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
  %.anon85 = alloca i64, align 8
  %v89 = 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 = getelementptr inbounds [4 x <4 x float>], ptr %a, i64 0, i64 0
  %11 = load float, ptr %cosr, align 4
  %12 = load float, ptr %x, align 4
  %13 = load float, ptr %x, align 4
  %fmul = fmul reassoc arcp contract float %12, %13
  %14 = load float, ptr %cosr, align 4
  %fpfpext = fpext float %14 to double
  %fsub = fsub reassoc arcp contract double 1.000000e+00, %fpfpext
  %fpfptrunc = fptrunc double %fsub to float
  %15 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul, float %fpfptrunc, float %11)
  %16 = insertelement <4 x float> undef, float %15, i64 0
  %17 = load float, ptr %x, align 4
  %18 = load float, ptr %y, align 4
  %fmul1 = fmul reassoc arcp contract float %17, %18
  %19 = load float, ptr %cosr, align 4
  %fpfpext2 = fpext float %19 to double
  %fsub3 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext2
  %fpfptrunc4 = fptrunc double %fsub3 to float
  %20 = load float, ptr %z, align 4
  %21 = load float, ptr %sinr, align 4
  %fmul5 = fmul reassoc arcp contract float %20, %21
  %22 = fneg reassoc arcp contract float %fmul5
  %23 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul1, float %fpfptrunc4, float %22)
  %24 = insertelement <4 x float> %16, float %23, i64 1
  %25 = load float, ptr %x, align 4
  %26 = load float, ptr %z, align 4
  %fmul6 = fmul reassoc arcp contract float %25, %26
  %27 = load float, ptr %cosr, align 4
  %fpfpext7 = fpext float %27 to double
  %fsub8 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext7
  %fpfptrunc9 = fptrunc double %fsub8 to float
  %28 = load float, ptr %y, align 4
  %29 = load float, ptr %sinr, align 4
  %fmul10 = fmul reassoc arcp contract float %28, %29
  %30 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul6, float %fpfptrunc9, float %fmul10)
  %31 = insertelement <4 x float> %24, float %30, i64 2
  %32 = insertelement <4 x float> %31, float 0.000000e+00, i64 3
  store <4 x float> %32, ptr %10, align 16
  %33 = getelementptr inbounds [4 x <4 x float>], ptr %a, i64 0, i64 1
  %34 = load float, ptr %y, align 4
  %35 = load float, ptr %x, align 4
  %fmul11 = fmul reassoc arcp contract float %34, %35
  %36 = load float, ptr %cosr, align 4
  %fpfpext12 = fpext float %36 to double
  %fsub13 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext12
  %fpfptrunc14 = fptrunc double %fsub13 to float
  %37 = load float, ptr %z, align 4
  %38 = load float, ptr %sinr, align 4
  %fmul15 = fmul reassoc arcp contract float %37, %38
  %39 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul11, float %fpfptrunc14, float %fmul15)
  %40 = insertelement <4 x float> undef, float %39, i64 0
  %41 = load float, ptr %cosr, align 4
  %42 = load float, ptr %y, align 4
  %43 = load float, ptr %y, align 4
  %fmul16 = fmul reassoc arcp contract float %42, %43
  %44 = load float, ptr %cosr, align 4
  %fpfpext17 = fpext float %44 to double
  %fsub18 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext17
  %fpfptrunc19 = fptrunc double %fsub18 to float
  %45 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul16, float %fpfptrunc19, float %41)
  %46 = insertelement <4 x float> %40, float %45, i64 1
  %47 = load float, ptr %y, align 4
  %48 = load float, ptr %z, align 4
  %fmul20 = fmul reassoc arcp contract float %47, %48
  %49 = load float, ptr %cosr, align 4
  %fpfpext21 = fpext float %49 to double
  %fsub22 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext21
  %fpfptrunc23 = fptrunc double %fsub22 to float
  %50 = load float, ptr %x, align 4
  %51 = load float, ptr %sinr, align 4
  %fmul24 = fmul reassoc arcp contract float %50, %51
  %52 = fneg reassoc arcp contract float %fmul24
