Cache all pure minimum translations

Cache all pure minimum translations

I'm getting mixed results for this - I can't tell if it's leading to a
uniform speed-up on all scores or not.  What I need most from people are
runs on every sort of score possible.  I'm averaging about a 7% speed up
but it's not statistically significant.

[Keith, 2012-08-29 21:25:18]

I tried in on a score where should have its largest effect, a 41-page score with
15 staves per system, and saw 8%.
 user	2m26.191s   ->	2m14.115s
 sys	0m1.097s    ->	0m1.049s

[Keith, 2012-08-31 06:35:26]

>  I'm averaging about a 7% speed up
>  but it's not statistically significant.

What types of scores get slower ?

[mike7, 2012-08-31 07:20:43]

On 31 août 2012, at 08:35, k-ohara5a5a@oco.net wrote:

>>  I'm averaging about a 7% speed up
>>  but it's not statistically significant.
> 
> What types of scores get slower ?
> 
> http://codereview.appspot.com/6498053/
> 
> _______________________________________________
> lilypond-devel mailing list
> lilypond-devel@gnu.org
> https://lists.gnu.org/mailman/listinfo/lilypond-devel

It's just erratic - sometimes I'll run it and it'll be the same speed, sometimes
I see a speed up.  My computer has all sorts of cruft hogging memory, so it's
tough for me to accurately evaluate if there is a speed up.  There's only been a
slow-down once, though, and it was on a score that almost always showed
speed-ups (a Mozart horn concerto I use regularly for tests).

So if you're finding that the patch categorically speeds things up, all is
well.

Cheers,
MS