You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
zstd::stream::copy_decode is 10-50x slower then zstd_safe::decompress on my test input. And 2 zstd::stream::copy_decode
40% slower then zstd_safe::decompress.
zstd::stream::copy_decode currently uses the streaming API, which has indeed some overhead compared to zstd_safe::decompress, which takes the entire input data in memory at once. Especially for small input data, I'd expect the overhead to be significant.
It would be nice to expose zstd_safe::decompress in the top-level zstd API to allow avoiding this overhead for small input.
zstd::stream::copy_decode
is 10-50x slower thenzstd_safe::decompress
on my test input. And 2zstd::stream::copy_decode
40% slower then
zstd_safe::decompress
.benchmark:
results:
The text was updated successfully, but these errors were encountered: