Does anyone know vertex processing instruction set of R300?

[Zephyr]

A presentation from nVidia "CineFX Architecture" claims: R300 only has "max instructions: 1024" and "max loops: 4", and does not have call and return instructions. I suspect it.

Someone claims that R300 has 32 temporay registers in VPE. I suspect it too.

Anyone can tell me whether i am right or not and the whole vertex processing instruction set of R300?

BTW, anyone can tell me when ATi will publish its OpenGL 1.4 ICD and specifications?

Thx

 

[OpenGL guy]

Re: Does anyone know vertex processing instruction set of R3

A presentation from nVidia "CineFX Architecture" claims: R300 only has "max instructions: 1024" and "max loops: 4", and does not have call and return instructions. I suspect it.

You suspect what? That nvidia is correct? Dave Baumann covered much of this in his Radeon 9700 review.

 

[Zephyr]

Re: Does anyone know vertex processing instruction set of R3

A presentation from nVidia "CineFX Architecture" claims: R300 only has "max instructions: 1024" and "max loops: 4", and does not have call and return instructions. I suspect it.

You suspect what? That nvidia is correct? Dave Baumann covered much of this in his Radeon 9700 review.

I know Dave's caculation, and want to be verified. Besides, I am curious about ATi's 32 temporal registers and DX9VS2.0's 1024 Instructions, because I read this in ATi's own R300 whitepaper:

Type--------name--------count
Temporary--r[n]---------12 read/write vectors

and it seems DX9VS2.0(Beta 2) does not limit instruction number, albeit it doesnot support nested loops and nested subroutines.

 

[Dave Baumann]

I know Dave's caculation, and want to be verified.

Well, that wasn't my calculation, so you can consider it to be verified.

Many of the specs ATI put in their initial documentation were DX9 limiations at that time, which is why you see a difference in what has been reported in my (and others?) reviews and what was initially showed from the ATI previews. You'll actually note that these specifications were from a document by ATI called 'DirectX9'. These were probably right as at the time of writing and there is now a difference with DX9 spec because Beta2 was not available at that period. I think I've seen a couple of time the ATI guys saying they weren't sure exactly what they will expose once DX9 actually becomes available - they may up it more to the limits of what the hardware can do.