cache

On Sunday November 01 1998 Kirill Frolov wrote to Roman Petrov:

KF> It doesn’t work. The hole in the drive is too small. THIS IS A HARDWARE
KF> SUCH!
I know it. And if you don’t understand, then I’ll translate from Russian into Russian: GIVE THE SCHEME!

RP>> 32-64?????????????????????? Since when did the functions of sector-by-sector
RP>> reading/writing to disk with direct operation with TR-DOS begin to take up
RP>> 32-64 bytes? Now a specialist has reassembled that piece - 370 bytes!
RP>> Or what do you suggest throwing into RAM?
KF> You need to go to zx.spectrum with such tales.
Tell me specifically what you didn’t like?

RP>> Calling the tr-dosa function? Every time you call, you will have to turn on 48
RP>> ROM, call tr-dos, turn on 128 ROM back, and only then
RP>> return.
KF> Why are you D00Mal? That's how it should be. And there is no need to write fairy tales about bars.
These are not fairy tales, this is harsh reality. When the read/write functions
ROM are running, tr-dosa is called more than once. And with each call you have to be so perverted, but these calls are included in the main cycle of reading sectors. So
we get the brakes, since the disk will probably have time to turn before the beginning
of the next sector => brakes 16 times!

KF> That’s the trouble. You don't even know where or what will happen. You are 100-200
I know what will happen and where. It meant that onI don't care about the speed of the emula, it's already fast. I'm worried about working on a real spectrum.
KF> I feel sorry for the clock cycles for each call to read _several_ sectors?
see above.

KF> If you get very perverted, you can even do it purely in ROMs in tr-dos
HOW??? Tell me, I'll just say thank you.
KF> move on. Only probably
KF> may work on some computers, but
KF> will not.
Oh! This is the most important point. So...

KF> Interesting question - is it possible to put at least a few bytes into TR-DOS from neos
KF>? This will make it a little easier.
It’s possible, but what’s the point? It’s not just a few bytes that will need to be thrown there, but these functions will need to be placed there. And this is ~350 bytes. Although if you put so much
in ROM, then you can.

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

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

RP> These are not fairy tales, this is harsh reality. When the read/write functions
RP> are running, the TR-DOS ROM is called more than once. And with each call you have to be so
RP> perverted, but these calls are included in the main cycle of reading sectors. So
RP> we get brakes, since the disk will probably have time to turn before the beginning
RP> of the next sector => brakes 16 times!

One byte requires 112 clock cycles. There are ~50 bytes between sectors.
In total, you have 5600 clock cycles. But if there are not 50 bytes between sectors, but 10
for example - only 1120 clock cycles. Can't make it in time?

On Friday December 04 1998 Kirill Frolov wrote to Roman Petrov:

KF> 10 for example - only 1120 bars. Can't make it in time?
We've already made it.

KF> I'm being lame. You also have one sector at a time.
Can you do several at a time? And how?

RP>> bytes. Although if you put so much into the ROM, then it’s possible.
KF> It’s possible, but not on Scorpio. Everything is busy there. Although the scorpion has
KF> everything you need is there.
Everything is already working, you don’t need to cut anything. You should have started right away with the bars.