Could you tell me what your test setup and corpus was?
replyFor reference, to test ARM64, I tested v0.1.0 on an M3 mac with this fork of lzbench: https://github.com/welcome-to-the-sunny-side/lzbench/tree/ad...
Here, lz4's decompression speed was far slower than misa77 and zxc. Results are here: https://github.com/welcome-to-the-sunny-side/misa77/blob/mai...
Source?
replyJust tested a few minutes ago on the ClickBench dataset. Overall quite good, but, depending on particular columns, most of the time slower - e.g., ~2.0 vs ~2.8 GB/sec on Graviton 4 machine in AWS.
replyYou probably meant MB/sec not GB.
replyNo? Decompression is rated in GB/s per core now.
reply