmul reassoc arcp contract float %98, %99
  %100 = load float, ptr %cosr, align 4
  %fpfpext57 = fpext float %100 to double
  %fsub58 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext57
  %fpfptrunc59 = fptrunc double %fsub58 to float
  %101 = load float, ptr %z, align 4
  %102 = load float, ptr %sinr, align 4
  %fmul60 = fmul reassoc arcp contract float %101, %102
  %103 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul56, float %fpfptrunc59, float %fmul60)
  %104 = insertelement <4 x float> undef, float %103, i64 0
  %105 = load float, ptr %cosr, align 4
  %106 = load float, ptr %y, align 4
  %107 = load float, ptr %y, align 4
  %fmul61 = fmul reassoc arcp contract float %106, %107
  %108 = load float, ptr %cosr, align 4
  %fpfpext62 = fpext float %108 to double
  %fsub63 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext62
  %fpfptrunc64 = fptrunc double %fsub63 to float
  %109 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul61, float %fpfptrunc64, float %105)
  %110 = insertelement <4 x float> %104, float %109, i64 1
  %111 = load float, ptr %y, align 4
  %112 = load float, ptr %z, align 4
  %fmul65 = fmul reassoc arcp contract float %111, %112
  %113 = load float, ptr %cosr, align 4
  %fpfpext66 = fpext float %113 to double
  %fsub67 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext66
  %fpfptrunc68 = fptrunc double %fsub67 to float
  %114 = load float, ptr %x, align 4
  %115 = load float, ptr %sinr, align 4
  %fmul69 = fmul reassoc arcp contract float %114, %115
  %116 = fneg reassoc arcp contract float %fmul69
  %117 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul65, float %fpfptrunc68, float %116)
  %118 = insertelement <4 x float> %110, float %117, i64 2
  %119 = insertelement <4 x float> %118, float 0.000000e+00, i64 3
  store <4 x float> %119, ptr %ptradd55, align 16
  %ptradd70 = getelementptr inbounds i8, ptr %b, i64 32
  %120 = load float, ptr %z, align 4
  %121 = load float, ptr %x, align 4
  %fmul71 = fmul reassoc arcp contract float %120, %121
  %122 = load float, ptr %cosr, align 4
  %fpfpext72 = fpext float %122 to double
  %fsub73 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext72
  %fpfptrunc74 = fptrunc double %fsub73 to float
  %123 = load float, ptr %y, align 4
  %124 = load float, ptr %sinr, align 4
  %fmul75 = fmul reassoc arcp contract float %123, %124
  %125 = fneg reassoc arcp contract float %fmul75
  %126 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul71, float %fpfptrunc74, float %125)
  %127 = insertelement <4 x float> undef, float %126, i64 0
  %128 = load float, ptr %z, align 4
  %129 = load float, ptr %y, align 4
  %fmul76 = fmul reassoc arcp contract float %128, %129
  %130 = load float, ptr %cosr, align 4
  %fpfpext77 = fpext float %130 to double
  %fsub78 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext77
  %fpfptrunc79 = fptrunc double %fsub78 to float
  %131 = load float, ptr %x, align 4
  %132 = load float, ptr %sinr, align 4
  %fmul80 = fmul reassoc arcp contract float %131, %132
  %133 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul76, float %fpfptrunc79, float %fmul80)
  %134 = insertelement <4 x float> %127, float %133, i64 1
  %135 = load float, ptr %cosr, align 4
  %136 = load float, ptr %z, align 4
  %137 = load float, ptr %z, align 4
  %fmul81 = fmul reassoc arcp contract float %136, %137
  %138 = load float, ptr %cosr, align 4
  %fpfpext82 = fpext float %138 to double
  %fsub83 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext82
  %fpfptrunc84 = fptrunc double %fsub83 to float
  %139 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul81, float %fpfptrunc84, float %135)
  %140 = insertelement <4 x float> %134, float %139, i64 2
  %141 = insertelement <4 x float> %140, float 0.000000e+00, i64 3
  store <4 x float> %141, ptr %ptradd70, align 16
  %ptradd85 = getelementptr inbounds i8, ptr %b, i64 48
  store <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 1.000000e+00>, ptr %ptradd85, align 16
  store i64 0, ptr %.anon, align 8
  br label %loop.cond

