Re: Windows performance

[Trevor Daniels]

Martin Tarenskeen wrote Saturday, April 18, 2015 6:29 PM

On Sat, 18 Apr 2015, Paul Morris wrote:

>> >  On Apr 18, 2015, at 11:07 AM, Trevor Daniels <address@hidden> 
>> >  wrote:
>> > 
>> >  But wouldn't this show a speed-up on systems other than Windows?
>>
>>  There are similar speed-ups on macs (see earlier in this thread).  I don’t
>>  think we’ve had anyone corroborate the results on GNU/Linux yet.
>>
>>  Thanks for looking into this and tracing it to GhostScript as the likely
>>  candidate.  And thanks to Masamichi HOSODA for the upgrade!
>
> Just a wild guess: did anyone on Windows try the same speed comparison using 
> the --ps option instead of pdf output?

Just done that.  The conversion from ps to pdf takes only a couple of seconds
with either version, so it looks like GhostScript is *not* involved.  My guess 
was
wrong.

The big change seems to be during the phase "Preprocessing graphical objects".
For my test example this takes around 13 secs of the 23 under 2.19.17 and around
3 secs of the 12 under 2.19.18, so this represents pretty well the entire 
saving.  
Using the -V option shows this phase is where all the fonts are accessed.  The
ones that are slow under 2.19.17 are the Windows Century Schoolbook fonts.
So it seems Keith's guess might be right.

On the Pango website I see an entry against the changes for Pango 1.28.2
 
- Improve performance on Windows especially for non-Latin scripts

and in 1.25.4

- Improved win32 performance 

Is this the version of Pango installed in 2.19.18?

Trevor