  %53 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul20, float %fpfptrunc23, float %52)
  %54 = insertelement <4 x float> %46, float %53, i64 2
  %55 = insertelement <4 x float> %54, float 0.000000e+00, i64 3
  store <4 x float> %55, ptr %33, align 16
  %56 = getelementptr inbounds [4 x <4 x float>], ptr %a, i64 0, i64 2
  %57 = load float, ptr %z, align 4
  %58 = load float, ptr %x, align 4
  %fmul25 = fmul reassoc arcp contract float %57, %58
  %59 = load float, ptr %cosr, align 4
  %fpfpext26 = fpext float %59 to double
  %fsub27 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext26
  %fpfptrunc28 = fptrunc double %fsub27 to float
  %60 = load float, ptr %y, align 4
  %61 = load float, ptr %sinr, align 4
  %fmul29 = fmul reassoc arcp contract float %60, %61
  %62 = fneg reassoc arcp contract float %fmul29
  %63 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul25, float %fpfptrunc28, float %62)
  %64 = insertelement <4 x float> undef, float %63, i64 0
  %65 = load float, ptr %z, align 4
  %66 = load float, ptr %y, align 4
  %fmul30 = fmul reassoc arcp contract float %65, %66
  %67 = load float, ptr %cosr, align 4
  %fpfpext31 = fpext float %67 to double
  %fsub32 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext31
  %fpfptrunc33 = fptrunc double %fsub32 to float
  %68 = load float, ptr %x, align 4
  %69 = load float, ptr %sinr, align 4
  %fmul34 = fmul reassoc arcp contract float %68, %69
  %70 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul30, float %fpfptrunc33, float %fmul34)
  %71 = insertelement <4 x float> %64, float %70, i64 1
  %72 = load float, ptr %cosr, align 4
  %73 = load float, ptr %z, align 4
  %74 = load float, ptr %z, align 4
  %fmul35 = fmul reassoc arcp contract float %73, %74
  %75 = load float, ptr %cosr, align 4
  %fpfpext36 = fpext float %75 to double
  %fsub37 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext36
  %fpfptrunc38 = fptrunc double %fsub37 to float
  %76 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul35, float %fpfptrunc38, float %72)
  %77 = insertelement <4 x float> %71, float %76, i64 2
  %78 = insertelement <4 x float> %77, float 0.000000e+00, i64 3
  store <4 x float> %78, ptr %56, align 16
  %79 = getelementptr inbounds [4 x <4 x float>], ptr %a, i64 0, i64 3
  store <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 1.000000e+00>, ptr %79, align 16
  %80 = getelementptr inbounds [4 x <4 x float>], ptr %b, i64 0, i64 0
  %81 = load float, ptr %cosr, align 4
  %82 = load float, ptr %x, align 4
  %83 = load float, ptr %x, align 4
  %fmul39 = fmul reassoc arcp contract float %82, %83
  %84 = load float, ptr %cosr, align 4
  %fpfpext40 = fpext float %84 to double
  %fsub41 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext40
  %fpfptrunc42 = fptrunc double %fsub41 to float
  %85 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul39, float %fpfptrunc42, float %81)
  %86 = insertelement <4 x float> undef, float %85, i64 0
  %87 = load float, ptr %x, align 4
  %88 = load float, ptr %y, align 4
  %fmul43 = fmul reassoc arcp contract float %87, %88
  %89 = load float, ptr %cosr, align 4
  %fpfpext44 = fpext float %89 to double
  %fsub45 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext44
  %fpfptrunc46 = fptrunc double %fsub45 to float
  %90 = load float, ptr %z, align 4
  %91 = load float, ptr %sinr, align 4
  %fmul47 = fmul reassoc arcp contract float %90, %91
  %92 = fneg reassoc arcp contract float %fmul47
  %93 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul43, float %fpfptrunc46, float %92)
  %94 = insertelement <4 x float> %86, float %93, i64 1
  %95 = load float, ptr %x, align 4
  %96 = load float, ptr %z, align 4
  %fmul48 = fmul reassoc arcp contract float %95, %96
  %97 = load float, ptr %cosr, align 4
  %fpfpext49 = fpext float %97 to double
