Beta 2:
Originally Posted by :
Background information:
After many days of logic analyzer work, blitter should finally be 100% cycle-exact, even in line mode.
DMA emulation cycle-exactness is greatly improved, most vector routines that had slowdowns, flickering or graphics garbage should work perfectly now (even Los Huivos/Virtual Dreams doubleglenz work, this has shown only major graphics garbage previously)
There are still some that won´t still work or work even worse, these could be caused by CPU emulation timing errors (even single cycle difference in single instruction that is used regularly enough can make the difference)
"Standard test" programs still not working: Rampage/TEK, Hulkamania/TSP (and left border garbage in one part of Absolute Inebriation/VD, reason is known but I am not sure how to implement the fix)
- "DMA cycle debugger", v [] lists selected scanline´s DMA activity, hpos, custom register, data, address. This made compatibility testing much easier. (just compare this data to logic analyzer data) Second row can contain extra characters: N = blitter cycle given for CPU, b = blitter interrupt, B = blitter finished. (not necessarily exactly same thing)
- chipset bitplane/copper/blitter cycle exact mode DMA sequence routines rewritten. Emulation may be slower due to some missing optimizations. Later.
- blitter cycle diagrams rechecked with logic analyzer, errors fixed
- CPU emulation fixes, lsl/asl/ror and friends had wrong cycle counts. bclr/bset/bchg are 2 cycles shorter if bit number is less than 16 (16-31 = add extra cycles) and more. Lots of testing to do..
- WASAPI sample format fallback didn´t work correctly
- command line parser now really works exactly like Windows console
- added "win32.exec_before" and "win32.exec_after" configuration entries, "before" command is run when emulation starts, "after" when emulation shuts down or before another config is loaded. Multiple commands can be separated with ´;´. Waits until command returns (=waits forever if program never exits..)
- interlace artifact removal works again (b1)
- automatically handle split dms files. If extracted data is only about half of standard DD disk and if file name is "*a.dms", attempts to open and merge "*b.dms" with already unpacked data
- do not enumerate openal recording devices
