cache

On Sunday March 08 1998 Kirill Frolov wrote to Andrey Mikheev:

KF> 2All: A similar request, at the same time I want to know what to do if
KF> the cache is 32 kilobytes? Now, instead of 48 and 128 BASIC
Here in our city, they seem to have made a similar scheme, but it’s buggy, it’s an infection.
They even support 32 kilos - you can replace 2 different pages.

KF> can be turned on. Do you need a reset in the cache (by analogy with Magic)?
What do you mean? So that the cache is cleared when you press reset?

KF> There are a lot of buttons. When should Magic in the cache be pressed -
KF> always or only in RAM? I have it whenever.
That’s right, that’s how it should be. Yes, but is it also possible to make such a feature (optional) - so that the TR-DOS address selector works in 128 ROM. Damn, I need this
for the operating system.

Bye, Kirill! With the best wishes, Megus of Brainwave.

[Team ZX-Spectrum] [Team NeOS] [Team Rock'n'Roll forever]

RP> What do you mean? So that the cache is cleared when you press reset?

No. So that it turns on after a reset. The computer has a second reset,
like a second magician.

RP> That's right, that's how it should be. Yes, but is it also possible to make such a feature (optional)
-
RP> so that the TR-DOS address selector works in 128 ROMs. Damn, I need this
RP> for the operating system.

Why??? do it through RAM. This is not available in Spectrums.

W!7H B3$7 W!5h3$, FK0.

RP>> it’s buggy, it’s an infection. There, even 32 kilos are supported - you can replace 2
RP>> different pages.
KF> ??? Replace RAM??? But my ROM is being replaced.
??? Did I write about RAM? Maybe I missed something? ROM, of course, is
meant.
KF> After ina, (#fb) the cache is turned on, and its 2 pages can
KF> be clicked through the 4th bit #7ffd.
Can you do it arbitrarily? Those. In 128 ROM we enable the cache, in 48 we do not enable it and vice versa.
Helpful. And my question, actually, was of this order - a scheme for FULL ROM replacement, i.e. Even writing to the cache in this mode is prohibited.

wrong. Well, imagine: nmi can work at any moment
(which means the cache will turn on at any time), but nmi is processed -
after the last mash. command cycle.
If the program uses a ROM, it will turn off too early -
we catch a glitch.

RP> feature (optional) -
RP> so that the TR-DOS address selector works in 128 ROM. It's necessary,
RP> damn it, I need this
RP> for the operating system.
it's possible. But how do you want to choose? Mikricom: the
ROM signal that goes to LA2 in the controller is passed through
micrik, and then sent to LA2. Here (on LA2) you hang the 1K summary. When the microphone is turned off (open circuit) - you can
apply from basic128.
I can’t give the numbering of the microcircuits for the scorpion - I haven’t seen it.
Ask the scorpion people.

Goodbye, Roman.

KF> you can click through the 4th bit #7ffd.

RP> Is it possible to click arbitrarily? Those. in 128 ROM we turn on the cache, in 48 not
RP> we turn it on and vice versa.
RP> Useful. And I have a question, actually, of this order -
RP> FULL-STAGE
RP> ROM replacement scheme, i.e. even writing to the cache in this mode
RP> is prohibited.

And then the ROM emulator circuit flew!? (I, however, never watched it)

In general, some people here do this: they install KP11 and a toggle switch. I loaded some new thread into the TR-DOS cache, clicked the toggle switch and played with this TR-DOS
(recording is prohibited). You can also replace the 128th quarter... Well, let NeOS be pushed there.

RP> Yeah, damn it, through RAM... The Spectrum already has little memory, you also need to remove the functions
RP> of reading/writing a block of sectors in RAM... No, it’s better to cut one wire
RP>, and close it back when the ROM is turned on.

FAKING SHIT! I won’t cut the wires to make the program work. And no one except
you will be, because you have ~32-64 bytes in RAM!

With best wishes, Kirill Frolov.

AK> is incorrect. Well, imagine: nmi can work at any moment
AK> (which means the cache will turn on at any time), but nmi -
AK> is processed after the last mash. command cycle.
AK> If the program uses ROM, it will turn off too early -
AK> we catch a glitch.

It is necessary to provide confirmation of a non-maskable interrupt or something
similar.
After all, even if you prohibit the inclusion of cache and nmi in ROM, then there may be a command
that sits in RAM and reads data from ROM.

AK> is possible. But how do you want to choose? Microcom: the
AK> ROM signal that goes to LA2 in the controller is passed through the
AK> microphone, and then sent to LA2. Here (on LA2) you hang
AK> summary 1K. When the microphone is turned off (open circuit) - you can
AK> contact from basic128.
AK> I can’t give the numbering of the microcircuits for the scorpion - I haven’t seen it.
AK> Ask the scorpion people.

If you do as you write, each program will write for you like this:

TURN ON THE MICRICS: 1,123,56,12.
TURN OFF MICRIKS: 3,15,98,67.

ERROR: MICRIC 12 NOT DISABLED!

And the computer will have a 1x1 meter control panel, all hung with switches...

If you really need it, you can make it programmatically accessible. Just inappropriately.
The bad emulator people want to ruin the whole Spectrum, they send spies here with such
proposals...

IT’S DIFFICULT TO FIND 32-64 BYTES IN RAMFOR A SMALL PROGRAM ???

RP>> FULL ROM replacement, i.e. even writing to the cache in this mode
RP>> is prohibited.
KF> 'ROM emulator'.
Yeah, I don't have one at the moment. Throw it into the echo please.

RP>> cut the wiring, and when the ROM is turned on, close it back.
KF> FAKING SHIT! I won’t cut the wires to make the program work. And
^^^^^^^^^^ Don’t worry, why do it right away.
KF> no one will be there except you, because you have ~32-64 bytes in RAM
KF> it’s a pity !
32-64?????????????????????? Since when did sector-by-sector read/write functions on a disk with direct operation with TR-DOS begin to take up 32-64 bytes? Now a specialist has reassembled that piece - 370 bytes! Or what do you suggest throwing into RAM?
Calling the tr-dos function? Every time you call, you will have to turn on 48 ROM,
call tr-dos, turn on 128 ROM, and only then return.
It’s going to be so slow! Although, I'll try this option on a real Spectrum
(on the emulator it doesn't matter, everything will work fine there). Or you try it. Let me know about the speed. Or, if I misunderstood you, tell me about your method.
I just haven’t been able to come up with anything simpler yet. Or offer shorter
read/write functions with conditions so that they process all errors
(no disk, disk error, read only, etc.), work with disks with arbitrary
sector size and numbersectors on the track. If you write, I’ll say a big
thank you, I’ll include it in NeOS and upon completion of the work I’ll personally send you
the complete set.