Re: Integrate ftbench into FreeType's build structure(s)

[Werner LEMBERG]

Hello Peter,


thanks for your interest in FreeType.

> I'd like to work on this project: Integrate ftbench into FreeType's
> build structure(s).  I have implemented both approaches mentioned
> therein.
> 
> I modified ftbench to save its screen output to a file, which is
> then processed by my module to produce an HTML file. The code can be
> found here: pgfjr/ftbench-html: ftbench with HTML output
> (github.com)<https://github.com/pgfjr/ftbench-html>.  Another
> version also modified ftbench to output an HTML file directly.  The
> code is here: pgfjr/ft-bench-html-v.2
> (github.com)<https://github.com/pgfjr/ft-bench-html-v.2>.
> 
> I have attached the output of the latter for your viewing. Both HTML
> outputs, of course, still need polishing, which I"m more than glad
> to do if I'm selected for this project.
> 
> I look forward to hearing from you soon.

Before doing any coding you should have contacted us.  The 'ftbench'
project was already tackled by a student last year, please look up the
many, many e-mails regarding this topic in our mailing list archive at

  https://lists.nongnu.org/archive/html/freetype-devel/

If you read the e-mails you will find out that the student (Ahmet
Goksu) has implemented almost exactly the same as you've shown to us.
In other words I'm sorry that you did a lot of work for something that
already exists.

  
https://gitlab.freedesktop.org/freetype/freetype-demos/-/tree/gsoc-2023-ahmet?ref_type=heads

The very problem, however, that couldn't be solved then, and which is
the topic of this year's 'ftbench' project, is to refine the results
so that the timing values become meaningful.  Right now, the timings
between identical runs vary so much that they are completely
meaningless, not helpful in any way.

We would be glad if you were interested in this topic.  Be warned,
however, that it is *not* trivial task.

[And yes, we should update the GSoC page with this information, sorry
for that.]


    Werner