bars

Hello, Kirill! KF> Can someone please explain how long it takes KF> in scorpion (yellow) to execute commands like LDI/DJNZ/LD A,(HL)...? KF> The fact is that they contain an odd number of measures, and the scorpion kF> reading COPA from RAM can insert /WAIT. We asked - we answer (ts)... :) By car Leningrad, Composite, Scorpio, Kau When fetching a command from RAM, one clock cycle is inserted expectations, but not always, but then, When the previous the command had an odd number of clock cycles. For simplicity, we can assume that the expectation occurs not in the current command, but in the previous command (odd). So, if a team takes an odd number of number of measures - an extra measure is inserted and, thus time is supplemented to more even value. LD A,(HL) 7 > 8 LD (0),A 13 > 14 DJNZ$13 > 14 etc. (you're wrong about LDI - 16 cycles, even, 16 will be) Thus, when counting clock cycles for any programs you are next to any odd team you add one more bar to the passport bar. The same is true for turbo cars. in non-turbo mode. KF> Another question: the clock frequency of the processor is 3.5 MHz, interrupts KF> 50Hz, and the clock cycles per interrupt are not 70000, but about 68000, KF> and this is in ROMe and RAMe (the results are the same). What's the matter? How did you count the bars in the interrupt? I think it should be like this - install IM2 and launch program like M0 INC DE:JP M0; 16 strokes/revolution And in interrupts we catch the current DE, write it down put it somewhere, reset it and run it again. We repeat this 129 times, unsubscribing each time the resulting DE (the very first value of the selection) collapses). We get 128 values. Each must be multiplied by 16 to get the result add the number of clock cycles it takes (including odd commands!) program on IM2 (writing DE to a list, working with a stack, looping 129 times, etc.), and add 19 (20) bars - correction for the INTACK cycle. Of 128 adjusted given values, calculate the average - this is number of clock cycles per frame. This is just a template, details may vary. only at this time the processor should execute how many even teams (in my example 6+10). And if there are odd ones, you will get the wrong one result, since the execution time is odd commands is different from what is documented. The resulting accuracy is 16 cycles. I was able to measure with an accuracy of up to 4x cycles - I filled the memory with NOPs, launched, and during interruptions I looked at what NOP The program has arrived. It is theoretically possible to measure accurately to the beat - but why? For example, I know that in my Pentagon 320 lines/frame, 224 clock cycles per line, without tor-mosa - it turns out 71680 clock cycles (frequency frames 48.828 Hz). On Skorpa there are 312 lines, 224 measures per line, we get 69888 clock cycles (frame rate 50.08 Hz). One more moment. In all Scorpi/Leningrad/ In composites the number of lines is always 312, but the number of bars in a line depends on the curvature pens that collected them, Maybe as many as 224, more or less... And also, some smart people with the aim of raising compatibility with the Pentagon (by the way, it has WAIT never occurs) they rape their Scorpios/Leningrads, increasing the number bars per line (improvement in the IE7 area). This solution, accordingly, reduces line and frame frequencies, and cannot be considered that the frame rate is 50 Hz - it works fewer, respectively, more clock cycles in Int. Measure the number of clock cycles per line is possible only by writing a program vertical stripes on the curb, and, at random, achieving its verticality, measure the amount cycles in the output program (taking into account odd teams!). would measure the number of cycles taking into account the brake on odd teams, no one bothered... That's how things are. All the best, Dmitry AKA -=LD=-/X-TRADE GROUP -+- ZXASM 3.0

