HDD speed

Hello All! Just recently I wrote here about a screw driver for scorpion, using smyk directly. The driver, weighing exactly one kilobyte, supports 8 army If you finish it a little, it will be able to work with two screws. There is only one condition: bit 2 in port #1ffd must include DOS ports to direct access from RAM (as in GMX). It is done simply: the bit is taken from TM9, inverted, and pez LI1 clings somewhere. I didn’t find a diagram, I’ll explain it with my fingers. We need to find the se- lecturer windows #3D??. In my opinion there is an 8I-HE element, 1533LA2 with RC chain at the exit for catching needles. LA2 ┌──────┐ ─┤ & │ ─┬─ ─┤ │ │ ─┤ │ === TM2 ─┤ │ │ cut ┌───┬───┐ DOS ─┤ o────████──┴──┬───── X ─┤ │ │ │ ├───│ │ ─┤ │ │ │ ──┤ D │ │ ─┤ │ │ │ ──/ C │ │ /DOS └──────┘ ┌────┐ │ ┌──────┐│ ├───┤ o── │1 │ └─┤ & ├┘ ──o R │ │ ────┤ o──────┤ │ └───┴───┘ #1ffd bit 2 │ │ └──────┘ └────┘ LI1 LH1 #1ffd itself is some kind of TM9 near the slots. I made the inverter from the free element under AY, LA3 seems to be there.The meaning is simpler than a steamed turnip, the above-mentioned bit now includes same as CALL #3D??. Now a little history. === Cut === Let your enemies, Aleksandr, die without sons! 13-07-98 at 19:05:18 Aleksandr Majorov ═> Oleg Grigoriev: OG>> Nope. Reading from the screw via #3d13 is still faster OG>>readings OG>> disk via #3d2f. I won’t tell you the numbers - I haven’t checked them; what the hell OG>> it's for me OG>> it is necessary - through #3d13 something. :) But purely visual OG>> faster. OG>> A little bit.:) AM> ??? AM> strange, I think Kir told me that according to the test results, AM> which he threw LD fast-loader slightly ahead of the screw! AM> Although I could be wrong - it was a long time ago... Oh... In short, the confusion in this matter bothered me a little... tala, and I spent 5 minutes with a stopwatch... And the writings are half- It'll probably be an hour... So... Reading 640 kilos from a floppy disk using turbo-load pa takes 32 seconds regardless of the address it goes to reading. Screw. I read 640 kilos through #3d13 in 3 different ways - mi: 1. Reading in pieces of 16 kilos per jar starting from #c000.2. Reading 16 kilograms into memory starting from #8000. 3. Reading in chunks of 32 kilograms into memory starting from #8000. The reading procedure is given in listing1. Naturally everyone times the register values were adjusted. Result: 1. Operating time: 33.5 s. Reading speed: 19.1 kb/s. 2. Operating time: 24.0 s. Reading speed: 26.6 kb/s. 3. Operating time: 28.7 s. Reading speed: 22.2 kb/s. Frightened by the results obtained, I decided to check that will be when using rst8. listing2. 1. Operating time: 16.0 s. Reading speed: 40.0 kb/s. 2. Operating time: 7.0 s. Reading speed: 91.4 kb/s. 3. Operating time: 11.5 s. Reading speed: 55.6 kb/s. You can already live with these numbers. I hope it’s clear why reading into memory above #c000 works melts much more slowly. I'll have to change the copier to mode 2.2 :) Now it's 2.1 Or maybe I won’t... I’ll mess around with the transfer. But recording slows down significantly... Moreover, regardless of the address sa. 1:33:5 at #3d13 and 23:8 via rst8... I went nuts... [hamster chews] === Cut === Do you remember? By the way, please explain, is it possible to use rst 8 to steal in the bank??? It seems like it was always #4000 - #C000? Now sit down somewhere and hold on tight. This is what the above-mentioned and excellent driver produced. There was a tube included. I read at 640k, but it turned out quickly and inaccurately, then I did itat 2560k and brought it to 640. Reading speed does not depend on the destination address. Operating time 3.2 s. Reading speed 203 kb/s. Everything is the same, but the driver is stuffed into cache: Operating time 2.3 s. Reading speed 280 kb/s. I went nuts. Then I turned off the tube and repeated everything, it worked 140 kb/s, as expected. And here is the pump itself: ;reads A*2 words to the address in HL. When A=#80 - full sector. _INPUT ticks PUSH DE 11 LD DE,#D8F8 10 _INP LD B,E 4 read from here INI 16 LD B,D 4 INI 16 LD B,E 4 INI 16 LD B,D 4 INI 16 DEC A 4 JR NZ,_INP 12 96 POP DE 10 96*128= 12288 t/sector (pump speed) 12400 t/sector (roughly) 24800 t/kb 3500000/24800 = 141.12903 ~ 140 kb/s on 3.5 MHz. When recording accordingly: 120*128= 15360. 15500 31000 3500000/31000= 112.90322 ~ 112 kb/s on 3.5 MHz 162 kb/s on 7 MHz 224 kb/s cache used. No comments. But in my probably very stupid head there is absolutely no such thing fits: HOW and WHY the shadow man manages even in the mostfast modes consume almost 70% of CPU power ??????? 2Ruiner: can you imagine HOW freaked out I was by these results- Comrade ??? WBR, Jason.

