methods for license conversion

sig siguctua at gmail.com
Sat Jul 21 08:10:33 UTC 2007


On 21/07/07, J J <azreal1977 at hotmail.com> wrote:
>
>
>
>
>
> ________________________________
> > Date: Mon, 9 Jul 2007 02:10:24 +0300
> > From: siguctua at gmail.com
> > To: squeak-dev at lists.squeakfoundation.org
> > Subject: Re: methods for license conversion
> >
> > I implemented a scratch replacement of Canvas/Display stuff to use GL
> > via FFI for drawing morphic with squeak. Premiliary, based on rough
> > calculations it renders desktop 3 to 5 times faster than currently
> > used Balloon/BitBlt.
>
> Even using FFI, it is so much faster?  I was under the impression that FFI
> was itself slow?  What would happen if a plugin were made the GL calls?  How
> much faster do you think that would make it (if any)?
>
I examined how FFI callout implemented. There are many tests to ensure
that you calling FFI function with sane parameters, also, most of them
require a conversion (or coercion) to corresponding to C types.
I don't think that primitives will be faster than ffi - you still need
to perform same checking and type conversions for each external
function you call.
Also, someone may be tempted to implement plugin with primitives which
have higher level API, than OpenGL have, but i simply against this
approach. Yes, we can gain substantional speed boost using primitives,
but for this we need to design own architecture and support it, extend
it , and debug it e.t.c.
All this leads to bulk and poorly manageable C code. Even if someone,
having tons of free time and dedication, will implement plugin to use
GL, i'll still will not use it, because of reasons above.

> ________________________________
> PC Magazine's 2007 editors' choice for best web mail—award-winning Windows
> Live Hotmail. Check it out!
>
>
>



More information about the Squeak-dev mailing list