Overview Features Coding Performance Forum Downloads Products OrderV4 Contact

Welcome to the Apollo Forum

This forum is for people interested in the APOLLO CPU.
Please read the forum usage manual.



All TopicsNewsPerformanceGamesDemosApolloVampireAROSWorkbenchATARIReleases
Information about the Apollo CPU and FPU.

Can I Have Some New AMMX Mnemonics?page  1 2 

Mr Niding

Posts 458
03 Aug 2017 14:00


I apologize for polluting this thread with non code posting, since this is far above my understanding but;

"i want to optimize the tinygl library"

Im just curious as to what programs/games will be able to utilize this optimization? I guess Quake which we all love to use as a performance scale. Anyhow, I hope you manage to tweak the library, and thanks for your efforts Andreas! :)


Gunnar von Boehn
(Apollo Team Member)
Posts 5408
03 Aug 2017 14:33


Andreas Timmermann wrote:

the problem in the code is the wrapping.
you need a compare and a branch every time.

 
Are you sure?
Can you give me an example to clarify this?


Andreas Timmermann

Posts 11
03 Aug 2017 16:04


not at this moment. first, the zbuffer must be finished ;-)


Gregthe Canuck

Posts 274
04 Aug 2017 04:24


Go Andreas!!  :)


Michael Borrmann

Posts 139
04 Aug 2017 07:58


Yeah, go for it!!! :)


Thellier Alain

Posts 126
04 Aug 2017 09:35


"i want to optimize the tinygl library"
NICE

>and then you have it. d4.w = YYXX
Nice too but then you are limited to 256x256 textures max
Also textures can repeat so UV values may have a value like (say) 4.0 meaning "repeat texture 4 times among the triangle"
So even a 256 pix texture can reach a 1024 value (here 256*4)

Certainly the most flexible solution would be to have
d0.l=XXXXxxxx
d1.l=YYYYyyyy
but then certainly  will need a new 68080 operator to get the pixel from (d0,d1,texture) at decent speed

About the Zbuffer: for the moment you only do the test/update zbuffer stuff but how do you keep a track of the test ?
I mean

for(segment size)
if(zbuffertest[n])
    drawpixel();

Alain Thellier



Andreas Timmermann

Posts 11
04 Aug 2017 15:32


thellier alain wrote:

    for(segment size)
    if(zbuffertest[n])
        drawpixel();
   

   
    you only draw at the end of the frame. so the algorithm goes like this:
   
  for every triangle
  {
      for every block in the triangle
      {
          calculate new_zmin and new_zmax
          if new_zmin > zbuffer_block_zmax
          {
                skip block
          }
          else
          if new_zmax < zbuffer_block_zmin
          {
                replace zbuffer_block with new_block (no zbuffer calculation needed, because all new values are smaller then all zbuffer_block_z)
                save only new_zmin/new_zmax
          }
          else
          {
                  insert new_block in zbuffer_blocklist and change zbuffer_block_zmin/max
          }
      }
  }
  update all zbuffer_block and render texture
   
    so, you don't need to update zbuffer_block's every time. and if there is no overlapping, you don't need to update zbuffer_block too :-)


Grzegorz Wójcik (pisklak
(Apollo Team Member)
Posts 84
04 Aug 2017 15:37


Whatever you do - good luck man !
Have some AMMX fun ! :)


Gunnar von Boehn
(Apollo Team Member)
Posts 5408
04 Aug 2017 15:43


thellier alain wrote:

  Certainly the most flexible solution would be to have
  d0.l=XXXXxxxx
  d1.l=YYYYyyyy
  but then certainly  will need a new 68080 operator to get the pixel from (d0,d1,texture) at decent speed

APOLLO has in instruction which can help here a lot PERM
PERM can pick any Bytes from 2 registers
PERM is single cycle and can be executed Super-scalar in both pipes = 2 instructions per cycle.

PERM should be useful for byte juggling you want to do in this Fix.Point math


posts 29page  1 2