Hayushki, Valerij! 05-10-98 at 21:48:00 someone Valerij Kozhevnikov wrote 2 All on the topic 'HDD speed'. [ ] VK> Do you remember? By the way, please explain, is it possible to chi- through rst 8 VK> thief in the bank??? It seems like it was always #4000 - #C000? maybe! You didn't ask yourself this question to begin with: LD DE.????? LD HL,#C000 LD BC,#??05 CALL #3D13 with the screw image connected? After all, we climb through the shadows! ;-))) And if in nature, some kind of buffering occurs data, don’t ask me what and where ;-))) Therefore, writing/reading to the jar slows down.... [ ] VK> But in my probably very stupid head it’s absolutely not SUCH VK> placed: HOW and WHY the shadow man manages even in the most VK> fast modes consume almost 70% of CPU power ??????? the biggest brake is entering the shadow according to RST#08, he arranges _such_ brakes... Yes, if you also take into account the fact that for each IN/OUT at the beginning makes a delay of 260 cycles... (at least This is done for watches!) Well that's it *MAS* with a sledgehammer

Let your enemies, Valerij, die without sons! 05-10-98 at 21:48:00 Valerij Kozhevnikov ═> All: VK> Just recently I wrote here about a screw driver for VK> scorpion, using smoke directly. Honestly, I don't remember. :( Probably blind... Or glued rose... Senile. ;) VK> The driver, weighing exactly one kilobyte, supports 8 yst- VK> property. If you finish it a little, it will be able to work with two VK> screws. Could you throw it here? Preferably with the source, You can even just the source. It would also not be superfluous to have any information fa about direct use of the scorpion propeller. At least the ports... VK> There is only one condition: bit 2 in port #1ffd must include DOS ports VK> to direct access from RAM (as in GMX). VK> It is done simply: the bit is taken from TM9, inverted, and what- VK> rez LI1 clings somewhere. 2LD: Well what? When should I pay? ;-) I want a fast propeller... :) [skip diagram from fingers] VK> Now a little history. Well, it’s worth getting a little distracted and you’re history... :( ;-) [an army of hamsters gnawed everything in their path] OG> We'll have to convert the copier to 2.2 mode :) Now 2.1By the way, I remade it, though in a lame way - in small pieces, i.e. 17 clock cycles per byte... It will be necessary to redo the stack... Climb into this 4th tasm... brrr... OG> Or maybe I won’t... I’ll mess around with the transfer. Gee... On the same day I redid it. :) [thrown into a pit of hamsters] VK> Do you remember? Of course not. I'm talking about sclerosis. :) VK> By the way, please explain, is it possible to chi- through rst 8 VK> thief in the bank??? It seems like it was always #4000 - #C000? #24 and #25 restart work "normally" with the upper memory. A #22 and #26 are no longer there. (Already? More?). I’m generally silent about #23. Glu- codrom... [ skip measuring screw with ruler ] VK> And here is the pump itself: The pump was mentioned in a letter from LD. VK> _INPUT bars VK> PUSH DE 11 ~~~~ Do you have Scorpio? :) That means 12. [skip prog, calculations and shadow business that ate up 70% of the CPU] VK> 2Ruiner: can you imagine HOW freaked out I was by these results - VK> comrade ??? Uh-huh. :( Honestly, I'm terribly jealous of you... Seriously. Best wishes, Oleg aka Ruiner.

