@MikeB

R8G8B8A8 is an integer based buffer. RGBA8 = 8 bits for Red, 8 bits for Green, 8 bits for Blue, 8 bits for alpha = R8G8B8A8. Information is stored in 32 bits.

FP16 comment is only made for storage comparison. HDR buffer is not FP. HDR FP16 = 16 bits for each component = 64 bits per pixel.

R8G8B8A8/RGBA8 is half the space of a HDR FP16 i.e. 32bits vs 64bits respectively.

The programmers are aware of RSX’s limitations and program accordingly i.e. you don’t program against known weakness.

http://realtimecollisiondetection.net/blog/?p=15
Converting RGB to LogLuv in a fragment shader.

"Running this code through NVShaderPerf gives (from memory) 5 cycles for 9 instructions. When inserted at the end of a longer shader where there is plenty of room for instruction pairing, the total overhead for the LogLuv conversion will be less than this, perhaps around 3 cycles".

It's not 1 cycle but 3 cycles are good enough. Notice PS3's optimisations are filtering to Microsoft XNA dev network.

CELL SPE's Log2 functions vs the rest.

Last edited by Hammer on 29-Mar-2009 at 03:40 AM.
Last edited by Hammer on 29-Mar-2009 at 02:52 AM.
Last edited by Hammer on 29-Mar-2009 at 02:39 AM.
Last edited by Hammer on 29-Mar-2009 at 02:37 AM.
Last edited by Hammer on 29-Mar-2009 at 02:36 AM.

_________________
Ryzen 9 7900X, DDR5-6000 64 GB RAM, GeForce RTX 4080 16 GB
Amiga 1200 (Rev 1D1, KS 3.2, PiStorm32lite/RPi 4B 4GB/Emu68)
Amiga 500 (Rev 6A ECS, KS 3.2, PiStorm/RPi 3A+/Emu68)