[AMIGA] Winuae 2.7.1 beta 6/5

Gee Bee Air Rally (Activision, Inc., 1987)

WinUAE 2.7.1 beta 6 [02/02/14]

Probably fastest beta update ever.. B5 is broken in non-ce modes if chipset modes are used. Only RTG worked.

• Display emulation last horizontal position (plfright) was not correctly updated, caused all kinds of bad side-effects, including non-immediate/non-ce blitter taking seconds to finish.

WinUAE 2.7.1 beta 5 [02/02/14]

• "Strange right edge overscan display shift" adjusted, in some situations it was visible when it shouldn't have been. Looks also correct in NTSC mode now (Sawtooth pattern caused by alternating line lengths)
• OCS Agnus requires one cycle between sprite DMA and bitplane DMA, ECS Agnus can have both cycles back to back. This means OCS Agnus will only fetch first sprite word (practically making sprite useless) where ECS Agnus can fetch both.
• Rarely during heavy interrupt activity in cycle-exact mode some delayed interrupt register set or clear operation may have been lost causing possible spurious interrupt.
• lea (d8,An,Xn) prefetch is last operation, not idle cycle.
• 68020 CE mode CAS instruction prefetch was broken.
• When UAE boot ROM is enabled, add fake autoconfig board (configdev) that has same board address and size as boot ROM. Allows MMU utilities (mmu libraries, enforcer etc..) to mark boot ROM as valid address space without extra command line parameters.
• Partially rewritten horizontal Agnus bitplane state machine. Can (as usual) break something..
• When BPLCON1 matches, only copy enabled planes to shift registers. (Back In 90 / Samplers)
• Power led didn't turn off after soft reset.
• Added sprite-only linetoscr routines, bordersprites and HAM do not mix well. (Satyanarayanan / Scoopex)
• AGA and 64-bit fetch modes: last 16 or so pixels were not flushed to display, depending on BPLCON1 value. (Satyanarayanan / Scoopex)
• New (still WIP) modern native library interface merged from FS-UAE.
• Added LSI 53C895A SCSI chip emulation from QEMU. Lots of changes to make it 53C710 compatible (53C720 and later have multiple registers changes but SCRIPTS code is fortunately backwards compatible) A4091 and A4000T NCR driver now "work", chip init works, returns timeout error (no device connected) when it tries to detect SCSI devices. Do not connect any drives, it does not work yet.
• pdk compressed (or dsq) files always unpacked to Amiga DD floppy size.
• Keyboard leds lost sync in some situations.

WinUAE 2.8.0 beta 19 [30/04/14]

• Increased space for UAE boot ROM extension resident module structures, it was too small and overflowed (causing boot crash) if too many UAE resident module expansions were enabled simultaneously.
• Windows specific configuration defaults were not set if config was command line or double click loaded (missed in b17 fix)
• AROS ROM updated.

Winuae 64 bits differences (authors comment):

• No JIT, can support up to 2.5G of Z3 address space (any AOS supported combination of Z3 fast, Z3 chip and RTG total can be 2.5G), possibly slightly faster because x64 has more registers (big and fast L1 caches can hide it).
  

Human Killing Machine (GO!, 1989)

WinUAE 2.8.0 beta 18 [25/04/14]

• 0xFFFF was not returned when reading write-only or non-existing custom register and previous cycle was idle cycle. Not exactly 100% correct but more compatible with programs that accidentally read wrong custom registers.
• Cirrus Logic emulation had some useless logging enabled.
• Added GUI option to enable/disable new automatic programmed mode resolution selection to Display panel, enabled by default. (For example if you want to use superhires all the time and use filter panel to scale the display back to correct aspect ratio)
• Unexpanded A1200 also reads last fetched opcode word when accessing non-existing memory, similar to unexpanded A500. (more compatible and cycle-exact only)
• Last possible displayable line was incorrectly drawn in border color in some situations when short field mode was active.

WinUAE 2.8.0 beta 17 [24/04/14]

• Debugger breakpoints work again. (b14)
• Full CPUSHA/CPUSHL/CPUSHP disassembly implemented.
• Integer scale mode compatibility improved with programmed display modes.
• Integer scale mode changes Filter panel zoom slider range to -99 to 99 and is used to adjust integer scaling calculation. For example -10 means if next largest scaling factor is selected even if display height is 10% too big to fit normally. (This was implemented already but it didn't work properly)
• Fixed fill_line() top/bottom background line drawing buffer overflow in some programmed modes.
• Some programmed modes had single short background colored line in left horizontal blanking area.
• Command line (or icon doubleclick) loaded config didn't reset previously loaded config (default.uae, if it existed)
• Emulated more Fat Gary address regions that are unmapped (CIA is same as Gayle, upper 32k in clock space and Gary register space are unmapped, 0xdf0000-0xdf7fff is also unmapped)
• Emulate Fat Gary invalid address space access timeout if Gary timeout bit is set, first 50 timeouts are always logged. To prevent last moment stupid bugs or bug reports, delay is not yet enabled

WinUAE 2.8.0 beta 16 [17/04/14]

• Borderblank blanked sprites incorrectly in some situations (b11, Roots AGA)
• SPRxDATx/SPRxPOS copper write/Denise internal sprite timing should now fully match real hardware.
• Ancient (and recently non-working) Superfrog flashing intro bee hack removed, properly emulated now.
• gfx_filter_keep_autoscale_aspect and gfx_filter_keep_aspect config entry change was not detected when using using uae-configuration.
• Last used AVI codec was not restored from registry if ICGetState() returned zero size data block.

Hybris (Discovery Software International, Inc., 1988)

WinUAE 2.8.0 beta 15 [13/04/14]

• Finally bumped version to 2.8.0. This isn't small 2.7.0 update (Should have been done since b6 or so..)
• FPU register negative zero transformed to positive zero when it was written to or read from memory in extended FP format (FMOVE.X FPx,<ea> or FMOVE.X <ea>,FPx or FMOVEM). No one should care as usual.
• When loading FPU register from memory in extended format, normalize any "unnormal" zero. (non-zero exponent but zero mantissa part). Motorola FPUs do this automatically, x86 FPU does not seem to like them. (Another mostly Next specific update)
• Programmed mode autoresolution should now work with strange modes like highgfx.
• Only allow CPU mode changes between instructions, previously change sometimes happened mid-instruction (during instruction internal memory read/write) which wasn't safe.
• Small tweak to recent sprite update (Hybris)