loop.cond:                                        ; preds = %loop.body, %entry
  %142 = load i64, ptr %.anon, align 8
  %gt = icmp ugt i64 4, %142
  br i1 %gt, label %loop.body, label %loop.exit

loop.body:                                        ; preds = %loop.cond
  %143 = load i64, ptr %.anon, align 8
  %ptroffset = getelementptr inbounds [16 x i8], ptr %a, i64 %143
  %144 = load <4 x float>, ptr %ptroffset, align 16
  store <4 x float> %144, ptr %v, align 16
  %145 = load <4 x float>, ptr %v, align 16
  %146 = extractelement <4 x float> %145, i64 0
  %fpfpext86 = fpext float %146 to double
  %147 = load <4 x float>, ptr %v, align 16
  %148 = extractelement <4 x float> %147, i64 1
  %fpfpext87 = fpext float %148 to double
  %149 = load <4 x float>, ptr %v, align 16
  %150 = extractelement <4 x float> %149, i64 2
  %fpfpext88 = fpext float %150 to double
  %151 = load <4 x float>, ptr %v, align 16
  %152 = extractelement <4 x float> %151, i64 3
  %fpfpext89 = fpext float %152 to double
  %153 = call i32 (ptr, ...) @printf(ptr @.str, double %fpfpext86, double %fpfpext87, double %fpfpext88, double %fpfpext89)
  %154 = load i64, ptr %.anon, align 8
  %addnuw = add nuw i64 %154, 1
  store i64 %addnuw, ptr %.anon, align 8
  br label %loop.cond

loop.exit:                                        ; preds = %loop.cond
  %155 = call i32 (ptr, ...) @printf(ptr @.str.1)
  store i64 0, ptr %.anon90, align 8
  br label %loop.cond91

loop.cond91:                                      ; preds = %loop.body93, %loop.exit
  %156 = load i64, ptr %.anon90, align 8
  %gt92 = icmp ugt i64 4, %156
  br i1 %gt92, label %loop.body93, label %loop.exit101

loop.body93:                                      ; preds = %loop.cond91
  %157 = load i64, ptr %.anon90, align 8
  %ptroffset95 = getelementptr inbounds [16 x i8], ptr %b, i64 %157
  %158 = load <4 x float>, ptr %ptroffset95, align 16
  store <4 x float> %158, ptr %v94, align 16
  %159 = load <4 x float>, ptr %v94, align 16
  %160 = extractelement <4 x float> %159, i64 0
  %fpfpext96 = fpext float %160 to double
  %161 = load <4 x float>, ptr %v94, align 16
  %162 = extractelement <4 x float> %161, i64 1
  %fpfpext97 = fpext float %162 to double
  %163 = load <4 x float>, ptr %v94, align 16
  %164 = extractelement <4 x float> %163, i64 2
  %fpfpext98 = fpext float %164 to double
  %165 = load <4 x float>, ptr %v94, align 16
  %166 = extractelement <4 x float> %165, i64 3
  %fpfpext99 = fpext float %166 to double
  %167 = call i32 (ptr, ...) @printf(ptr @.str.2, double %fpfpext96, double %fpfpext97, double %fpfpext98, double %fpfpext99)
  %168 = load i64, ptr %.anon90, align 8
  %addnuw100 = add nuw i64 %168, 1
  store i64 %addnuw100, ptr %.anon90, align 8
  br label %loop.cond91

loop.exit101:                                     ; preds = %loop.cond91
  ret void
}