oyvindln

Yeah we could probably abuse loops a bit.

Seems like we're still a bit behind on 32-bit (tested on old core (1) duo laptop): ``` Running target/release/deps/bench-c6ee7e382af50fa6 running 13 tests test compress_default ... bench: 38,803,194 ns/iter (+/- 167,060)...

I don't think decompression uses much 64-bit buffers so I'm not sure why the difference is so massive there. Compression is more as expected (actually expected it to be more...

Nice work! Helped a fair bit on 32-bit machine too: ``` running 3 tests test decompress ... bench: 3,832,808 ns/iter (+/- 18,652) test decompress_mem_to_heap_miniz ... bench: 2,696,180 ns/iter (+/- 33,146)...

>I don't think decompression uses much 64-bit buffers Actually, I just realized the Action enum is 64-bit (as status and state are 32-bit values.) I don't know whether this makes...

With my last commit, the decompression benchmark is now faster than miniz on my machine! ``` running 4 tests test decompress ... bench: 740,342 ns/iter (+/- 50,835) test decompress_mem_to_heap_miniz ......

@jrmuizel I am aware of the library, it's probably worth looking into for potential ideas for improving performance further. Granted, libdeflate doesn't support streaming, so the use case is a...

Neat. I added some extra checks to avoid some overflows/out of bounds issues that showed up in fuzzing and from people using this library after zip-rs adopted it as default....

What do zlib and C miniz when given this data? People often expect corrupted/non-conforming stuff to work regardless so if it e.g decompresses in zlib we might want to let...

Okay, then we should also reject it.