Is their somewhere an overview which video-modes the MorphOS CyberGFX Gfx-card drivers do support in which MorphOS version?

For example I want to know which GFX-card driver supports 16bit color modes with big-endian GFX-mem layout and which GFX driver runs in little Endian slow mode.


The reason for this is that little endian modes of course need a endian conversion which slows down the video a lot.
SDL uses for endian conversion a 128KB lookup table.
The runtime conversion doubles the required memory bandwidth so it basicly kills most games.


Many thanks for the info

Gunnar


[ Edited by BigGun on 2006/4/22 17:44 ]

The quickest way to generate such an overview would be to get ppl with different kinds of gfxcards to run "sys:tools/showcgxconfig debug" and paste the "Available pixel formats" line .. f.ex. right here in this thread. ;)


- CISC

I fear you ask the wrong question to solve the wrong problem :-)
What you really want is to avoid on the fly pixel format conversion during blits.
So, you should check SDL_DisplayFormat() docs.

CISC: Thanks for the tip:

Okay SO let me begin to send some info then:


cgxsystem.library 50.1 (PPC mode)

1 graphics device(s) installed

DEVICES:
Voodoo3: Driver Version 4.9 (PPC mode)
Prod=4634/5($121a/$5)
BaseName=Voodoo3
CardID=14($e) Number=1($1)
Developed by 3Dfx Interactive, Inc.
BoardBase=$ebff0000
DrvrBase=$20875350
Available pixel formats: LUT8 RGB16PC ARGB32
Supports: Doublescan Gamma
Maximum memory available: ~16.0 MBytes
Memory clock: ~166.2 MHz

Henes' suggestion is probably more sensible though... ;)


- CISC

@Henes,

Display format conversion during blits is *very slow* of course.

But to be honest I'm abit worried about SDLs feature to emulate big endian GFX memory on little endian GFX-Cards.
I think this feature is really slow, isn't it?

If I look at the pixel swapping routines in cgximage.c
The old Amiga version used the work with 128kb swapping tables even - LOL. But the new PowerSDL is still not optimal.

The surface is read (one read per pixel),
the pixel is swapped, and then written back (one write per pixel)
This is very uneffective of course and could be done with one memory access per two pixels. And on PPC we could save all the swap command by using loads that do endianconvertion.

I'll need to look a bit more into this to get a better understanding ...


Henes, I have trouble compiling PowerSDL
I'm getting this error.

IMG_jpg.c :174 'JFIF_BASE_NAME' undeclared

Do you know by chance how to fix this?


Cheers
Gunnar

Okay here the CyberGFX information of another card

cgxsystem.library 50.1 (PPC mode)

1 graphics device(s) installed

DEVICES:
Radeon: Driver Version 4.1 (PPC mode)
Prod=4098/20825($1002/$5159)
BaseName=Radeon
CardID=9($9) Number=1($1)
Developed by ATI
BoardBase=$e7ff0000
DrvrBase=$2085cc8c
Available pixel formats: LUT8 RGB15 RGB16 BGRA32
Supports: Doublescan Gamma
Maximum memory available: ~32.0 MBytes
Memory clock: ~182.7 MHz

Here is mine, Radeon 9000:

Installed libraries:
cgxsystem.library 50.1 (PPC mode)

1 graphics device(s) installed

DEVICES:
Radeon: Driver Version 4.1 (PPC mode)
Prod=4098/18790($1002/$4966)
BaseName=Radeon
CardID=9($9) Number=1($1)
Developed by ATI
BoardBase=$ebfdb000
DrvrBase=$20774b04
Available pixel formats: LUT8 RGB15PC RGB16PC BGRA32
Supports: Doublescan Gamma
Maximum memory available: ~64.0 MBytes
Memory clock: ~200.7 MHz

Radeon 8500

cgxsystem.library 50.1 (PPC mode)

1 graphics device(s) installed

DEVICES:
Radeon: Driver Version 4.1 (PPC mode)
Prod=4098/20812($1002/$514c)
BaseName=Radeon
CardID=9($9) Number=1($1)
Developed by ATI
BoardBase=$e7fea000
DrvrBase=$20440154
Available pixel formats: LUT8 RGB15PC RGB16PC BGRA32
Supports: Doublescan Gamma
Maximum memory available: ~64.0 MBytes
Memory clock: ~275.0 MHz

Radeon 9200

cgxsystem.library 50.1 (PPC mode)

1 graphics device(s) installed

DEVICES:
Radeon: Driver Version 4.1 (PPC mode)
Prod=4098/22881($1002/$5961)
BaseName=Radeon
CardID=9($9) Number=1($1)
Developed by ATI
BoardBase=$e7fef000
DrvrBase=$2034c354
Available pixel formats: LUT8 RGB15PC RGB16PC BGRA32
Supports: Doublescan Gamma
Maximum memory available: ~128.0 MBytes
Memory clock: ~200.7 MHz

Radeon 9200 SE
Available pixel formats: LUT8 RGB15PC RGB16PC BGRA32

I do not understand what SDL emulation feature you are talking about.

Just use SDL_DisplayFormat() at load time to convert all your bitmaps. Not in-game. Then you do not care if it is dead slow or not.
RTFSDLM :-p

About the error... look at the source, see the define around it and disable it in the makefile.

@Henes

LOL, I know about SDL_DisplayFormat()

What I was talking about was SDL ability to emulate a big endian gfx-memory to the user (developer) even if the gfx-card is in reality in little endian.

INFO of another MOS-supported GFX_CARD

cgxsystem.library 50.1 (PPC mode)

1 graphics device(s) installed

DEVICES:
SiS6326: Driver Version 4.4 (PPC mode)
Prod=4153/25382($1039/$6326)
BaseName=SiS6326
CardID=9($9) Number=1($1)
Developed by Silicon Integrated Systems Corp.
BoardBase=$ef7f0000
DrvrBase=$20873d20
Available pixel formats: LUT8 RGB15PC RGB16PC BGR24
Supports: Doublescan Interlace Gamma
Maximum memory available: ~4.0 MBytes
Memory clock: ~66.0 MHz

And Another GFX card.

This time an ELSA GLORIA SYNERGY SA
AGD Card (with Permedia2 Chip)

cgxsystem.library 50.1 (PPC mode)

1 graphics device(s) installed

DEVICES:
Permedia2: Driver Version 4.4 (PPC mode)
Prod=4172/15623($104c/$3d07)
BaseName=Permedia2
CardID=12($c) Number=1($1)
Developed by 3Dlabs
BoardBase=$ef7d0000
DrvrBase=$2085cbe0
Available pixel formats: LUT8 RGB15 RGB16 ARGB32
Supports: Doublescan Gamma
Maximum memory available: ~8.0 MBytes
Memory clock: ~83.0 MHz