mirror of https://github.com/mpv-player/mpv
314 lines
11 KiB
Plaintext
314 lines
11 KiB
Plaintext
ATI chips hacking
|
|
=================
|
|
Dedicated to ATI's hackers.
|
|
|
|
Preface
|
|
~~~~~~~
|
|
This document will compare ATI chips only from point of DAC and video overlay.
|
|
There are lots of difference from 3D point, dual-head support, tv-out support
|
|
and many other things but it's already perfectly different story.
|
|
This document doesn't include information about ATI AIW (All In Wonder) chips.
|
|
|
|
What are units on modern ATI chips:
|
|
DAC - (Digital to Analog Convertor) controls CRTC, LCD, DFP monitor's output
|
|
Consists from:
|
|
PLL - (Programable line length) registers
|
|
CRTC - CRT controller
|
|
LCD/DFP scaler
|
|
surface control
|
|
DAC2 - controls CRTC, LCD, DFP monitor's output on second head
|
|
TVDAC - controls Composite Video and Super Video output ports
|
|
Consists from:
|
|
TV_PLL
|
|
TV scaler & sync unit
|
|
TV format convertor (PAL/NTSC)
|
|
TVCAP - controls Video-In port
|
|
MPP - Miscellaneous peripheral port. (includes macrovision's filter - copy
|
|
protection mechanism)
|
|
OV - Video overlay (YUV BES) (include subpictures, gamma correction and
|
|
adaptive deinterlacing)
|
|
CAP0 - Video capturing
|
|
CAP1 - Video capturing (second unit)
|
|
RT - Rage theatre: video encoding and mixing
|
|
MUX - video muxer
|
|
MEM - PCI/AGP bus mastering
|
|
2D - GUI engine
|
|
3D - 3D-OpenGL engine (There are lots of stuff)
|
|
I2C - I2C Bus control
|
|
|
|
This document is mainly related only with OV unit ;)
|
|
Video decoding diagram:
|
|
|
|
RAM memory: [ App ] Copies YUV image to overlay memory
|
|
| <-- (It's possible to program DMA here)
|
|
overlay memory:[ OV ] performs scaling and YUVtoRGB convertion
|
|
/\
|
|
RGB memory: / \
|
|
/ [ macrovision ] performs copy protection filtering
|
|
/ \ (unneeded but presented by default thing;)
|
|
[ CRTC/LCD/DFP DAC ] [ TV DAC ] convert RGB memory to CRTC and NTSC/PAL signals
|
|
| |
|
|
[CRTC/LCD/DFP Monitor] [TV-screen]
|
|
|
|
History
|
|
~~~~~~~
|
|
What is history of ATI's chips? I can be wrong but below is my vision
|
|
of this question:
|
|
|
|
0. I don't know any earlied chips :(
|
|
1. Mach8
|
|
2. Mach16
|
|
3. Mach32
|
|
|
|
4. Mach64.
|
|
It's first chip which has support from side of open
|
|
source drivers. Set of mach64 chips is:
|
|
mach64GX (ATI888GX00)
|
|
mach64CX (ATI888CX00)
|
|
mach64CT (ATI264CT)
|
|
mach64ET (ATI264ET)
|
|
mach64VTA3 (ATI264VT)
|
|
mach64VTA4 (ATI264VT)
|
|
mach64VTB (ATI264VTB)
|
|
mach64VT4 (ATI264VT4)
|
|
|
|
5. 3D rage chips.
|
|
It seems that these chips have fully compatible by GPU with Mach64
|
|
which is extended by 3D possibilities. Set of 3D rage chips is:
|
|
3D RAGE (GT)
|
|
3D RAGE II+ (GTB)
|
|
3D RAGE IIC (PCI)
|
|
3D RAGE IIC (AGP)
|
|
3D RAGE LT
|
|
3D RAGE LT-G
|
|
3D RAGE PRO (BGA, AGP)
|
|
3D RAGE PRO (BGA, AGP, 1x only)
|
|
3D RAGE PRO (BGA, PCI)
|
|
3D RAGE PRO (PQFP, PCI)
|
|
3D RAGE PRO (PQFP, PCI, limited 3D)
|
|
3D RAGE (XL)
|
|
3D RAGE LT PRO (AGP)
|
|
3D RAGE LT PRO (PCI)
|
|
3D RAGE Mobility (PCI)
|
|
3D RAGE Mobility (AGP)
|
|
|
|
6. Rage128 chips.
|
|
These chips have perfectly new GPU which supports memory mapped IO
|
|
space for accelerating port access (It's main cause of incompatibility
|
|
with mach64). Set of Rage128 chips is:
|
|
Rage128 GL RE
|
|
Rage128 GL RF
|
|
Rage128 GL RG
|
|
Rage128 GL RH
|
|
Rage128 GL RI
|
|
Rage128 VR RK
|
|
Rage128 VR RL
|
|
Rage128 VR RM
|
|
Rage128 VR RN
|
|
Rage128 VR RO
|
|
Rage128 Mobility M3 LE
|
|
Rage128 Mobility M3 LF
|
|
7. Rage128Pro chips.
|
|
These chips are successors of Rage128 ones.
