Controlling the emulator from the zx program
ZXNet echo conference «zxnet.pc»
Hello, All
Behind this mysterious title lies a request to respected emulator authors
make it possible to slightly control the emulator from the emulated program.
Specifically interested in the ability to turn on/off maximum speed
(it would be convenient to correct it to enable it during compilation), modes
screen (in the same setting enable/disable Anti64 mode).
To avoid false positives, you can enter a disabling setting in the config
this feature.
What will the management itself look like? For example like this:
LD HL,#0102
LD HL,(#1234)
Where in H is the number of the operation, in L is the value of this operation, and according to codes #21 xx yy 21 34
12 the emulator determines that it is knocking on its door.
You can enable and read the current settings, for this in H or L
set the 7th bit.
Of course, there are a lot of implementation options, I’ll adjust the design to suit any one and post it for
everyone.
p.s. As far as I remember in the US there was once something similar...
Hello Specter
why bother with emulator bullshit if KAY has been around for a long time (the turbo hangs on the port)
Hello Sinus
There is also Scorpio, where the turbo is turned on and off by software, but
Emulators don't support this anyway.
In general, the question was broader, not because of one turbo “to block out emulator bullshit.”
Read the first message carefully.
Hello Specter
If I do, I don’t like this interface. Why slow down even one
command (LD HL,(xxxx)), adding a check of the address there or some other conditions,
when ZX has a bunch of unused commands. How, for example, is SLT Trap made for
loading levels: ED FB.
In the same way, it is better to do ED FF - check for support for a set of functions,
changes the value of A. That is if nothing is done and nothing is maintained, then
This command doesn't change anything. If there is support for extensions (control
emulator), then XOR A: DB $ED, $FF - and in A - a set of up to 8 flags. Let's say
bit 0 - speed control support. And then LD HL,xxxx : ED FE - changes
speed. The xx values are for example:
0 - pause, 10 - 10%, 50 - 50%, 100 - 100%, 255 - 255%, 65000 - 65000%, 65535 -
full throttle
Hello, Vladimir Kladov
All this is bullshit, because it is necessary to unify the interface for all emulators, which
hardly possible due to differences in features that emulator authors will want to include
via API (EDFF, by the way, was previously in unreal)
Hello SMT
SMT> all this is bullshit, because it is necessary to unify the interface for everyone
SMT> emulators, which is hardly possible due to differences in features that
SMT> emulator authors will want to enable it via the API (EDFF, by the way, was previously available
SMT> unreal)
Absolutely true. I think you need something like this:
Functions 0..#3f - common to all emulators, such as speed control, timing,
memory models, etc
Functions #40..#4f - US specific features
Functions #50..#5f - specific features of EMUZWIN
etc.
To have something to start with, let's make function N0, speed control
emulator, where 0..250 is the Z80 speed as a percentage of the “standard” one, 255 is
maximum possible.
Hello Specter
Let's not touch ED 00 - ED 3F, but (or vice versa ED C0 - ED FF, but everything is already there
equals SLT Trap, it seems to be a de facto standard)? Moreover, in some
in the ancient emulsion ED 00 - ED 3F they were full to capacity. In general, you need to use
less less "stairways". The function number can be transferred, for example, to A, additional.
parameters if necessary in DC, DE, HL, also at the output if you need to receive something.
Then, in this sense, ED FF is a check for the presence of a feature, and ED FE is a function (number in A
- you can keep the same numbers, and there will still be a reserve). Will that work?
Then A=0 is speed control, A=1 is “reading” speed in HL (one could
do it at once, in HL - return the previous one, but if you suddenly don’t do this
is it necessary, it’s better not to pull the emulator, or then you need to highlight some value -
"leave speed" unchanged).
As for 254 maximum - it won't be enough. I just drove RZXs today, I have
Almost all are running at 1500%, some are running at 3000%, and some are running at 5000%. Why not
and then don’t ask in HL. In 10 years, 65000% will be normal.
Hello, Vladimir Kladov
I think what is missing is the definition of the emulsion type. Let in the same ED FF in HL
returns two letters: H='U
Hello SMT
These ED nn commands have given in to you. I like the sequence better
consists of documented commands (so that problems with assemblers and debuggers
was not), but with near-zero probability it occurs in real code.
I don’t see the point in detecting an emulator at all. And what should I do?
program if it wants to enable pentagon mode, but it fails?
Nothing!
Therefore, it is enough to turn on something, and if it does not turn on, the user is to blame,
which has disabled it in the emulator or is using the wrong version. I somehow
I doubt that even reading the current values will be needed.
Hello, Vladimir Kladov
Finally!!! It's high time!!!
Hello Specter
2Smt: Actually, EDFF is pure NOP. So you can’t double-check it 100 times.
necessary. If this ladder changed anything, then this is the ladder :)
2Spectre: if you do something, then let it be on separate teams, and it won’t affect
not a fraction of a microsecond on the speed of execution of other commands. And we earlier
they didn't do anything and everything was fine. Let it be now. You can always
option/configuration to disable/disable.
Hello SMT
if such a miracle comes across, it won’t take long to chop it off - in the options (especially if the options
are cut off by the config on the fly). This is when ED FF herself is in this Xorka
is being performed. Why would anyone in their right mind do it? Just to make things worse
isn't it...
Hello, Vladimir Kladov
you need to use a minimum of traps (that is, one, for example, #EDFF) and
recheck 100 times (for example, I used to have a 6-byte signature
'unreal' in registers bc,de,hl) for some undocumented super-test
team did not accidentally run into a trap. on the other hand, coders will say that
It’s inconvenient to use very long API call procedures. now I
I thought about this option - make 1 function that checks a bunch of conditions
(to avoid accidental triggering), returns the type and version of the emulator and
sets a flag that allows EDFF to be accessed with lightweight
parameters (without double-checking). checkbox cleared on reset or boot
new snapshot
Spe> Functions #40..#4f - US specific features
Spe> Functions #50..#5f - specific features of EMUZWIN
one person wants an auto-enable feature in his editor
text64, and in the registers indicate the address and other parameters of the font. and then this one
The function will appear in another emulator, but with a different enable code. here it will be
It's a joy to detect the emulator type and enable the mode in different ways ;-)
Hello, Vladimir Kladov
Vla> if such a miracle comes across, it won’t take long to chop it off - in the options (especially if
Vla> options are cut off by the config on the fly). This is when ED FF herself is in this
Vla> xorke is executed
but try to guess what exactly needs to be cut off. especially when this #EDFF was
It was made six months ago and there were no such problems before
Vla> Why would anyone in their right mind do it?
Why in the right mind would you put greetings in the bootloaders between the XORs? why on
these greetings transfer control, fall into ROM-48 by mistake or
calculator and run the following raskorka as a message handler rst#08?
go understand these spektrumists. and foreign ones were no different ;-)
Hello, Vladimir Kladov
Vla> Actually, EDFF is pure NOP. So 100 times
Vla> There is no need to double-check. If this ladder changed anything, then this ladder
necessary. this #EDFF may be among the garbage inside the xorca. if it affects
registers, it won’t get messed up
Hello Specter
line is fine. or maybe, in fact, make it so that the emulated
the machine had no way of knowing that TRAP had been activated? all the necessary functions (even
hypothetical) do not require data return: speed, post-processing modes
picture/sound, stereo layout for AY
Hello, Vladimir Kladov
Vla> In theory, a lot of things can be done through the ladders. Ask for a cut
Vla> equipment. Enable the desired model
the model is incorrect. for good reason, after changing the model (i.e. ROM and ports
memory) needs reset. the trap that calls reset is useless. unless it's a model,
and temporary buildings (what for are coders)
Vla> Yes, at least implement a 3D graphics engine
and there will be exactly 1 game for this engine, written by the author of the emulator ;-)
Vla> the current speed can be done to find out. For example -
Vla> found out, saved, put up a new (large) one, did some work, turned on the old one
I very much doubt that they will use values other than normal/max.
