void siren
I have ported a decompression algorithm to Zig, and I want to maximise performance. It reads directly from an in-memory buffer and writes directly to another in-memory buffer, which can obviously potentially take a lot of memory and sucks compared to dealing with chunks at a time that can be stored into/read from files, but I also want the option to not have chunked reads at all. I also don't want to have a bunch of duplicated code with slight changes.
This sounds like a job for some reader/write interface with different possible underlying implementations - like std.io.GenericReader. Suppose I make my algorithm more generic so that it can take such a reader, and I make a reader implementation that reads from the in-memory buffer, wouldn't that run like ass due to added overhead? I suspect the compiler wouldn't be smart enough to reduce that down to just buffer reads like I have now. It kind of sounds like I would have to fight the compiler quite a bit to make this dream a reality. I'm very interested in best practices and I know none of them, so if you can point towards libraries you think have solved this problem (or similar), show me!
untold pawn
queen birch