IBM.H is gone, video stuff re-organized. Keyboard stuff reorganized.
Machines that have their own video, mouse and/or keyboard now have all this in their machine file.
Fixed and other cleanups here and there.
Fixed (hopefully) pass through in the Windows CD-ROM IOCtl code;
Fixed the legacy I/O address handling on the BT-958D, now it behaves correctly - legacy I/O address disabled by default, can be enabled in AutoSCSI (removed the setting from the device config struct);
The BT-958D PCI-configured I/O space now correctly gets 32 ports and not 4.
Fixed MCA write bugs for the AHA-1640 and BT-640;
Fixed a warning in the PC87306 Super I/O chip emulation;
Each renderer now only has a 32-bit blitter - video_blit_memtoscreen_8() now only converts buffer to buffer32 and then calls video_blit_memtoscreen(), 8-bit blitters are now gone.
Fixed the MCA write function of the AHA-1640 MCA - the host ID is now read from the correct variable and the sync and parity check values are now read as well.
The AHA and BusLogic SCSI controllers now support the bus reset (0x81) CCB opcode;
The AHA and BusLogic SCSI controllers now terminate with the correct mailbox completion codes on invalid ID, invalid LUN, invalid opcode, or invalid direction control byte.
The CD-ROM buffer is now dynamically malloc'd;
Did a lot of bugfixing in cdrom.c and scsi_disk.c, Debian 8's modprobe aha1542 and the NT 3.10.404.1 Trantor 310b driver now work;
Fixed the IDE IRQ/MIRQ mess and ATAPI DMA transfer length;
Fixed the operation of mmutranslate() a bit;
Commented out several instances of excess logging;
Applied the mainline PCem commit prefetch commit;
Readded missing Sound Blaster DSP MIDI behavior and gave it the correct MPU struct;
All non-IBM 486 machines now correctly have the VLB flag set.
The CD-ROM and SCSI disk REQUEST SENSE commands no longer incorrectly set the phase to status;
REQUEST SENSE with non-data direction control byte is now treated as a request to send sense bytes from the command directly to the sense buffer;
SCSI and SLiRP threads now terminate themselves again when out of the while loop;
SCSI and SLiRP mutexes are now closed when the threads terminate themselves;
Fixed a bug causing the "Press F12+F8 or middle mouse button to release" string on the title bar to get corrupt.
Fixed AHA-154x scatter/gather transfers;
Two-phased SCSI read-direction commands;
Made scsi_bus.c use a dynamically malloc'd buffer instead of a giant fixed-size buffer;
The AHA-154x/BusLogic thread now uses mutexes;
The BusLogic BT-545C is now the BT-545S;
Added the BusLogic BT-542BH;
The AHA-1542CF now again uses the v2.11 BIOS as the CD booting now works;
Applied PCem commit that adds NMI support to the recompiler;
Applied PCem commit that adds the IBM XT 286;
Applied PCem commits that have to do with sound, including the ES1371;
Fixed the NCR5380 output data register;
Moved the SLiRP mutex stuff from win.c to the appropriate network files;
Added sanity checks to everything in win_thread.c.
Cleaned up the handlers for the BusLogic Execute SCSI Command command;
Made the AHA-1542CF use version 2.01 of the BIOS;
The CD-ROM GET EVENT STATUS NOTIFICATION fix is back.
The Settings dialog now correctly filters devices by bus;
Split the dual VLB/PCI graphics cards into VLB and PCI versions;
Device filtering is now done using the new device_is_valid() call that compares device flags to machine flags;
Moved the NCR-based SCSI controllers to the main branch;
Moved the NE1000 to the dev branch until it's finished.
Two-phases SCSI disk write direction commands (needed for the NCR5380);
Renamed cdrom_ioctl.c/h to win_cdrom_ioctl.c/h and moved them to win/;
Moved some CD-ROM and floppy initialization blocks to the appropriate files.
