Spectrofon #17 01 февраля 1996

Железо

System - discussion of the pressing issues of standardization Spectrum-compatible computers.

         C & C T E M A



   The proposed lower material lifts,
In our opinion, a very interesting topic. Pass the word to our 
correspondents from Kharkov.


Jemmini: In a nutshell the problem. Us
quite amazed at the number of programs from
Russia, which focuses only on
"Pentagon" and not for anything else. These products are in 
most cases looks very bad on branded machines and machines with 
separate field memories. One gets the impression that this is 
not for SPECCY write, but for some unknown reason and

What! Maybe the standard was already not
SPECCY, and its home-grown equivalent? One
In short, I beg you, pick up this
painful question in his journal, for
Once your log can here something to help!
(C) Jemmini 'Metallica


       "Slow" MEMORY SPECCY



      ... Many of them are permanently left in

      past, some transformirova
      were, some were born from "mon
      Stry "...


      (From the preface to the book Rodionova)



               ATTENTION!


   Through the magazine "SPECTROFON" We want to appeal to all 
programmers and hackers that may not her own fault, forgot or 
did not know one very important Property SPECCY, which we try

tell.



Jemmini:


   Before I begin to expound the essence of the problem, I want 
to note one difference in the sea sovdepovskih schemes. There 
are two implementation: a common field of memory and a split. 
The first is "The Pentagon" LENINGRAD-1, 2 "," Baltika ", etc. 
Representatives of the second implementation is "Krasnodar", 
"Kharkiv", "MOSKVA-48" and ... Corporate SPECCY!



   In corporate SPECCY, video controller for reading data, 
there is a "lock" clock, going to the CPU. Thus stopping the 
processor is implemented on time- information from RAM to the 
screen. As is known, SPECCY video memory is measured in pages

RAM with the numbers 5 and 7. Therefore, the video controller 
makes the choice of data is from this memory location. But 
SPECCY arranged so that these pages are from

pages 4 and 6, one block. Only
the sake of convenience I'll call him a block of video memory 
(hereinafter referred to OHR). Jobs VWP because of references 
to it slowed down the video controller, namely by 25% (The fact 
that the treatment "cover" two clock cycles out of every 
eight). This slowdown in certain cases may affect the operation 
of the computer in general, and on the last speed affects the 
frequency dependence requests the CPU to the OHR.

I note that most of its value
This frequency is reached for sample codes
teams from the said memory. Therefore, for
to minimize accesses to the VWP, a rational place in the data. 
It for data and apply this memory.



   One of the severe cases is the one
when placed in the VWP program. With this
placement, performance on this segment of the program falls on 
the fourth. 



   But not even that made me write
the real message, but something less obvious but no less 
serious and unpleasant! AND these troubles are related to the 
habit of many homegrown ...- soft'ov set to register interrupt 
vectors I a number, indicating that this unfortunate OHR. This 
lining leads to an even more tragic results. 


   As I've written a program fragment
caught in the VWP, loses 25% of the speed, but
If there comes a table of interrupts that
The whole machine becomes slower by 25%! Not
delving into the reasons why I say what this
leads. No, the car did not explode and did not hang, and 
all-but everything stopped "Stowed in the interruption," as 
adopted to speak with programmers. But this is not

all and not even the worst when it would seem, and
well enough. The trouble begins with the
boot from a disk - DOS can not read
drive!


   Indeed, its rate of rotation remains
the former, and the speed DOS fallen almost 25% in
resulting in all the time delay
were wrong! Error reading
and I was there to tell you how many prog
respond to this ...


   It is worth noting that the above
"Property" does not depend on the mode interrupt!


   As a solution to this problem, we recommend that all 
programmers be installed in I register values ​​in the range 
from # 00 to # 3F, and from # 80 to # BF. In this setup,

overlap occurs. You can also install a range of up to # C0 # 
FF, provided that the current in the vast majority of pages are 
not owned by the VWP. In the case of the values ​​from # 40 to 
# 7F, the interrupt vector will always point to the OHR, which 
will inevitably lead to a loss of speed. 


   Dear colleagues from Russia! Did you
has only "Pentagon"? Can not you see how to do all firms, as
Poles have done (they have been branded machines!), and 
finally, as is done in all known SATISFACTION and INSULT? Maybe 
you're just not paying attention, and yet there all with this 
in mind: the block of video memory - the data, the rest of 
memory - a program. And you have companies will not find, so I 
register it with the values ​​from # 40 to # 7F.



   What about the situation in Ukraine
I can assure you that the machines with a separate field of 
memory is plenty. And judging by the publications in the journal

"SPECTROFON" in Russia similar machines
also available.



Jemmini 'Metallica:


   All who read these lines! Take into
account the above and do not be lazy on
When you work with DOS in case I bring
# 3F, and RAM to install interrupt table so that it does not 
hit the block of memory! And we can only hope that those who 
admit such mistakes, the more they will not do that our story 
will be useful to all of you, dear sinkleristy. Bye! 

Matviychuk MV (Jemminisoft)
310066, Kharkov, trans. Progressive,
d.28, kv.1 tel.: (0572) 98-81-35

Grebennik VA (Metallica)
310091, Kharkov, ul.Tankopiya, 5, kv.44
tel.: (0572) 90-45-24


                  *


"S": It should be noted that this problem in detail all the 
three volumes of "Programming in machine code and assembly 
language" published "Inforkomom" in 1993. But since this 
edition is now turned into discharge a rarity, information 
Kharkov programmers certainly cause of interest. Once again I 
draw your attention to the fact that I hit the register in the 
range from # 40 to # 7F slowing your computer is NOT DEPEND ON

Interrupt mode! In each machine cycle
Command Z-80 refers to the memory cell
addressable register I, displaying it in the senior eight bits 
address bus and activating the line of inquiry memory MREQ. 
This is interpreted as the ULA address to the OHR, leading to 
conflict with the controller display regardless of the mode 
interrupt. However, we also do not recommend the use of 
register values ​​I from # 00 to # 3F for the second treatment 
interruption, as 


with the table of cut falls in the ROM that on computers with 
an unstable system bus data leads to unpredictable results. 
Apparently, the optimum will use the I register of

range from # 80 to # BF.