[EMAIL PROTECTED] wrote:
> version 1.2 has anti-aliasing implemented already via ghostscript.
> it consists of (simplifying a *lot*) rendering the page at a higher
> resolution and averaging the values for the lower resolution.
>
> regarding speed, xdvi is a completely different beast, being at the
> same time a viewer and an interpreter -- therefore it can be much much
> faster. i am working slowly on a ggi dvi viewer now, but you can try
> my 'dview' hack (on ggi ftp i guess) and see for yourself that the
> speed is much higher.
>
> gv and psv are only interfaces to ghostscript. the difference is that
> there is a driver in ghostscript that is tailor-made for gv, namely
> the 'x' driver. with this driver, gv itself needs to do little (i
> think), everything is done directly to the x window by ghostscript.
>
> with psv i have to (a) dump the page contents to ghostscript, which is
> reasonably fast, (b) wait for ghostscript to process the page, and (c)
> get the page back. step (c) kills everything - as you can see -
> because loading the big page description takes time. step (b) is also
> slower than it could be because it generates a big page description
> that would otherwise be unecessary. if there was a `ggi' driver for
> ghostscript, then things would be much faster, because step (c) would
> be eliminated, and step (b) could then be made much more efficient.
so how do you do (c) ? Are you asking gs to write to a bitmap (png, say)
and then load that (via file or pipe) into your process ?
Having never looked into gs code I don't know how difficult it would be,
but it appears that the natural thing to do would be to write a 'ggi target'
for it, i.e. to let gs render directly into a ggi visual.
With my recent experiments in embedding ggi applications into the berlin
scene graph I think a gs-on-ggi renderer would be a natural candidate for
that.
> on the other hand, if someone can patch psv and make it as fast as gv
> (or more), i would be vary happy to incorporate it to the official
> distribution. maybe some obscure driver that i missed would make it
> faster... (i actually looked into the gdevmem driver, but found little
> that could help -- not even counting the fact that debian ghostscript
> does not have this driver)
>
> what i find rather amazing is that psv is much faster under the
> console than under x. i can't quite explain it, except perhaps for x
> being slow by nature...
by the way, ghostscript has a new maintainer: Raph Levine, who is known
for a couple of very interesting projects (the advogato site with its trust
metric, the libart library, etc.)
It might be interesting to tell him about your goal, i.e. to have a ghostscript
driver targetted at ggi. Who knows, may be he'd be willing to help. It shouldn't
be difficult. After all a visual is just a specification for a particular memory
layout...
Anyway, Raph is now more than busy implementing the pdf 1.4 specs for
ghostscript...
Best regards, Stefan
_______________________________________________________
Stefan Seefeld
Departement de Physique
Universite de Montreal
email: [EMAIL PROTECTED]
_______________________________________________________
...ich hab' noch einen Koffer in Berlin...
