Daniel Berlin wrote:
>On Tue, 2002-02-26 at 12:48, Branko �ibej wrote:
>
>>Warning: I haven't actually reviewed the patch (it's too big, or my
>>attention span is too short :-), but judging by the description, I think
>>it's *very* nice.
>>
>>Dan -- do you have profiling data to compare between the current
>>implementation and your patch?
>>
>
>Okay, i get the feeling people think i've replaced vdelta.
>
Not me. :-)
>I have not.
>I've simply compressed the instruction/data/address stream it outputs.
>
>vdelta is, as we all know, dictionary based.
>
>The output of it is reasonably compressable, once you seperate it out
>into the components (instruction stream/address stream/plain text).
>
>So i've compressed it, which buys us quite a bit.
>My point with the .68 vs 19 seconds is that the extra compression done
>doesn't slow us down at all.
>
I meant comparison between the current (svndiff 0) and your (svndiff 1)
performance. Obviously the underlying vdelta stays the same, and should
cancel out.
>
>If you do, I'd really like to see them.
>
>>I'd also like to compare the times for gzip vs. svndiff1, although
>>that's a bit harder to do
>>
>
>It's hard to say, since gzip is more optimized than our vdelta
>implementation.
>
Yeh, I figured.
--
Brane �ibej <brane_at_xbc.nu> http://www.xbc.nu/brane/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:37:10 2006