Re: Windows performance

[Trevor Daniels]

Phil Holmes wrote Thursday, April 16, 2015 3:43 PM
>
> From: "Trevor Daniels" <address@hidden>
> Sent: Thursday, April 16, 2015 3:28 PM
>>
>> Phil Holmes wrote Thursday, April 16, 2015 2:00 PM
>>
>>> The performance of LilyPond 2.19.18 on Windows is _much_ better than
>>> previous versions.  Some examples:
>>>
>>> A 26 page multi-score piece I've been working on:
>>> 2.19.16: 114s to compile
>>> 2.19.18: 52s
>> ... [etc]
>>
>> That's remarkable.  I can see no change between 2.19.16
>> and 2.19.18 that might account for this enormous change.
>> The only effect of this magnitude which I've seen in the
>> past is to do with setting up the font library when LP
>> is run for the first time.  But that would be an increase.
>>
>> Any chance your 2.19.18 is using a different hard disk,
>> an SSD maybe, which the others weren't?
> 
> No.  They're all installed on an SSD.  It's CPU limited anyway.
> 
> I thought 2.19.16 might be quicker because of the change to the compiler, 
> but, as you say, have no idea why .18 is so much quicker than .16.  I'm 
> pretty certain it's a genuine difference: I only noticed because the score I 
> was working on suddenly appeared more quickly!

Well, I remain mystified, but I can confirm the speedup on my Windows Vista 
laptop with a 4-page score:

With 2.19.16 this took 25.2  22.6  22.9 secs
With 2.19.18 this took 12.6  12.5  11.8 secs

It is amazing, but very welcome!  In my career as a systems programmer, in the 
days when performance was critical, I never saw an improvement of this 
magnitude resulting from (presumably) a single change.  Usually we were 
struggling to gain a percent or two.

Were there any changes to GUB between .16 and .18?

Trevor