WinUAE 3.4.0 Beta 5-6 - The latest Amiga emulator update

It's a Sunday and it's a late evening, but that doesn't stop us from mentioning the latest WinUAE release of ' WinUAE 3.4.0 Beta 5 and 6 which has recently surfaced on the Amiga forums. Toni Wilen is still hard at work as always and is progressing rapidly towards a finished final Amiga Emulator 3.4.0. Not as if that will stop him from doing another one later in the year however as I'm sure 3.5.0 beta will be along next!

Beta 5:

- Buddha Z2 IDE controller emulation added. Catweasel Z2 model optionally selectable (Has nearly identical Buddha hardware, CW part is not emulated). No ROMs added, flash rom images are available from manufacturer's web site. IDE doubler is not emulated.
- Reset drive sound state when emulated Amiga is reset.
- Z3 mapping = Automatic didn't choose UAE mode if JIT direct was enabled and Real mapping would not fit in allocated natmem space but UAE mapping would fit.
- Select UAE mode if Z3 mapping is automatic and UAE fits inside 2G barrier but Real does not. (To better match pre-3.4 behavior, selecting Real in this mode can cause non-booting config if config has enough Z3 RAM)
- 3G of Z3 RAM is now possible (3x1G Z3 boards). 64-bit WinUAE only, can't be fully JIT direct compatible.
- Hardware info GUI panel shows also last Z3 address.
- Hardware info shows Z3 board status info: dark background = outside of 2G border (can cause problems because some exec memory functions that use bit 31 for error status, reset proof programs mainly use them, usually causes reboot loop). Yellow background = board does not fit in address space (32-bit overflow, start address is also 0xffffffff in this situation). Red background = memory allocation failed. Lighter text = JIT direct enabled but memory is not JIT direct capable.
- SCSI emulation request sense always returned allocation field number of bytes even if command returned less data.
- If ATAPI CD data transfer size is odd (which is not valid), ignore last byte.
- NE2000 ISA is now available in x86 bridgeboard emulation. 100% untested!
- Portaudio driver converted to pull mode.
- Added Blizzard SCSI Kit III emulation and boot ROM v1.139.
- Blizzard 1230 MK III ROM layout changed to match SCSI Kit III ROM. (Still no proper dump but I am sure this is correct)
- Fixed M-TEC IDE controller emulation rom handling.
- Low latency vsync + no buffer + WASAPI pull mode should be stable again. (RTG mode: todo)
- Sound audio switch off didn't select correct timing method in WASAPI pull mode.
- Small Z2 RAM (256k or smaller that goes in IO Z2 space) shadowed f00000 UAE Boot ROM in some configurations.
- If only one or more Z3 RAM boards are outside of natmem and JIT direct is enabled: don't disable JIT direct completely but only force non-JIT direct capable Z3 memory banks to use indirect mode.
- Show Tecmar T-Card as "Zorro I" (prototype/unreleased version of autoconfig) board in Hardware info GUI.
- Certain accelerators with autoconfig board didn't appear correctly in Hardware info panel.
- Z2/Z3 memory bank that failed to allocate (out of memory) caused a crash.
- Save image extended adf image was created with smaller max PAL size if emulation was not already started and in NTSC mode.
- Some fake-DMA harddrive controllers with internal buffers crashed the emulation if GUI was opened during internal DMA operation and CPU mode was compatible/cycle exact.
- Separated memory bank pre-allocation and after allocation memory size variables.
- Return correct SASI request sense data.

Blizzard SCSI Kit III:
- PIO-only! I thought all Phase 5 SCSI controllers were true DMA...
- FAS408 SCSI chip which is FAS216 + 128 byte PIO FIFO buffer. (was annoying to emulate without datasheet..)
- v1.39 ROM ("$VER: A1233_VERSION 1.139 (9.5.95)") added to ROM scanner.
- Blizzard 1230 "MK I" to go.. (Does anyone have it? With or without SCSI Kit)

Beta 6: (b5 fixes only)

- Workaround for possible NCR53C720+ emulation bug when read is shorter than command's allocation size. CSPPC SCSI driver gets confused if it causes phase mismatch error. (b5 REQUEST SENSE fix caused this side-effect)
- CSMK3/CSPPC/BPPC without UAE Boot ROM: accelerator boot ROM didn't work very well.
- Picasso IV enabled + reset: always caused hard reset.
- If extended ADF's track space is smaller than write length, ignore first bits in sector buffer (it is most likely part of the gap).
- Fill REQUEST SENSE Information Bytes (if SCSI) or Address bytes (if SASI) if error was LBA related.
- x86 VGA board didn't correctly skip autoconfig. (It is technically shared with normal autoconfig boards)
- b5 memory bank update broke ECS 0.5M+0.5M config.
- Compiled again with pre-release MSVC compiler. Does speed now match 3.3.0? (Few reported that 3.4 betas have slower performance) This breaks most SHIFT+<something> shortcuts.. (It appears to be optimizer bug related to 64-bit variables in 32-bit mode) Temporary change, either I'll add workaround or I'll use official version or final/fixed version will be released soon. Only confirming if this helps with performance. (3.3.0 was accidentally compiled with pre-release version..)

- Thanks Vintageisthenewold for the heads up

Links : 1) Forum 2) 32bit 3) 64bit