On Tue, Dec 13, 2011 at 09:27:10AM +0100, Joakim Tjernlund wrote: > "Darrick J. Wong" <djwong@xxxxxxxxxx> wrote on 2011/12/13 07:32:28: > > > > On Mon, Dec 12, 2011 at 05:10:45PM -0600, Bob Pearson wrote: > > > That choice was for Joakim who measured better performance on his 32 bit PPC > > > platform with "by 4". > > > > Ok. On my 1.33GHz PowerBook I get ~255MB/s with slice by 4 and ~270MB/s with > > slice by 8. I think it's a PPC 7447, and definitely 32-bit. In any case, it > > reports having 32K of L1D cache. > > I tested Bobs early version on my mpc8321(266MHz, embedded CPU) and it was just > half the speed compared with current crc32. I wonder, given the patch "crc32: Speed up memory table access on powerpc" would you mind retesting to see if slice by 8 still trails slice by 4 on your powerpc? I see that your mpc8321 has 16K of L1D cache and a 32-bit memory bus whereas my 7447 has a 64-bit memory bus. I wonder if memory bus size could be a defining characteristic...? I tried it out the crc32c code on a s390x today; apparently by-8 trails by-4 there too. It's unfortunately difficult to figure out the hardware details of whatever's going on underneath that VM. --D > > Jocke > > -- > To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in > the body of a message to majordomo@xxxxxxxxxxxxxxx > More majordomo info at http://vger.kernel.org/majordomo-info.html > -- To unsubscribe from this list: send the line "unsubscribe linux-crypto" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html