  %fsub50 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext49
  %fpfptrunc51 = fptrunc double %fsub50 to float
  %98 = load float, ptr %y, align 4
  %99 = load float, ptr %sinr, align 4
  %fmul52 = fmul reassoc arcp contract float %98, %99
  %100 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul48, float %fpfptrunc51, float %fmul52)
  %101 = insertelement <4 x float> %94, float %100, i64 2
  %102 = insertelement <4 x float> %101, float 0.000000e+00, i64 3
  store <4 x float> %102, ptr %80, align 16
  %103 = getelementptr inbounds [4 x <4 x float>], ptr %b, i64 0, i64 1
  %104 = load float, ptr %y, align 4
  %105 = load float, ptr %x, align 4
  %fmul53 = fmul reassoc arcp contract float %104, %105
  %106 = load float, ptr %cosr, align 4
  %fpfpext54 = fpext float %106 to double
  %fsub55 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext54
  %fpfptrunc56 = fptrunc double %fsub55 to float
  %107 = load float, ptr %z, align 4
  %108 = load float, ptr %sinr, align 4
  %fmul57 = fmul reassoc arcp contract float %107, %108
  %109 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul53, float %fpfptrunc56, float %fmul57)
  %110 = insertelement <4 x float> undef, float %109, i64 0
  %111 = load float, ptr %cosr, align 4
  %112 = load float, ptr %y, align 4
  %113 = load float, ptr %y, align 4
  %fmul58 = fmul reassoc arcp contract float %112, %113
  %114 = load float, ptr %cosr, align 4
  %fpfpext59 = fpext float %114 to double
  %fsub60 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext59
  %fpfptrunc61 = fptrunc double %fsub60 to float
  %115 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul58, float %fpfptrunc61, float %111)
  %116 = insertelement <4 x float> %110, float %115, i64 1
  %117 = load float, ptr %y, align 4
  %118 = load float, ptr %z, align 4
  %fmul62 = fmul reassoc arcp contract float %117, %118
  %119 = load float, ptr %cosr, align 4
  %fpfpext63 = fpext float %119 to double
  %fsub64 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext63
  %fpfptrunc65 = fptrunc double %fsub64 to float
  %120 = load float, ptr %x, align 4
  %121 = load float, ptr %sinr, align 4
  %fmul66 = fmul reassoc arcp contract float %120, %121
  %122 = fneg reassoc arcp contract float %fmul66
  %123 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul62, float %fpfptrunc65, float %122)
  %124 = insertelement <4 x float> %116, float %123, i64 2
  %125 = insertelement <4 x float> %124, float 0.000000e+00, i64 3
  store <4 x float> %125, ptr %103, align 16
  %126 = getelementptr inbounds [4 x <4 x float>], ptr %b, i64 0, i64 2
  %127 = load float, ptr %z, align 4
  %128 = load float, ptr %x, align 4
  %fmul67 = fmul reassoc arcp contract float %127, %128
  %129 = load float, ptr %cosr, align 4
  %fpfpext68 = fpext float %129 to double
  %fsub69 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext68
  %fpfptrunc70 = fptrunc double %fsub69 to float
  %130 = load float, ptr %y, align 4
  %131 = load float, ptr %sinr, align 4
  %fmul71 = fmul reassoc arcp contract float %130, %131
  %132 = fneg reassoc arcp contract float %fmul71
  %133 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul67, float %fpfptrunc70, float %132)
  %134 = insertelement <4 x float> undef, float %133, i64 0
  %135 = load float, ptr %z, align 4
  %136 = load float, ptr %y, align 4
  %fmul72 = fmul reassoc arcp contract float %135, %136
  %137 = load float, ptr %cosr, align 4
  %fpfpext73 = fpext float %137 to double
  %fsub74 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext73
  %fpfptrunc75 = fptrunc double %fsub74 to float
  %138 = load float, ptr %x, align 4
  %139 = load float, ptr %sinr, align 4
  %fmul76 = fmul reassoc arcp contract float %138, %139
  %140 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul72, float %fpfptrunc75, float %fmul76)
