[forum] Some perspective from the cheap seats...

[Allen Akin]

On Wed, Mar 26, 2003 at 06:18:32AM -0800, Jim.Gettys@hp.com wrote:
| What modification?  Sounds like you don't like my FPU constraint.

In my previous message, I said:

	As far as hardware assist for 3D, we're probably looking at the
	last generation of PDAs, cell phones, and similar devices
	without it.

And we seem to agree that the host doesn't need an FPU for the device to
include hardware assist for 3D:

| > Going by what I heard at the last ARB meeting, we're likely to have 3D
| > assist before we have FPUs in the host processor.
| 
| Wouldn't surprise me in the least.  ...

So why should I accept a bet that doesn't match my original assertion?  :-)

Your points about power are well-taken, and explain why FPUs are so slow
to appear on the host.  The same isn't necessarily true of
floating-point capability in the graphics accelerator (although those
may also be fixed-point in the beginning).  You can budget additional
power for applications like gaming, since most PDAs and phones won't be
running those apps all the time.

| > In the near term
| > generating FP data is likely to be less important than rendering it.
| > Particularly for multiplayer and wireless.
| >
| Depends which market...

As far as I can tell, everything begins with games.  Once that hurdle is
cleared, other apps are worth considering.

| I dunno the current state of integer implementations in freely available
| OpenGL implementations; maybe one of the DRI folks would know and care
| to enlighten this discussion; other wise, it is just hand waving on
| the rendering side.

I don't know of an Open Source implementation (but I haven't looked for
one).  I've been told that proprietary implementations exist.  Their
relevance depends a lot on how the hardware vendors tackle the
acceleration question.


All bets aside, let me know next time you're out this way and I'll
spring for dinner.  It would be good to catch up.

Allen