Curious if this would have any impact on zram? lz4 and lzo are more recent defaults, I think, but zstd offers better compression. If this improves reads by 70%, it may make zstd the go-to choice.
From this test, lzo-rle is a bit more than twice as fast for IO swap throughput (a best case multi page, LZO-RLE: 8,248 MiB/s vs Zstd: 3,005 MiB/s). But with a 70% increase in speed, that gap may be significantly lessened, to the point where there may be a noticeable improvement due to the significantly better compression ratio (2.68:1 vs 3.96:1).
I didn’t read the code, but my impression is that it likely won’t help zRam much. The article says this was happening every time a zstd context was created, for zRam I imagine that this only happens once (per CPU core/thread?) at boot time.
