Added new cmd-line switch: -videorom <file> to replace the video ROM for the Enhanced //e.
- Support video ROM sizes of 4K, 8K and 16K (top 8K only).
- NB. The rocker switch is set to European video ROM.
F10 (for //e or Enhanced //e models) emulates the PAL //e's rocker switch (under the keyboard) to toggle between European or US video ROM.
Other:
- Fixed debugger's view of the AltCharSet soft-switch (it was showing the opposite state).
So added a 2nd "Pure CPU MHz" for full-speed (ie. no video update), eg:
---------------------------
Benchmarks
---------------------------
Pure Video FPS: 2860 hires, 2868 text
Pure CPU MHz: 10.3 (video update)
Pure CPU MHz: 241.3 (full-speed)
EXPECTED AVERAGE VIDEO GAME
PERFORMANCE: 476 FPS
(ref: #424, #540)
Changes to ease code compilation in GCC.
Disk_t and HDD contain a std::string and for this reason they need a proper constructor, ZeroMemory is not guaranteed to work.
. .dsk / .nib images
. ProDOS format / DOS 3.3 init
. authentic / enhanced disk access speed
For zero-length files, resize to the complete file size when first opened (#506)
. Support both .dsk and .nib
Created a new class FormatTrack to encapsulate the new track formatting logic
Improved precision of 'authentic' drive mode's spin emulation (#125)
Save-state: (save-state DiskII unit v2)
. support Format Track state
. save DiskLastCycle
DenibblizeTrack(): added some debug asserts and comments
Updated for VS2008/VS2013/2015/2017 projs & remove dependency on ddraw.lib for VS2013/2015
Updated disk logging:
. Moved all LOG_DISK_xx macros to new DiskLog.h (since shared by Disk.cpp and DiskFormatTrack.cpp)
. For write nibble: option to log cycle count & sync byte count
. For written track: option to log gap1/2/3 and track size
. For disk latch r/w: option to log when D5AA96 detected
Other:
. Debugger: Fix CD cmd to support absolute paths (#505)
. Debug 'G(o)' cmd now defaults to normal speed (and precise video updates)
. New debug 'GG' cmd enables full speed (and periodic video updates)
. Single-stepping (normal or full speed) now routed through ContinueExecution()
. Removed Cpu6502()/Cpu65C02() check for debug breakpoints
. Removed the (undocumented) SHIFT+F7 feature to exit debugger for 'normal speed breakpoints'
. Removed the g_bDebugNormalSpeedBreakpoints variable
. Removed NTSC_VideoVbl(), since it's not accurate during full-speed. (Nothing was using it)
. Renamed VideoGetVbl() to VideoGetVblBar()
. Correctly pass Vbl (not VblBar) to Mouse's SetVBlank()
. Specifically IBIZA.DSK demo which streams data from the disk throughout (ie. so runs at full-speed all the time)
. Update the screen every frame, but only if video memory has changed (check AZTEC.DSK loading time)
. NB. When running at full-speed, then 6502 emulation doesn't do cycle-accurate video updates
Fixed crash that could occur when switch video mode (F9) when running at full-speed
. Occured when g_nVideoClockVert was >= 192
- old 640x480 full-screen deprecated
Fixes for:
. Logo & Debug window scaled/positioned correctly
. Buttons & disk activity (on RHS) drawn in correct position
. Crosshairs for mouse (and when using mouse as joystick)drawn in correct position
. means that disk accesses (eg. loading) is much quicker
Correct naming of video modes (in UI and code)
Pixel adjust for NTSC B&W and Color video modes to align with other video modes
. final rendered window
. print-screen bmps (both sizes)
Fixed full-screen so that:
. all mode (RUNNING, DEBUG and LOGO) all occupy the same screen position
. there's no intermediate data drawn out of position when first switching to full-screen
Tested on Win7 and Win10
Fixed JoyUpdateButtonLatch() to debounce in 5mecs instead of 5secs!
SpkrToggle() was using redundant global ('cyclenum') when using the PC speaker.
. I reworked the code (as it looked out-of-date),
but probably better to just remove all the PC speaker support.
Removed complex case below for:
. VideoHasRefreshed(), 'anyupdates'
. VideoCheckPage()
Detailed notes below.
---
Video updates in ContinueExecution() loop:
'anyupdates' gets set if there were any page-flip(s) in last ~17030 cycles:
anyupdates |= VideoHasRefreshed();
ie. VideoRefreshScreen() was called outside of this loop.
If there's been a call to VideoRefreshScreen() outside of this loop,
and then the video framebuffer gets written to, ie. VideoApparentlyDirty() returns TRUE,
then don't call VideoRefreshScreen() from this loop for 3 frames.
(If a VideoRefreshScreen() is called outside of this loop then restart the 3 frame count.)
So..
if the game is flipping, the VideoApparentlyDirty() will return FALSE (since game writes to other framebuffer).
if the game is not flipping, then VideoHasRefreshed() will return FALSE (since no flips occur).
Therefore this complex case above probably only arises at a boundary eg. when the game is transitioning between these 2 modes,
and so if the emulator does the very occasional screen update in this main loop, it is of no consequence.
(I guess this extra logic was to throttle video updates on very old slow machines)
---
VideoCheckPage(BOOL bForce) was called twice in main-loop:
UnexpectedPage if g_bVideoDisplayPage2 != SW_PAGE2
Once each time through the loop (ie. every 1ms), with bForce=0
if UnexpectedPage && >500ms since last flip then VideoRefreshScreen()
Once each video frame (ie. ~17030 cycles) when not flipping, with bForce=1
if UnexpectedPage then VideoRefreshScreen()
Basically this was all about supporting FullSpeed mode, and limiting the calls to VideoRefreshScreen().
