Carsten Haitzler (The Rasterman)
raster at openmoko.org
Tue Jul 8 12:24:34 CEST 2008
On Sat, 5 Jul 2008 21:22:27 +0200 Jay Vaughan <jayv at synth.net> babbled:
> > that's very sad to read! I just bought my FreeRunner just this week
> > because
> > I head on seeral talk that "all docs are free" (except GSM stack, I
> > don't care)
> > and now this ?
> yeah, get over it though, if you make too big a deal out of it, the
> machine itself will probably break in your hands.
> for sure, the /software/ aspect of openmoko is more open than the
> hardware. such is the nature of all hardware.
> > oh well, I should have read the wiki more detailed before beleaving
> > those talks...
> > or are there any plans for a "solution" for this issue ??
> on gp2x, another arm was pressed into doing defacto 3d and it turned
> out great. i think the 3d accel unit on freerunner is borked by the
> memory/interface bottleneck, and it thus becomes impractical to use.
well 3d accel isn't borked by the bus bottleneck - but it WILL have a hefty
impact for uploading textures which you will need to do. the limits are more
resolution (u'll need to stick to qvga. 511x511 is the max 3d buffer size, so
not able to do vga. sure you could try some odd resolutions between qvga and
vga and see "interesting" resampling output... but i wouldn't try - i'd just
stick to qvga.
BUT... the idea of a 2nd arm with some shared memory interface pressed into
service as a "general purpose accelerator" (could do anything from graphics to
video decode to audio playback, or anything else you can think of) would be a
great idea - but it's something to mull about for the future. in fact this was
discussed at some point - but we'd need an extra pmu (this was in the context
of 2 samsung 6400's) and so it was fairly complex. in the end - a nice idea,
but unlikely to actually happen if it's a separate SOC. if it's a multi-core
SOC... now we are talking.
Carsten Haitzler (The Rasterman) <raster at openmoko.org>
More information about the openmoko-devel