how does DX8/9 card support T&L?

DOOM III

Newcomer
is the support hardware or software or through the simulation done by shader?
just want to make this issue clear though it doesn't have much meaning :oops:
 
That depends entirely on the chip manufacturer. There could be a hardwired T&L unit, it could be emulated by vertex shader units or done on the host CPU. All three possibilities are currently used by some chip.
 
Xmas said:
That depends entirely on the chip manufacturer. There could be a hardwired T&L unit, it could be emulated by vertex shader units or done on the host CPU. All three possibilities are currently used by some chip.

can you name some chips that applied different approaches?
 
DOOM III said:
can you name some chips that applied different approaches?

NV2x have additional hardware pice for that. Mostly it reuses vertex shaders silicon but defenitely have its own control and performs noticeable faster so dedicated part is here. For NV3X seems that this approach also used but only on area of interconection of data - short shaders benefit, but more or less complex not - TCL for 2 and more light sources almost equal to handwrited shader so i guess that here is additional hardware support for TCL but only for data setup and flow - so its minimal and helps in case of TCL parameters changing.

R2XX have some dedicated hardware but it not realy strong performer, and R3XX DEFENITELY did'nt have additional TCL hardware but emulate TCL by standart shader on drivers level. So DEVREL of ATI recomends do not use TCL and minimise switches between TCL and vertex shaders becouse its equal to switch and recompile shader that not so fast task. Changing of TCL parameters also equal to reloading constants into shader that not strong point of ATI.

see www.digit-life.com for base reviews about R200 R300 and NV300 - we always obtain numbers and investigate this case.

So, conclusion - R3XX case is outstanding from others - here you must avoid TCL usage
 
Back
Top