On Mon, Sep 10, 2007 at 03:04:27AM -0400, Jeff King wrote: > On Mon, Sep 10, 2007 at 08:50:19AM +0200, Martin Koegler wrote: > > > You could try to raise the config variable pack.deltaCacheLimit, as it > > avoids to do the delta again for all objects smaller then > > pack.deltaCacheLimit in the writing phase. > > Hmm. I bumped this to 100,000 and it seemed to make zero difference. I > was able to see some improvement by turning the compression level to > '0'. It should lower you CPU time. To verify it, I would check, that delta_cacheable returns 1 for most objects. If most deltas are cacheable, maybe the CPU time is used by compressing the deltas at write time. I noticed some other (potential) problem: delta_cache_size is not protected by locks, although it is modified by concurrent threads, eg: delta_cache_size -= trg_entry->delta_size mfg Martin Kögler - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
