[Help-tar] Extraction performance problem

[Mark Lehrer]

Hello, I am running into an extraction speed limitation of around 250MB/sec, 
and I would like to find a way to improve this.  On this test machine it is 
possible to write at > 1GB/sec.

I haven't yet done a full blktrace analysis yet, but the 250MB/sec speed 
seems to match other tests I have done with a 512 byte block size.  Is it 
possible to do the equivalent of dd's "obs" option, to increase the output 
block size to 4k, 64k, or even 1M?

Also, has anyone tried to make a multi-threaded version?  I can imagine that 
the code would be a little bit complicated, but with a huge memory buffer 
and 8 threads writing it would help quite a bit even with a smaller block 
size.

Of course, if there are already ways to help or better ideas on how to 
improve things that would be great.

Thanks,
Mark