5 Replies Latest reply on May 13, 2010 9:07 PM by Chris Cox

    Parallel layer compression

    sean.r.baker

      Could we maybe get ZIP layer compression for PSD / TIF saves parallelized to take advantage of all available cores when writing out a file?  On my home setup (Q6600 stock, 8GB, Vista64, 2TB of RAID5 over 4 disks, file sizes 500MB++), saving files pings a single core of my processor between 21 and 25 percent with bursts of HD writing of roughly the same % of its throughput capacity.  If layer data were being compressed into memory / cache in a parallel fashion (spawning new threads per layer to compress) and from there streamed to the controller, I would expect (I'm sure Chris will tell me if it's naively ) to see a 2-3x improvement in my write times.  As processors increasingly become parallel, files larger, and fast-writing drives & arrays more common, such a reworked saving routine would be beneficial to a wider and wider audience.  IMO.

       

      My current workaround is saving with everything uncompressed, but then I'm also spending a longer time for the write to occur in the first place and bloating my files a good bit (lots of channels for adjustment layers, many of which respond well to ZIP).

       

      Thanks,


      Sean