  %141 = insertelement <4 x float> %134, float %140, i64 1
  %142 = load float, ptr %cosr, align 4
  %143 = load float, ptr %z, align 4
  %144 = load float, ptr %z, align 4
  %fmul77 = fmul reassoc arcp contract float %143, %144
  %145 = load float, ptr %cosr, align 4
  %fpfpext78 = fpext float %145 to double
  %fsub79 = fsub reassoc arcp contract double 1.000000e+00, %fpfpext78
  %fpfptrunc80 = fptrunc double %fsub79 to float
  %146 = call reassoc arcp contract float @llvm.fmuladd.f32(float %fmul77, float %fpfptrunc80, float %142)
  %147 = insertelement <4 x float> %141, float %146, i64 2
  %148 = insertelement <4 x float> %147, float 0.000000e+00, i64 3
  store <4 x float> %148, ptr %126, align 16
  %149 = getelementptr inbounds [4 x <4 x float>], ptr %b, i64 0, i64 3
  store <4 x float> <float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 1.000000e+00>, ptr %149, align 16
  store i64 0, ptr %.anon, align 8
  br label %loop.cond

loop.cond:                                        ; preds = %loop.body, %entry
  %150 = load i64, ptr %.anon, align 8
  %gt = icmp ugt i64 4, %150
  br i1 %gt, label %loop.body, label %loop.exit

loop.body:                                        ; preds = %loop.cond
  %151 = load i64, ptr %.anon, align 8
  %152 = getelementptr inbounds [4 x <4 x float>], ptr %a, i64 0, i64 %151
  %153 = load <4 x float>, ptr %152, align 16
  store <4 x float> %153, ptr %v, align 16
  %154 = load <4 x float>, ptr %v, align 16
  %155 = extractelement <4 x float> %154, i64 0
  %fpfpext81 = fpext float %155 to double
  %156 = load <4 x float>, ptr %v, align 16
  %157 = extractelement <4 x float> %156, i64 1
  %fpfpext82 = fpext float %157 to double
  %158 = load <4 x float>, ptr %v, align 16
  %159 = extractelement <4 x float> %158, i64 2
  %fpfpext83 = fpext float %159 to double
  %160 = load <4 x float>, ptr %v, align 16
  %161 = extractelement <4 x float> %160, i64 3
  %fpfpext84 = fpext float %161 to double
  %162 = call i32 (ptr, ...) @printf(ptr @.str, double %fpfpext81, double %fpfpext82, double %fpfpext83, double %fpfpext84)
  %163 = load i64, ptr %.anon, align 8
  %add = add i64 %163, 1
  store i64 %add, ptr %.anon, align 8
  br label %loop.cond

loop.exit:                                        ; preds = %loop.cond
  %164 = call i32 (ptr, ...) @printf(ptr @.str.1)
  store i64 0, ptr %.anon85, align 8
  br label %loop.cond86

loop.cond86:                                      ; preds = %loop.body88, %loop.exit
  %165 = load i64, ptr %.anon85, align 8
  %gt87 = icmp ugt i64 4, %165
  br i1 %gt87, label %loop.body88, label %loop.exit95

loop.body88:                                      ; preds = %loop.cond86
  %166 = load i64, ptr %.anon85, align 8
  %167 = getelementptr inbounds [4 x <4 x float>], ptr %b, i64 0, i64 %166
  %168 = load <4 x float>, ptr %167, align 16
  store <4 x float> %168, ptr %v89, align 16
  %169 = load <4 x float>, ptr %v89, align 16
  %170 = extractelement <4 x float> %169, i64 0
  %fpfpext90 = fpext float %170 to double
  %171 = load <4 x float>, ptr %v89, align 16
  %172 = extractelement <4 x float> %171, i64 1
  %fpfpext91 = fpext float %172 to double
  %173 = load <4 x float>, ptr %v89, align 16
  %174 = extractelement <4 x float> %173, i64 2
  %fpfpext92 = fpext float %174 to double
  %175 = load <4 x float>, ptr %v89, align 16
  %176 = extractelement <4 x float> %175, i64 3
  %fpfpext93 = fpext float %176 to double
  %177 = call i32 (ptr, ...) @printf(ptr @.str.2, double %fpfpext90, double %fpfpext91, double %fpfpext92, double %fpfpext93)
  %178 = load i64, ptr %.anon85, align 8
  %add94 = add i64 %178, 1
  store i64 %add94, ptr %.anon85, align 8
  br label %loop.cond86

loop.exit95:                                      ; preds = %loop.cond86
  ret void
}