lroathe
15th January 2006 19:11 UTC

Smp
Hi,

I saw the post from 1/04 on SMP and it not being a priority, but I thought I'd bring it up again now that a couple years have passed :)

The only need for SMP (imho) is during compression, and it seems that here it should be the simplest (haha); spawn a thread to compress each file, one per processor.

Could be that directory ordering is an issue (say a large file in dir A takes longer than all the rest and we want to move to dir B). Waiting for all files in dir A to complete before moving to dir B would be fine (could be some time w/only one processor working).

If you have a couple hundered MB of data this is just a nicety... but if you have a few GB of data this would be extremely helpful! We're well over the 2GB limit in raw data, but with lzma and non-solid compression we get to ~1.6GB on the install size. On a 3Ghz SMP machine my project takes ~6 hours to compress it's data. Not great when you are trying for quick turnaround on uploading your next beta submission :(

kichik
17th January 2006 14:15 UTC

It's not currently possible to create a new thread with the current code. The script is parsed sequentially and files are compressed as they come. Also, with solid compression you can't compress one file at a time, you must compress the entire block.

It's not that this is not a desired feature, it's just that it's not that simple.