Hello Alexandr! Quoting your msg of [hamster chew] to me: [...] VK>> Do you remember? By the way, please explain, is it possible through rst 8 VK>> chi- VK>> thief in the bank??? It seems like it was always #4000 - #C000? AM> maybe! a more detailed? AM> you didn't ask yourself this question to begin with: AM> LD DE.????? AM> LD HL,#C000 AM> LD BC,#??05 AM> CALL #3D13 AM> with the screw image connected? AM> because we climb through the shadow! ;-))) AM> And if in nature, some kind of buffering occurs AM> data, don’t ask me what and where ;-))) So, it’s cool for me to ask, we pulled out the shadow programmer, and I climbed it up and down ;-))))))) You yourself think about what you say. :)))) But seriously, when reading through #3D13, at address #5D00 created There is a 256 byte buffer, and everything is pumped through it. The logic of the work is something like this: (before all this, the current drive is checked for connection to screw, and all sorts of pointers are determined) M1 LD HL,#5D00 LD DE, INC LD BC,#0105 RST 8 DEFB #81LD DE, ADD ,#100 LD HL,#5D00 LD BC,#100 LDIR DEC JR NZ,M1 RET [hamster chews] AM> the biggest brake is entering the shadow along RST#08, AM> he arranges _such_ brakes... well :) it even mutes AY while it's working. AM> yes, if you also take into account the fact that for each IN/OUT at the beginning AM> makes delay for 260 cycles... (at least AM> this is done for watches!) think about it ;-)))) so in your opinion: CALL PAUSE LD B,D INI CALL PAUSE LD B,E INI Then, in general, a tape recorder would turn out from a screw. =))) But reliable - awesome -=8-( ) WBR, Jason. -+- is'edit 6.06+

Hello Dmitry! Quoting your msg of [hamster chew] to me: VK>> Do you remember? By the way, please explain, perhaps via rst 8 VK>> you can chi- VK>> thief in the bank??? It seems like it was always #4000 - #C000? DL> you can, you can... it works, however. DL> and Ryinep sits next to him and adds: DL> "only some restarts..." :) WHAT??? And what is the ROM version? [the hamster ate the pump] DL> yes, but if they also addressed the registers DL> on A8, it would be another 20% faster. Yeah. In my scheme on zx_server this is how it is done. And looking at the scheme, I didn’t see any particularly evil restrictions. ny. A13 is co-addressing, not selecting. But A8 is set to A0 in the same slot, maybe that’s why it’s not allowed? DL> by the way, I advise you to expand the indicated procedure 16 times instead DL> four, maybe we can meet at 8? :-/ DL> and at the end JR NZ,.. replace with JP NZ,.. DL> if you do this, you will have 20.875 clock cycles/byte DL> (currently 24 clock cycles/byte). you can do 32 repetitions, but DL> win DL> will be insignificant (20.4375).This is, firstly, isdos, with her lack of memory, and secondly, my bad habit of chasing every byte and viciously saving it wash. ;-EEE In addition to this, there are two more pumps, for the first and second ones separately. half of the sector, they will also have to be opened. A battery - common cycle counter for all three. By type: LD A,#80 SRL D JR C, SRL D JR C, And half the sector is skipped through DEFB #ED,#70 aka INF, 80 times per cycle. By the way, I’ve been wanting to ask you for a long time, everyone has such a command Z84C0010? So that there are no glitches, like with OUT (C),0. And the JP command in the resident takes up as much as 5 bytes. Many. ;) But opening it all 8 times is already done. :) By the way, what does a CD-ROM look like from the program's point of view... uh ny that is, CD-ROM? Are there any special perversions? And does it live on the same cable with the propeller? But who knows... ugh, so as not to jinx it. VK>> 2Ruiner: can you imagine HOW freaked out I was by these VK>> result- VK>> Comrade ??? DL> Ryinep sits next to him and freaks out ;) And I tried to carry megabytes in large quantities here according to the catalogues, I just went crazy. %-) WBR, Jason.

Hayushki, Valerij! 12-10-98 at 00:08:48 someone Valerij Kozhevnikov wrote 2 Aleksandr Majorov on the topic 'HDD speed'. [...] VK>>> Do you remember? By the way, please explain, is it possible through rst 8 VK>>> chi- VK>>> thief in the bank??? It seems like it was always #4000 - #C000? AM>> maybe! VK> more details? Well, I don’t know what and how was done in the shadows, but the limitation is already removed. [ ] VK> But seriously, when reading through #3D13, at the address #5D00 created - VK> there is a 256 byte buffer, and everything is pumped through it. wow! so it can work with #C000...#FFFF. but it slows down IMHO due to page switching ;-) [ ] AM>> yes, if you also take into account the fact that for each IN/OUT at the beginning AM>> makes delay for 260 cycles... (at least AM>> this is done for watches!) ~~~~~~~~~~~~~~~~~ VK> lift your head ;-)))) so in your opinion: VK> CALL PAUSE VK> LD B,D VK> INI VK> CALL PAUSE VK> LD B,E VK> INI VK> then, in general, a tape recorder would turn out from a screw. =))) I stressed! _for_hours_! I specifically traced the process of writing/reading to/from the CMOS clock. Tama this (approximately) _OUT PUSH BC LD B,? DJNZ$-1 POP B.C. OUT(C),A RET _IN PUSH BC LD B,? DJNZ$-1 POP B.C. IN A,(C) RET I'm serious! The same thing happens when recording shadows to the #FFBA port (by the way, this port is write only ;-))) Well that's it *MAS* with a sledgehammer

