Re: [Freetype] LoadGlyph Time

[David Turner]

Hi James,

"James H. Cloos Jr." a écrit :
> 
> >>>>> "David" == David Turner <address@hidden> writes:
> 
> David> (you basically run a bytecode interpreter where opcodes can
> David> perform complex 2d operations like dot products and vector
> David> length measurements, all in fixed-point arithmetic, so don't
> David> count on a FPU to speed everything up)
> 
> Some subset ot the SIMD instructions on modern µPs deal with integers
> rather than floats.  Would cpu-specific optimizations using such
> instructions help speed things up?
>
This wouldn't add much I believe. Since it's not possible to
automatically parallelize TrueType bytecode.
 
> How about a jit for the bytecodes?
>
You won't win much in terms of performance, and you're going to
explode the memory requirements of the engine (and the code's
complexity).
 
> Or do patent issues start popping up with such optimiaztions?
>
Nope, the patents protect the _result_ of computations, not the
way they're done. Otherwise, we would have side-stepped them for
long by now :-)

Regards,

- David Turner
- The FreeType Project  (www.freetype.org)