|
|
Rage128Pro GL PA
|
|
Rage128Pro GL PB
|
|
Rage128Pro GL PC
|
|
Rage128Pro GL PD
|
|
Rage128Pro GL PE
|
|
Rage128Pro GL PF
|
|
Rage128Pro VR PG
|
|
Rage128Pro VR PH
|
|
Rage128Pro VR PI
|
|
Rage128Pro VR PJ
|
|
Rage128Pro VR PK
|
|
Rage128Pro VR PL
|
|
Rage128Pro VR PM
|
|
Rage128Pro VR PN
|
|
Rage128Pro VR PO
|
|
Rage128Pro VR PP
|
|
Rage128Pro VR PQ
|
|
Rage128Pro VR PR
|
|
Rage128Pro VR TR
|
|
Rage128Pro VR PS
|
|
Rage128Pro VR PT
|
|
Rage128Pro VR PU
|
|
Rage128Pro VR PV
|
|
Rage128Pro VR PW
|
|
Rage128Pro VR PX
|
|
Rage128Pro Ultra U1
|
|
Rage128Pro Ultra U2
|
|
Rage128Pro Ultra U3
|
|
|
|
8. Radeon chips.
|
|
Indeed they could be named Rage256 Pro. (With minor changes is fully
|
|
compatible with Rage128 chips).
|
|
Radeon QD
|
|
Radeon QE
|
|
Radeon QF
|
|
Radeon QG
|
|
Radeon VE QY
|
|
Radeon VE QZ
|
|
Radeon M6 LY
|
|
Radeon M6 LZ
|
|
Radeon M7 LW
|
|
9. Radeon2 chips.
|
|
Indeed they could be named Rage512 Pro.
|
|
Radeon2 8500 QL
|
|
Radeon2 7500 QW
|
|
|
|
10. Radeon3 and newest are cooming soon, but I hope that they will be fully
|
|
compatible with Radeon1 chips.
|
|
|
|
In Radeon famility there were introduced also FX chips: Radeon FX and
|
|
Radeon2 8700 FX. Probably they have the same possibility as other Radeon
|
|
but currently it's unknown for me.
|
|
|
|
What about video overlay and DAC?
|
|
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
Currently it's known that there is only difference between
|
|
Mach64 and Rage128 compatible chips:
|
|
- They have different logic of io ports programming!
|
|
- They are incompatible by port numbers!
|
|
But:
|
|
- They use the same program logic from register's name point.
|
|
(Indeed exists slight difference even between Radeon and Rage128
|
|
chips. AFAIK only Radeon has OV0_SLICE_CNTL register which currently
|
|
is not used by driver. But I know only its name ;). Also there
|
|
is difference in slight adjust of BES position but it's configured
|
|
by #ifdef blocks).
|
|
|
|
Please compare:
|
|
|
|
(The piece of Back-End Scaler programming)
|
|
|
|
Sample for Mach64 compatible chips:
|
|
***********************************
|
|
|
|
#define SPARSE_IO_BASE 0x03fcu
|
|
#define SPARSE_IO_SELECT 0xfc00u
|
|
|
|
#define BLOCK_IO_BASE 0xff00u
|
|
#define BLOCK_IO_SELECT 0x00fcu
|
|
|
|
#define MM_IO_SELECT 0x03fcu
|
|
#define BLOCK_SELECT 0x0400u
|
|
#define DWORD_SELECT (BLOCK_SELECT | MM_IO_SELECT)
|
|
|
|
#define IO_BYTE_SELECT 0x0003u
|
|
|
|
#define SPARSE_IO_PORT (SPARSE_IO_BASE | IO_BYTE_SELECT)
|
|
#define BLOCK_IO_PORT (BLOCK_IO_BASE | IO_BYTE_SELECT)
|
|
|
|
#define IOPortTag(_SparseIOSelect, _BlockIOSelect) \
|
|
(SetBits(_SparseIOSelect, SPARSE_IO_SELECT) | \
|
|
SetBits(_BlockIOSelect, BLOCK_SELECT | MM_IO_SELECT))
|
|
#define SparseIOTag(_IOSelect) IOPortTag(_IOSelect, 0)
|
|
#define BlockIOTag(_IOSelect) IOPortTag(0, _IOSelect)
|
|
|
|
...
