Re: Issue 1211 in lilypond: [PATCH]Optimizations for pure-height approximations. (issue1817045)

[Graham Percival]

On Sat, Nov 20, 2010 at 12:17:55AM +0100, Valentin Villenave wrote:
> On Fri, Nov 19, 2010 at 11:25 PM,  <address@hidden> wrote:
> > Unfortunately there's been a drop in performance since 2.13.36; I've
> > narrowed it down to Joe's fix for issue 1240 (5266aa).
> 
> That's annoying.

...

> Should I reopen #1240?

No.

> (Or possibly open a new issue about it; we should also consider
> making it a High-prio, since I doubt we want to release a new
> stable version with that much of a performance drop.)

Correctness is more important than speed.


If you want to improve our speed, then look into the "time" fields
in the regtest comparison.  It appears to be currently broken, but
fixing this will likely be a 10-line patch.  Granted, a 10-line
patch which may require up to 10 hours to write for anybody
unfamiliar with the system, but since we have nobody familiar with
that stuff, nobody else is likely to tackle it.

If you want to improve our memory handling, then look into the
"cells:" fields in the regtest comparison.  As far as I know
they're working, but we have no documentation about what they
mean, and the bug squad certainly isn't checking them.

The best time to notice a drop in speed or memory importance is
before a patch is pushed.  The second-best time is after the devel
release immediately following the commit.  Improving the regtest
comparison code and docs will do more in the long term than any
amount of complaining about this specific commit.

- Graham