Hi All,

Request your help in understanding why the below code is producing different outputs, code 1 output: 4294966806, code 2/3 output: 50331648, and which one is the correct output.

import std.stdio;

auto floatswap(T)(T[] array) @trusted @nogc pure nothrow
{
    foreach (ref item; array)
    {
        import core.bitop : bswap;
        // Code 1 output: 4294966806
        const swapped1 = bswap(*cast(uint*)&item);
        item = *cast(const(T)*)&swapped1;
        return cast(uint) item;

        // Code 2 output: 50331648
        uint tmp = cast(uint) item;
        tmp = bswap(tmp);
        uint swapped2 = cast(float) tmp;
        return swapped2;

        // Code 3 output: 50331648
        return bswap(cast(uint) item);
     }
    return assert(false, "Unsupported T: " ~ T.stringof);
}

void main () {
    float[1] floatswapBuffer = [3.14f];
    writeln(floatswap(floatswapBuffer[]));
}

On Monday, 14 October 2024 at 15:59:07 UTC, vinoB wrote:

cast(unint) of a float does truncation. So none of your version is correct.
See https://dlang.org/spec/expression.html#cast_floating, paragraph 2.

On Monday, 14 October 2024 at 20:38:38 UTC, user1234 wrote:

Well "Code 1" does the right thing expected for the cast on the return which wastes it.