Re: cleaner/better zlib sources?

From: Matt Mackall <mpm@selenic.com>
Date: Mar 16, 2007, 01:14 UTC
Message-ID: <20070316011457.GA4892@waste.org>
In-Reply-To: <45F9EED5.3070706@garzik.org>

On Thu, Mar 15, 2007 at 09:11:49PM -0400, Jeff Garzik wrote:

Show 7 quoted lines

> Linus Torvalds wrote:
> >Now, it's possible that I'm just wrong, but the instruction-level profile 
> >really did pinpoint the "look up state branch pointer and jump to it" as 
> >some of the hottest part of that function. Which is just *evil*. You can 
> 
> ISTR there are a bunch of state transitions per byte, which would make 
> sense that it shows up on profiles.

Yep, not surprising.

Show 10 quoted lines

> >Now, I'm just wondering if anybody knows if there are better zlib 
> >implementations out there? This really looks like it could be a noticeable 
> >performance issue, but I'm lazy and would be much happier to hear that 
> >somebody has already played with optimizing zlib. Especially since I'm not 
> >100% sure it's really going to be noticeable..
> 
> I could have sworn that either Matt Mackall or Ben LaHaise had cleaned 
> up the existing zlib so much that it was practically a new 
> implementation.  I'm not aware of any open source implementations 
> independent of zlib (except maybe that C++ behemoth, 7zip).

I cleaned up the version in lib/ that's used for boot on most systems. It's quite a bit simpler and cleaner than the code lib/zlib (and elsewhere!). But making it faster is another matter entirely - I don't know off-hand how the two compare.

-- 
Mathematics is the supreme nostalgia of our time.