There is no point in speeding up an interactive program or demo a little (if there is one, the user himself
This will be done in the emulator settings. but there is no need to force it), it will be in real money
slow down It’s easier then to write a full-fledged pisish version
Spe> it’s another matter to save IX, call the ladder, calculate the CRC received
Spe> lines, compare with the required one, restore IX. This is already byte 50
Spe> will pull
a little smaller -
┌─- CODE ───
call nn
db 'emulator traps'
nn: pop ix
dw #FFED
└── CODE ───
I knew that the coders would rush to count the bytes ;)
Hello SMT
SMT> line - normal.
The line is wild. It’s one thing to cram 5 bytes into the program code with STS,
turning on the turbo, it’s a completely different matter to save IX, call the ladder, calculate the CRC
the resulting string, compare with the required one, restore IX. This is already 50 bytes
will pull. This means you won’t remember them from memory, and you’ll have to look for the place.
Moreover, I repeat once again: there is no need for ED and other prefixes if I use this code
I want to disassemble, will the disassembler understand it? Plus in different assemblers
different designations for undocumented commands.
SMT> and can, indeed, make sure that the emulated machine does not
SMT> had the opportunity to know that TRAP was triggered? all the necessary functions (even
SMT> hypothetical) do not require data return: speed, modes
SMT> image/sound post-processing, stereo layout for AY
Here, we are getting closer and closer to Spectrum programmers. :)
Hello SMT
Well, why, you can speed up the current one to find out. For example -
I found out, saved it, put up a new (larger) one, did some work, turned on the old one. Other
The thing is that for this feature you can add another option - like do not react. A
cool for example: an audio player that reads the emulation speed, adjusts
audio speed so that it goes at normal speed hehe.
In theory, a lot of things can be done through the ladders. Ask to cut the equipment.
Enable the desired model. Yes, at least implement a 3D graphics engine. :)
Hello SMT
I'll set it to off by default. If he turns it on and doesn’t know why he did it,
This is a user problem. But it's unlikely to be a serious problem.
Well. You can do this thing: let IX point to when calling these features
line in memory: 'Emulator traps'. Is that enough?
Hello SMT
I’ll say right away: whether it’s good or bad, I did it on purpose so that
the model could be changed on the fly and, if possible, without having to do a reset. By
at least if ishra is already working - what for does it need ROM. And for RAM, many RAMs and so on
converge. “Changing the model” in this case is simply a transition to another decryption
ports, different multicolor, timing, sound. Overall it helped me a lot
debugging multicolor, when I still didn’t know what color this or that deme was for
sharpened. True, in this case, I took care of the timing, I have models as such
It didn’t exist then. Let the user do the reset if he wants.
Hello SMT
Well, why doesn’t it work like it did in Unreal? That's right: you can diss it, etc. too
I doubt the need for normal and max speeds... well, maybe 200% turbo???
I will also support independence from the emulator. just functions. why do the authors of emuline
make a standard, so to speak. There are many function numbers in 1 byte, here they are
divide, support similar ones, divide different ones and all things...
in short, Specter - let's see how it will be exactly... I'm FOR it with all my limbs!
oft: SPECTER!!! - how is QC 311B with sprView??? :confused:
Hello, Vladimir Kladov
and I don’t give a damn whether a specific disassembler will understand or not understand this
instructions (mine will understand, he's smart :)). It is more important to me that with the prefix ED - I
I just go through the table of jumps in the emulator, if it’s a trap, then it starts
there is something to do there, and if it’s not Trapovsky, then there’s no need to do anything. From what
By the way, I have to add PC clock cycles to analyze additional conditions in the usual
team. She must do what she is supposed to do - for the minimum amount
beats, and run further. It just seems like the emulator has a lot of time to spend.
frame emulation (70,000 clock cycles - big deal). And if any
graphics mode like HQM or Blur motion? What if the emulator is running on a slow PC?
