• xthexder@l.sw0.com
    link
    fedilink
    arrow-up
    12
    ·
    edit-2
    4 days ago

    I came up with a kind of clever data type for storing short strings in a fixed size struct so they can be stored on the stack or inline without any allocations.
    It’s always null-terminated so it can be passed directly as a C-style string, but it also stores the string length without using any additional data (Getting the length would normally have to iterate to find the end).
    The trick is to store the number of unused bytes in the last character of the buffer. When the string is full, there are 0 unused bytes and the size byte overlaps the null terminator.
    (Only works for strings < 256 chars excluding null byte)

    Implementation in C++ here: https://github.com/frustra/strayphotons/blob/master/src/common/common/InlineString.hh

    Edit: Since a couple people don’t seem to understand the performance impact of this vs regular std::string, here’s a demo: https://godbolt.org/z/34j7obnbs This generates 10000 strings like “Hello, World! 00001” via concatenation. The effect is huge in debug mode, but still has performance benefits with optimizations turned on:

    With -O3 optimization
    std::string: 0.949216ms
    char[256] with strlen: 0.88104ms
    char[256] without strlen: 0.684734ms
    
    With no optimization:
    std::string: 3.5501ms
    char[256] with strlen: 0.885888ms
    char[256] without strlen: 0.687733ms
    
    (You may need to run it a few times to get sample numbers due to random server load on godbolt)
    Changing the buffer size to 32 bytes has a negligible performance improvement over 256 bytes in this case, but might be slightly faster due to the whole string fitting in a cache line.
    
    • anton@lemmy.blahaj.zone
      link
      fedilink
      arrow-up
      1
      ·
      4 days ago

      I came up with a kind of clever data type for storing short strings in a fixed size struct so they can be stored on the stack or inline without any allocations.

      C++ already does that for short strings while seamlessly switching to allocation for long strings.

      It’s always null-terminated so it can be passed directly as a C-style string, but it also stores the string length without using any additional data (Getting the length would normally have to iterate to find the end).

      Also the case in the standard library

      The trick is to store the number of unused bytes in the last character of the buffer. When the string is full, there are 0 unused bytes and the size byte overlaps the null terminator.

      Iirc, that trick was used in one implementation but discontinued because it was against the standard.

      (Only works for strings < 256 chars excluding null byte)

      If you need a niche for allocated string you can get to 254 but the typical choice seem to be around 16.

      • xthexder@l.sw0.com
        link
        fedilink
        arrow-up
        1
        ·
        edit-2
        4 days ago

        C++ already does that for short strings

        I’ve already been discussing this. Maybe read the rest of the thread.

        Also the case in the standard library

        I think you’re missing the point of why. I built this to be a nearly drop in replacement for the standard string. If this wasn’t the case it would need to do even more processing and work to pass the strings to anything.

        discontinued because it was against the standard.

        Standards don’t matter for an internal type that’s not exposed to public APIs. I’m not trying to be exactly compatible with everything under the sun. There’s no undefined behavior here so it’s fine

      • xthexder@l.sw0.com
        link
        fedilink
        arrow-up
        5
        ·
        6 days ago

        22 characters is significantly less useful than 255 characters. I use this for resource name keys, asset file paths, and a few other scenarios. The max size is configurable, so I know that nothing I am going to store is ever going to require heap allocations (really bad to be doing every frame in a game engine).

        I developed this specifically after benchmarking a simpler version and noticed a significant amount of time being spent in strlen(), and it had real benefits in my case.
        Admittedly just storing a struct with a static buffer and separate size would have worked pretty much the same and eliminated the 255 char limitation, but it was fun to build.

        • FizzyOrange@programming.dev
          link
          fedilink
          arrow-up
          1
          ·
          5 days ago

          22 characters is significantly less useful than 255 characters.

          You can still use more than 22 characters; it just switches to the heap.

          nothing I am going to store is ever going to require heap allocations

          I would put good money that using 256 bytes everywhere is going to be slower overall than just using the heap when you need more than 22 characters. 22 is quite a lot, especially for keys. ThisReallyLongKey is still only 17.

          • xthexder@l.sw0.com
            link
            fedilink
            arrow-up
            1
            ·
            4 days ago

            I don’t use 256 bytes everywhere. I use a mix of 64, 128, and 256 byte strings depending on the specific use case.
            In a hot path, having the data inline is much more important than saving a few hundred bytes. Cache efficiency plus eliminating heap allocations has huge performance benefits in a game engine that’s running frames as fast as possible.

            • FizzyOrange@programming.dev
              link
              fedilink
              arrow-up
              1
              ·
              4 days ago

              having the data inline

              It’s not as simple as that, depending on the architecture. Typically they would fetch 64-byte cache lines so your 128 bytes aren’t going to be magically more cached than 128 bytes on the heap.

              Avoiding allocations may help but also maybe not. This is definitely in “I don’t believe it until I see benchmarks” realm. I would be really really surprised if the allocation cost was remotely bad enough to justify the “sorry your file is too long” errors.

              • xthexder@l.sw0.com
                link
                fedilink
                arrow-up
                1
                ·
                4 days ago

                Check out the benchmark I edited in to my original post. These are not user-provided strings in my case.