The current implementation still uses the shifter forced to Shift 256 mode to keep things generic, but we can always opt for a separate scanline renderer.
Seems the Tseng cards are as incompatible as the S3 cards in terms of pure IBM VGA emulation, especially the chain4 stuff, so accomodate the fix to the renderer. Fixes 8bpp rendering on Tseng cards when invoked.
This gives more likely behaviour when one loads more often than the address is incremented.
The behaviour matches an Intel GMA 4500MHD (2008) which is the earliest hardware I'm able to test right now.
It appears that the S3 Trio does something weird with its 8bpp modes.
Specifically, it seems to ignore some of the flags needed for dword mode.
I will keep looking into that and see if I can find a good solution.
1. Made the 8514/A mode switch to VGA and viceversa (including ATI mode) actually usable and eliminating the last bugs related to them. Fixes 8514/A mode in OS/2 1.x/2.x (both IBM and Microsoft).
2. More sanity checks to the 8514/A and XGA pointers. Should fix emulator crashes on hard resets.
1. Its execution is exactly (actually 99%) the same as the S3 Bresenham procedure, so I thought of making the Bresenham command of the TGUI9440/96x0 behave like the s3 one but adapted to use Trident's ports and meanings and now OS/2 and Windows lines in certain programs work fine.
Bit 11 of ports 0x82e8 and 0x86e8 (Current X and Current Y positions registers, respectively) are taken into account again without conflicting with OS/2 Warp 3/4's software cursor, fixes blitting issues with Neptune Build 5111.1's S3 driver and possibly some other builds and other stuff.
the XGA memory test takes account of Win3.x' XGA memory test again without conflicting XGAKIT's test, fixes DOS-based boxes using Win3.x XGA drivers in 386 mode.
ATI Mach64 is temporarily unthreaded because of blitter desyncing.
ATI Mach32 now returns to ATI mode from IBM mode and vice versa more correctly.
Make the IBM 8514/A compatible poll not run if a 3D add-on card (Voodoo1/2) takes precedence (seen on a Mach32 PCI and Voodoo2 3D combo).
Finally ended the mess ups of the TGUI9440/96x0 pitch, it actually was the CRTC offset reg (SVGA CRTC 0x13) that defined it with some shifting based on the BPP.
1. Cirrus Logic GD54xx, Paradise/WD VGA now reset the interlace once a text mode is issued if not done automatically.
2. Paradise/WD's 15/16bpp modes using the 800x600 resolution now have the correct ma_latch, should fix most operating systems drivers using this combo.
3. More fixes (hopefully) to the accelerated pitch and rowoffset of the Trident TGUI cards (9440AGi and 96x0XGi), should fix issues with delayed displays mode changes under various operating systems (e.g.: Win3.1x).
4. Preliminary implementation of the Area Fill command of XGA, which is issued while using various painting and/or calc utilities on Win3.1x (IBM XGA updated drivers, e.g.: 2.12).
5. Preliminary (and incomplete) 4bpp XGA mode.
6. The XGA memory test for the 0xa5 using writes (used by various operating systems) no longer conflicts with DOS' XGAKIT's memory detection.
7. Small ROP fixes to both XGA and 8514/A.
8. Re-organized the mapping of the Mach32 chipset, especially when to enable the ATI mode or switching back to IBM mode, should fix LFB conflicts with various operating systems.
9. According to The OS/2 Museum, the Adaptec AHA-154xB series of SCSI cards fail the ASPI4DOS.SYS 3.36 signature check, so now make the changes accordingly.
10. Remove useless and crashy bios-less option of the Trantor T128.
11. The Image Manager 1024 card can also be used on a XT (although only if it has a V20/V30).
12. Re-organized the IBM PS/2 model 60 initialization as well as its right POS machine ID (though an update to sc.exe is still required for the POST memory amount to work normally).
Replaced svga->bpp == 24 with dev->accel_bpp == 24 as well as proper ATI graphics engine offsets added to 24bpp mode in the 8514/A rectangle fill commands (cmd 2, 3, 4) as well as on a command val of 8514/A bitblt (mainly 0xc2b5, actual cmd 6). This should fix 24bpp mode under Windows/NT and other operating systems, like OS/2 2.x/Warp.
Fixed the Win98 Trident PCI (and 9440 VLB) card accelerator pitch while maintaining compatibility with other OSes, despite being undocumented (this is at least an attempt to fix it properly based on the logs).
