Request/question : -T parameter

[sebbellob]

It would be nice to change the (MT) -T parameter to lowercase ,because zstdmt cant be used as external compressor with multithreading in Freearc. In my experiences Freearc dont understant uppercase characters after the -m parameter ! (after -m you can chain compression algorithms ,like this : arc a -mdelta+zstd myarc.arc mydata.dat)

Ok ,in Arc.ini it works fine ,but i need to change frequently the number of threads and it is not a good way to do that....

ZSTD/MT:
1 thread : it eats ~750 MB RAM.
2 threads : ~4 GB RAM.
The RAM consumption is non-linear ,i cant predict the required RAM. Is it normal ?
How can i calculate the required memory ?

Thanks !

[Cyan4973]

Thanks for feedback @sebbellob

The RAM consumption is non-linear ,i cant predict the required RAM. Is it normal ?

RAM consumption is expected to be linear starting from 2 threads.

From 1 thread to 2 threads though, there is a larger difference, especially for high compression mode, because buffers are larger for parallel processing. They are much larger in --ultra mode. (I suspect you're using -22 in your example, which is an extreme case).

It would be nice to change the (MT) -T parameter to lowercase ,because zstdmt cant be used as external compressor with multithreading in Freearc

Interesting.
We selected -T# because that's the way it's defined in xz (see https://linux.die.net/man/1/xz).
-t is already taken, for "test".

Can freearc support long commands ? In which case, I would rather support --threads=#, which is the long form of -T#.

[Cyan4973]

In latest dev branch update,
zstd cli receives support for long command --threads=#.
It works the same as -T#, though the command itself needs no uppercase character.
Hope it helps for your issue with freearc.

[Cyan4973]

--threads=# command will be part of soon-to-be-released zstd v1.2.0