I suddenly see on 01/12/98, at 01:43 someone Dmitry Lomov wrote to Paul Falcon about the bars: Hi Dmitry! PF>> for example, I’m not familiar with the theory of this phenomenon, PF>> but I know on what machines this happens. PF>> the person asked - I answered. DL>BEST! DL> question: DL> - How many ticks are there on Scorpio? DL> response: DL> - Yes, and also at the Pentagon! ^^^^^^^^ Why don’t you notice anything? I didn’t write _about_the Penatagon_! DL> This is it - “anyhow”... DL> If you don’t know, don’t answer. If you noticed, I answered wrong. the man asked when an extra beat is added on commands that an odd number of clock cycles are executed. I replied that this always happens on scorpions, kayas, composites and on the crooked Pentagon! I can repeat it again so that you I understand, at the _crooked_Pentagon_. because Hacker's zone 128 has somehow been converted into a Pentagon. subtleties I don't know about this change. PF>> and why did you suddenly snap at me like that? here's your _it_ PF>> the letter is both off-topic and offensive. DL> First I answered the question. And then he turned DL> your attention to soap bubbles. “Falling off” a cliff... DL> Shall we go to the soap? :) you can go anywhere, I just don’t want to swear because you want it. you're tempted to quarrel with the Mokvichs so swear, just write to them that I have nothing to do with I don’t have anything to say to you and I don’t want to have it, otherwise they write to me 3 times. If you write such a blizzard in the echo again, I’ll cut you off for a week. If you really need it, you can call, we’ll solve everything over the phone. ▌▌║▌█▐│▌▌▐▐ WiTh The BeST wIsheS fROM CREATOR product ▌812/03.00▐ Powered by Cyrix 6x86L 200+/16DIMM/S3TrV2 2MB/1.6GB -+- SMM version 1.05

Hello, Kirill! [annihilated] DL>> The above is also true for turbo cars DL>> in non-turbo mode. KF> Or maybe vice versa - for turbo cars in _turbo_ mode? No. Exactly as I said. In turbo mode what was said is doubly true :) *skip* KF> How is this done? If quartz is 14 it should be 224... KF> (and if you didn’t configure it with the D4=IE7 counter...) The number of clock cycles in a line depends on the quartz frequency doesn't depend. It depends on the division factor counter, i.e. on the number that is loaded into it. Different boards have different numbers. (meaning old Zones). In Scorpa, a lot depends on the IE7 series and from the RC circuit in the meter loading wire. *skip* KF> What should the even/odd ratio be? KF> Imho this is a very difficult question... You need to test this way: first check for even teams, it turns out? clock cycles/INT. Then check on odd ones. If it turns out otherwise, it means that the system there is a characteristic Zonov brake... KF> P.S. Well, the scorpion slows down like a skateboard without rollers... Mine doesn't slow down after I leave it The Pentagon did. I had to work hard... All the best, Dmitry AKA -=LD=-/X-TRADE GROUP -+- ZXASM 3.0

Hello, Kirill! KF> T.e. Turbo cars in non-turbo mode slow down, but in turbo mode they don’t brake? Some non-turbo cars slow down (except ATM and Pentagon, IMHO). All turbocharged cars in Turbo slow down even more, but the gain still there is. The fastest car, IMHO, KAY-TURBO. There's a turbo mode, the number of clock cycles in the command will be rounded up to a multiple four: 4 > 4 6 > 8 7 > 8 8 > 8 9 > 12 13 > 16 ... But if we take into account that most of the commands are 4 and 7 clock cycles, we get almost double productivity. *skip* KF> Why doesn’t the Pentagon slow down? (I’ve never seen his diagram!) Due to the peculiarities of the circuitry. Cars slow down because: The memory clock frequency is equal to the processor frequency. From the same memory, information is displayed on the screen, in this case, every second memory cycle is occupied. Memory is divided: one memory cycle to the processor, another video controller. The command reading cycle takes one and a half cycles. Conclusion - sometimes it can happen that when the processor wants a command, memory is occupied by displaying it on the screen. The processor is waiting. The turbo increases the processor frequency, but memory performance will remain the same.Conclusion - sometimes the processor will have to wait in the turbo two bars. In the Pentagon, division of cycles between the processor and the video is made randomly, from every two The first memory cycle goes to the video controller, second to the processor, but if the processor wants memory is on the first clock cycle, it works first, and then behind that is the video controller. In Zonov machines, the cycles are fixed - The first is required for the processor, the second is video. That's it... All the best, Dmitry AKA -=LD=-/X-TRADE GROUP -+- ZXASM 3.0