- up to 200 MHz? I only agree to ED prefixes (other unused commands
would also fit, but there are not many of them equivalent to NOP). Actually, I already
done, all that remains is to add that along IX lies “that line”. But the code is true
why do you start counting bytes? But it’s already 99.999999% that no Xorks
will break. I don't understand about CRC. What other CRC? Here is the interface I
implemented, verbatim:
PUSH IX:LD A,0:LD IX,strap:DEFB $ED,$FF:POP IX
RRCA:JR NC,@1;if there is no transition, then there is support for speed-control, HL='EW'
PUSH IX:LD A,1:LD IX,strap:DEFB $ED,$FE:POP IX
;HL=current speed
HL=new_speed
PUSH IX:LD A,0:LD IX,strap:DEFB $ED,$FE:POP IX;new speed set
strap: DEFB 'Emulator traps'; the final byte or flag is not needed in the line
(No, well, you can optimize it, I wrote it in a simple way for clarity. Already
it is clear that this is the same code, and by self-modification it is possible
reuse so as not to duplicate 3 times).
Another potential possibility: make a plugin that will
treat additional ladders. Pass block addresses to this plugin
processor memory and registers, and let him do what he wants with them. Possible
application: drain-stub for future functionality, which is simpler
implement at first while the program is debugged not by code in the spec, but by code in
PC. Or a trap-stub for accessing some external equipment that is still
not ready for real connection to the spec (but there is a chance to emulate it on
PC or just contact him at the PC), but I want to test it and make software
necessary (and possible) at the same time or “before”.
Hello Specter
Spe> In short, almost the entire US config needs to be rewritten here
What are you doing, people! what except max. speed will actually be used in
programs? yeah, you know better than the user what he simply needs
FullScr, and that with vsync it has enough performance and the video card does not glitch
Spe> Just one question: is it necessary to check for the presence of ladders
no one is forcing you. turn on the speed immediately, without checking what happened
(and if it doesn’t work out, who should I complain to?) unless I implement the idea with a flag,
when the #EDFF trap requires a line in IX and returns the emulator version, checks the box. a
subsequent calls to #EDFE no longer require the use of IX (if the flag
installed)
Hello, Vladimir Kladov
Vla> Here is the interface that I implemented, verbatim:
Vla>
Vla> PUSH IX:LD A,0:LD IX,strap:DEFB $ED,$FF:POP IX
Vla> RRCA:JR NC,@1;if there is no transition, then there is support for speed-control,
Vla> HL='EW'
Vla>
Vla> PUSH IX:LD A,1:LD IX,strap:DEFB $ED,$FE:POP IX
Vla> ;HL=current speed
Vla>
Vla> HL=new_speed
Vla> PUSH IX:LD A,0:LD IX,strap:DEFB $ED,$FE:POP IX
Vla> ;new speed set
Vla>
Vla> strap: DEFB 'Emulator traps'; final byte or flag is not needed in
Vla> line
Well, if it’s more convenient for the authors of emulators, then do it that way. I, as I wrote in the first
message, I will support any standard. Just one question: is it necessary?
check for the presence of ladders (it would be more convenient for me not to) before
starting to work with them?
All that remains is to decide on the necessary operations and their numbers. My option is this:
Memory model: PENTAGON, SCORPION, KAY, PROFI, ATM450, ATM710
Memory size: 0 = 0Kb, 255 = 4096Kb (that is, each one corresponds to 16Kb)
CMOS type: none, DALLAS, 512Bu1CMOS size: 0..32Kb (in 1Kb increments)
EFF7mask
SMUC features
RESET to: BASIC,DOS,MENU,SYS,CACHE,...
Keyboard Layout
Mouse type: NONE/KEMPSTON/AY/emulate kempston joystick on mouse
Mouse Scale
Swap mouse buttons
Frame ; t-states in frame
Paper ; t-states before paper
Line ; t-states in line
int ; int frequency
intlen ; int length in t-states
EvenM1 ; align M1 cycle to even T-state (for scorpion)
4TBorder ; update border every 4T states (for scorpion)
FloatBus ; unstable data bus, unstable data bus in TR-DOS mode
AtmPalette
VSync
Full Screen
Border size
[...]