|
|
|
|
#define OVERLAY_Y_X_START BlockIOTag(0x100u)
|
|
#define OVERLAY_Y_X_END BlockIOTag(0x101u)
|
|
|
|
...
|
|
|
|
#define OUTREG(_Register, _Value) \
|
|
MMIO_OUT32(pATI->pBlock[GetBits(_Register, BLOCK_SELECT)], \
|
|
(_Register) & MM_IO_SELECT, _Value)
|
|
|
|
...
|
|
|
|
OUTREG(OVERLAY_Y_X_START,((drw_x)<<16)|(drw_y)|(1<<31));
|
|
OUTREG(OVERLAY_Y_X_END,((drw_x+drw_w)<<16)|(drw_y+drw_h));
|
|
|
|
|
|
Sample for Rage128 compatible chips:
|
|
************************************
|
|
|
|
#define OV0_Y_X_START 0x0400
|
|
#define OV0_Y_X_END 0x0404
|
|
|
|
...
|
|
|
|
#define INREG(addr) readl((rage_mmio_base)+addr)
|
|
#define OUTREG(addr,val) writel(val, (rage_mmio_base)+addr)
|
|
|
|
...
|
|
|
|
rage_mmio_base = ioremap_nocache(pci_resource_start (dev, 2),RAGE_REGSIZE);
|
|
|
|
...
|
|
|
|
#ifdef RADEON
|
|
#define X_ADJUST 8
|
|
#else /* rage128 */
|
|
#define X_ADJUST 0
|
|
#endif
|
|
|
|
OUTREG(OV0_Y_X_START,(drw_x+X_ADJUST)|(drw_y<<16));
|
|
OUTREG(OV0_Y_X_END,(drw_x+drw_w+X_ADJUST)|(drw_y+drw_h)<<16));
|
|
|
|
Thus - these chips have almost the same logic from register's name point.
|
|
(except the fact that they have swapped 16-bit halfs).
|
|
Yes - programming of Rage128 is much simpler of Mach64.
|
|
|
|
|
|
What about other ATI's chips?
|
|
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
I suggest you have latest copy of GATOS-CVS:
|
|
http://www.linuxvideo.org
|
|
GATOS was designed and introduced as General ATI TV and Overlay Sowfware.
|
|
You will be able to find out there a lots of useful hacking utilities
|
|
(at location gatos-ati/gatos):
|
|
gfxdump - Program for dumping graphics chips registers on Linux and Windows 9X.
|
|
(it's more useful for Win9x to hack their values).
|
|
xatitv - For working with tv-in (currently is under hard development)
|
|
atitvout- For working with tv-out
|
|
and lot of other stuff.
|
|
BUT: After studing of Gatos and X11 stuffs I've found that they are bad
|
|
optimized for movie playback.
|
|
Please compare:
|
|
radeon_vid - configures video overlay only once and provides DGA to it.
|
|
(doesn't require to be MMX optimized)
|
|
gatos and X11 - configures video overlay at every slice of frame, then
|
|
performs unoptimized copying of source stuff to video memory
|
|
often with using CopyMungedData (it's C-analog of YV12_to_YUY2)
|
|
since there are lacks in yv12 support.
|
|
(is not MMX optimized that's gladly accepted, but probably
|
|
will be never optimized due portability).
|
|
|
|
hardware IDCT support diagram:
|
|
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
|
|
|
|
|
[ Video parser ] <---------- [ Transport demuxing ] --> [ Audio ]
|
|
| | |
|
|
[ Variable length decoder] |D |
|
|
| |V |
|
|
[ Inverse quantization ] |D |
|
|
| | |
|
|
-------|---[ video card ]---------+ |s |
|
|
| | |u |
|
|
[ Run level decode & de-zigzag ] | |b |
|
|
| | |p |
|
|
[ IDCT ] | |i |
|
|
| | |c |
|
|
[ Motion compensation ] | |t |
|
|
| | |u |
|
|
[ Advanced deinterlacing ] | |r |
|
|
| | |e |
|
|
[ Filtered X-Y scaling ] [SUBPIC]-|-----+s [ OSD ]
|
|
| | | | |
|
|
[ 4-bit alpha blending ] <---+ | +-------+
|
|
| |
|
|
[ YUV to RGB conversion ] |
|
|
-------|--------------------------+
|
|
TV-screen or CRT-display
|
|
|
|
|
|
Conslusion:
|
|
~~~~~~~~~~~
|
|
|
|
That's all folk!
|