DMAEN vs normal GPU priority

[Cyprian_K]

is there any difference in writing/reading (to/from main) performance?

And is it worth to use STOREP / LOADP vs STORE / LOAD?

thanks

 

 

[Zerosquare]

3 hours ago, Cyprian_K said:

And is it worth to use STOREP / LOADP vs STORE / LOAD?

The main bus is 64-bit wide and so is the RAM, so using STORE / LOAD when you could be using STOREP / LOADP reduces performance. Of course, it only matters if you need to transfer more than 32 bits of data.

[Cyprian_K]

ok, and what about DMAEN? does it boost read/write performance?

[Zerosquare]

Never used it, but according to the documentation, it doesn't. It does raises the GPU priority on the bus, but if there's no bus contention, it won't make any difference. And if there is bus contention, it causes the GPU to have a higher priority than the object processor, which is something you don't want as it will cause display glitches (as documented in the errata, which tells you not to use DMAEN).

[Cyprian_K]

ok,

can it be considered as a lock the bus for the other devices?

It could allows to avoid additional waitstates when other devices touch a different memory page.

 

I guess glitches could avoidable, when GPU access is scheduled in the right moment e.g. top/bottom border

[Zerosquare]

You can try it and see if it does anything useful

[Cyprian_K]

that's the plan,

at the beginning I have to learn about GPU and OP GPU interrupt

[SCPCD]

STOREP/LOADP worth it when you can use it : if you can handle the restriction about those instruction.

 

I will not recommend to use DMAEN, I think that the headache to make it works as you want with all Tom bugs will not worth the time/boost ratio.

A better way is to optimize the GPU code.

 

 

 

Edited September 29 by SCPCD

[Cyprian_K]

ok thanks