In short, almost the entire US config needs to be rewritten here. Just throw it away
the ability to change the screen frequency, disk and other system features, that is,
that only the user himself should change.
Hello SMT
Is it really usable? Anti64 is a must!
Hello Specter
Right now. Whatever we have enough of, we will support. What is anti64? What is 64 against?
Video effects and full screen inclusion are the prerogative of the user after all, and not
Spekovskaya program. Spec's programs are actually necessary for real life (hm, sort of)
be done, not for the emulsion. Emulsion, as I understand it, is for people who cannot tolerate
the smell of burnt rosin, and vice versa for hardware developers to get their ideas
implement faster and more accurately.
Hello, Witch Doctor
So I didn’t understand. Let it turn on the text video mode under ATM and that’s it. Or
Why am I not getting it?
Hello, Vladimir Kladov
What are you doing, Volodya???
Many of us need this for 1 only ALASMA with a 64 character font. That's all. So that we can
I turned this thing on myself at startup, and turned it off during vacation. This is an idiot's dream :)
Specter, am I speaking correctly???. Same for max. speed when compiling the same
alasma. Of course, it’s better in ZXemuASMe - not everyone wants something... But
the rest of the options... IMHO optional, probably...
Hello, Witch Doctor
Ah, well, that’s for unreal. I don't see such a filter. Author alasma
alive and well, but doesn’t want to convert the ASAM to ATM? This is not good. There are more
Pentagon with 512x192, also high resolution. (There are also 384x304, a little
there will be less, right now I’m downloading ZX-Guide, like someone somewhere gave it to someone in
The quality of the answer to the question about this regime is a hint. Somehow this doesn’t suit me
I like it. Electronic media should be more accessible. Like clicked on Google -
and uploaded the text. Is it difficult for the authors of their “media” to post texts? No, let him
suffer, load it into the emulator and read it in the original, in a poor square font,
poor keyboard control... And most importantly, Google does not index the content
archives, and even in the form of scl-trd files!)
Hello, Vladimir Kladov
So under ATM there is no asma - Alasma :( And who will rewrite it? AlCo? I don’t think so...
I'm trying to clarify:
This is what just alasm looks like... and alasm with an anti64 filter is completely different
thing :) is much more readable.
File: al64.png http://zx.pk.ru/attachment.php?attachmentid=2807
File: al1.png http://zx.pk.ru/attachment.php?attachmentid=2806
Hello, Vladimir Kladov
SMT!!!! Why are you silent when a “colleague” asks? I don't think anyone is here
Volodya couldn’t explain better than you what’s what... especially since he clearly has a reason
interested :), i.e. Apparently it will be implemented. Help progress!
Hello, Witch Doctor
I have no time for that. such a video card burned out :-( and other things piled up :-(
Hello SMT
Actually, I agree that “the entire US config” is not needed. ;) Personally, I would
enough:
1. normal speed
2.max speed
3.anti64
4. default video mode (that is, which is specified in the config, for some
double, I have quadro)
5.Memory model
6.Memory size
Maybe something else will be needed later in the process of implementation.
I'm waiting for support from the authors of emulators...
Hello, Vladimir Kladov
Vla> The author Alasma is alive and well,
This is not the author, but the 5th “continuator”. Global changes Alasma AlCo not
will pull (I said it myself once).
Vla> and doesn’t want to convert the ASM to ATM? This is not good. There is also the Pentagon
Vla> with 512x192, also high resolution.
The conversion of Alasm for ATM is approximately comparable to the conversion of EmuZWin for
MS-DOS... Or maybe an order of magnitude heavier.
Under the Pentagon 512 it’s easier, but is this effort worth the final result?
Hello Specter
Indeed, remaking the alasma is a very grim task. Thank God that although
I would like to slowly improvise what is...