Let your enemies, Valerij, die without sons! 21-10-98 at 23:10:54 Valerij Kozhevnikov ═> Oleg Grigoriev: OG>> Come on. :) VK> oh... it's big. The volume won't scare me. :) VK> Hy lana. I may still have a piece of shadow business disas- VK> Zipana collected for you here. Read while sitting. Or better VK> lying down - you will still fall. :) I picked the shadow a little, so I’m unlikely to fall. Yes, and I’m sitting I'm in a chair - it's hard to fall. :) VK>>> To him nyna prog what thread, which- VK>>> he will tell you where which devices are located. VK>>> But I have neither the time nor the desire to write it. OG>> Basically, I have a dedicated ZhddServ. So OG>> by- OG>> if Uncle MK doesn’t see it, you can rename it OG>> adapt OG>> for this purpose... ;) VK> that’s it. What is “this and that”? I have a disassembled ZhddServ. He is doing exactly what you wrote - looking in memory driver and writes all sorts of bytes into it. Remake it won't be difficult. VK> My brains couldn’t stand such abuse... VK> In the morning I went to the store, and I saw on the price tag: 1000. I looked and du- VK> May - #08? Is Chivo so cheap? :))))) :) VK> And y me it sticks out in the slot. I have everything in slots. Kind of VK> garland. Honestly, it’s a pity to take the last place for the sake of such nonsense slot... OG>> What does the process of reading/writing letters look like for you? VK> no way. chaos and anarchy. It's annoying with your eyes. VK>>> And our coordinator also changed and Cyp sent SMM :))) OG>> Yep. SMM. So the format is similar to ours? A EV OG>> LDashny y OG>> are you there? VK> NO! Nothing at all! Like in a swamp :( Well, that means I’ll throw him too. With source. :) VK> Although the headings are very similar. SMM even almost digested them - VK> But it still hangs later. VK> By the way, I taught that evil SMM to understand 4 drives. 2DP: Do you see the person suffering? Give the source of SMM... to Him. :) Although I can do it too, you look, I’ll do something clever... VK> And ZHvord is usually used as a mail editor. Well, why is it sick? He's a Mazda... I wish I could have seen him then,if you really are so disgusted by the nerdy Isedite... (Although, the speed Isedita on the screw suits me quite well.) VK> That's why no one writes anywhere - try quoting with your hands. In isedite almost automatically. The macro adds checkmarks, and Quoting is done by pressing 3 buttons. Rulez isdosa consists of that it can be configured for any email format. Well, almost any. :) VK> They call by voice, or communicate using sysopnik. :*) The voice of course rules, but the advantage of letters is that over You can think normally in writing. And the sisop usually turns - going on a drinking spree, peeling potatoes and smashing chandeliers. :) (Joke for those who are in the know). VK> Many of us here are sure that CODE.ZX is generally a fidish echo :) :) Yes. 2All: As the moderator of this gathering of cunning perverts, I I declare with full responsibility that echo has nothing to do with Fido. th relationship. Almost. :) VK> ps VK> and I received a bucket of yuyukov, thank you. Catch another one. :) VK> This happens with us: eh VK> is always clear, but soap may be late. Hm... I threw the soap on the 17th, I just didn’t change the date nope. Answer me a simple question: have you seen Yuyuki in the echo? VK> Such a question: Rhddsave, Rhddconn - it requires _boot VK> th_ device or any? Is boot.sys required? No. VK> And then I started to launch it, and it viciously shit on my bitcard, VK> and some files nearby. VK> Or maybe I’m just so old-fashioned? Compilation date - VK> beginning of 1994. And it was sold as 5.0. :-/ 1. As far as I know, Slot installed versions from the idiot... My number is 4.5, date 7/16/96. 2. Wonderful, wonderful, why did I throw you a swap? There's isdos and a lot of residents of all sorts of useful... And the 11-letter catalogs are takes it. Try connecting to this swap, booting from it and configure the system to suit your needs. Now there is a,b,c,d - disco vye, e - ramdisk, f,g,h - screw. f - system. I can throw is_dos.sys. 3. I’ll warn you right away - I won’t go to Moscow to get the isdos. VK> And the help to Run_me.B was clearly written by someone completely stupid. Details VK> describes where and how to press the enter button, and not a damn thing anymore. The help for run_me is self-sufficient. Or rather, it’s simply not needed. Copy it to a floppy disk, run it and press reset. Should appear- whole menu with processing. Unless, of course, you have a hundred- it autoconfig on.Best wishes, Oleg aka Ruiner.