add some thoughts about clock

This commit is contained in:
Brad Grantham
2018-07-30 11:18:57 -07:00
parent 3bf5f5ed6a
commit 064a9704d6

45
notes
View File

@@ -1,10 +1,53 @@
per-scanline modes
Actual Apple II TEXT/GR and HGR video address patterns are in Understanding the Apple II
Not for DHGR or 80-col but they're probably just same address twice alternating banks
tracking clock and reporting video memory contents
clock
actual fastest clock is 14.31818MHz
one tick of ~1MHz clock maps to one tick of address counter
so is 7 HGR bits or one GR/TEXT horizontal cell
unsigned long would overflow after 20 minutes
But could modulo byteclock by 14318180 at interface update
time and not overflow
as long as it doesn't take longer than 20 minutes between calls to APPLE2Einterface::iterate
than 20 minutes...)
Want current CPU clock count after "cycle()" - map that to 14MHz clocks, can map that to actual wall clock
"most 6502 cycles are 978 nS long [14 14mhz clocks], but every 65th cycle is 1117 nS long [16 14mhz clocks]."
So wall clock time is (((clockcount / 65) * 912 + clockcount * 16) / 14318180 seconds) but only starting from 0
so need seconds from t_phase to t_clocks in clock counts where t_phase != 0
initially t_phase = 0
t_elapsed_14mhz = t_clocks * 14 + (t_clocks + t_phase) / 65 * 16
t = t + t_elapsed_14mhz / 14318180
t_phase = (t_phase + t_clocks) % 65
So clk would remain machine clock cycles, but actual wall clock time won't be known until calculated
Could increment wall clock at clk++ depending on phase, but then probably accumulate error because 14MHz clock rate is so high
Current CPU clk (in 1MHz CPU cycles) :
is used to determine to which audio sample (at libAO sample rate) in the audio waveform we should fill
OK - should work with higher fixed rate
is used to mark after how many cycles the open apple key should be raised
checked against clk in
HM - need to convert to wall clock?
doesn't need to be that
address generator
honor softswitches
convert byteclock to address
read empty memory address on bus no longer fails but instead
generate address, get byte from RAM and return it
recording softswitch mode changes
store vector of switch or mode change and byteclock
instead of setting before APPLE2Einterface::iterate, pass into iterate
honoring softswitch change in APPLE2Einterface
initialize scanline mode array to text
receive vector of tuple of byteclock and modes
fill scanline mode array from passed in vector
draw framebuffer by scanline
final release:
help screen
gamepad calibration screen (for desktop GLFW anyway)
need to handle asynchronous gamepad arrival (a la HTML5)
screenshots for README.md
link to emscripten version
Linux build
emscripten:
safari misses 'V' for some reason