In addition, AlCo pursues the line of maximum compatibility and “maximum
minimalism" i.e. should go under 128k and take up a MINIMUM of memory. Offer
It was not possible to expand the 6x6 font into a separate page to speed it up precisely for this reason
considerations... Hmm... But this has its own logic, of course... (although for myself I
I'll expand the font anyway :) )
Hello Specter
Well, there is also conditional compilation. The MINIMAL_VERSION symbol, and with it
maximum minimalism, SUPPORT_EXTVIDEO symbol, and additional
video modes. As a programmer who works a lot with Asm, this is not a problem for me at all.
seems complicated. I also make the emulsion - the main part - on Asma. Exactly
Thanks to conditional compilation and macros, I managed to solve an unsolvable problem:
combine speed and support for such a huge number of combinations of all kinds
modes. As a result, I don’t need to write out the code separately for mode 48 (so that
it was optimal, and the memory was addressed simply as 64K of memory in a row, at least
least for reading), separately - at 128K with multicolor, and separately - without
multicolor (I mean branded multicolor, domestic
multicolor does not require such optimizations. But before everything was exactly
separately, even for the 256 color mode there was a separate code. (I'm just tired of them
synchronize with each change, it turned out to be easier to make a procedure for
compiling PC-Asm into memory on the fly).
Or the output of the current page of text is so difficult to separate into a separate one
procedure so that you can use advanced modes? The plan is to replace
this procedure with at least conditional compilation and get the ATM version of the same
or Pentagon 512x192 mono at least.
Hello, Vladimir Kladov
Something happened to the connection yesterday - the message disappeared. I repeat:
Actually, I agree that “the entire US config” is not needed. Personally I would
enough:
1. normal speed
2.max speed
3.anti64
4. default video mode (that is, which is specified in the config, for some
double, I have quadro)
5.Memory model
6.Memory size
Maybe something else will be needed later in the process of implementation.
I'm waiting for support from the authors of emulators...
I already supported 1.2.
3.4 is for SMT.
5.6 - this option:
ED FF returns set bit 2 to A (not 1 and not 0, bit 1 will be left by SMT). And
then:
ED FC register H, L - two letters of the model, DE - memory in kilobytes. At the entrance -
sets the model/memory, if the corresponding register is not 0000, to the output
- returns the model and memory as they were before the call. Still convenient on the way out
return the model number. You can use your number at the entrance, but it seems to me
letters are more convenient:
S4 = Spectrum48
S1 = Spectrum128
S+ = Spectrum128+
S2 = Spectrum128+2
SA = Spectrum128+2A
SB = Spectrum128+2B
S3 = Spectrum+3
SE = SpectrumSE
T4 = Timex2048
T6 = Timex2068
SC = Sam Coupe
Di = Didaktik
Sc = Scorpio
Pe = Pentagon
Pr = Profi
A1 = ATM-1
A2 = ATM-2
A+ = ATM-2+
A3 = ATM-3
Sp = Sprinter
KA = KAY
Kv = Kvorum
Ho = Hobbit
PL = PLM Automatica
Le = Leningrad
Mo = Moscow
Hello, Vladimir Kladov
this is all clear, but there is another more terrible problem - places in
There are almost 40 bytes left on the Alasma page (I don’t remember exactly right now, but it’s a fact)
- i.e. as if it were more cultural... there is no place to fart there anymore. By the way, this would be in the department.
branch and milk well on the answers of AlCo himself...
Hello, Witch Doctor...!
I believe that the emulator should be controlled from the program by writing to
ports.
This is the only way to clearly track access to resources as real.
Spectrum and virtual. Let's select a range of ports that were not initially
existing in a real computer, which we will select to control the emulator.
With the best - Wlodek # wlblack()newmail.ru # http://wlodeks.narod.ru
[ZX] [500:95/462@ZXNet] [2:5016 Forever] [Golyanovo] [Old Russians]
Hello, Vladimir Kladov
Vla> It is not clear what default is. Is it difficult to save HL, DE? Well, excuse me
Vla> didn't know. 8-]
Probably haven’t written on a real Spectrum for a very long time? ;) I alternate:
I’ve been writing on PC for several months, then on Spectrum. So when switching to Spectrum
some things are unpleasantly surprising. But what to do... :(
Vla> (Why restore them at all?)
Culture. We played with the settings, and when exiting the program, we returned those that the user
exposed.