Info Guide #02

Самиздат

Story - The story of Mel, Hastoyaschego programmer.

   This epic poem was sent to
USENET by its author Ed Neserom (utastro!
nather) May 21, 1983.


 History Mel, Real Programmers.



   In a recent article devoted to the courage of professional 
programmers, with no all sorts of ploys the author declared: 
"The real programmer writes only in FORTRAN. 

   Maybe someone else is doing it in
our decadent era of clarified beer
pocket calculators and "user-friendliness", but in the good old 
days when the word "Software" sounded like something

strange, but the real computer to do
of drums and vacuum tubes, Real
Programmers working in machine code. Not
FORTRAN. Not on RATFORe. And not even on the
assembler. Just machine codes. No
prikrashennye, obtuse raw hexadecimal numbers.

   Has grown up a generation of programmers who do not know 
something about our glorious past. And I realized that I must 
fill that gap, I have to tell young people about how to write 
your code present Programmer. I'll call him Mel,

because that was his name really is.

   I met Mel by going to work at the Royal MisiBi Computer 
Corporation, which is now no longer exists, but those days was 
hoo! The firm has produced LGP-30 is very small and very cheap

computer (of course, by the standards of those
days), which used drum memory organization, and only just
starts manufacturing RPC-4000
large, advanced and faster
computer and also with a drum memory.
Magnetic core memory in those days was very expensive - 
probably why you never did and never heard anything about this 
company, nor on this computer. 

   I had to write a Fortran compiler
for this supercar, and Mel was my guide. He never approved of 
compilers. 

   He was asking me: "Well, what good can be in a program that 
can not even change my code?" 

   In hexadecimal native Mel
written the most popular program of
All that only existed in the company. Her
run on LGP-30 on various computer trade shows, and she began to 
play with potential buyers in the Black Jack and always 
produced a sensation. At each exhibition hall around the LGP-30 
crowd sellers of IBM, but we never broke

their heads over so well or do not sell our computers.

   Mel had to rewrite his toy
for RPC-4000. (Port? What is this?) In
new car used for addressing
the scheme "one plus one", so each
machine instruction, except for the operation code and
addresses the right operand input of the second address that 
indicated where the rotating drum recorded the following 
statement. 

   In modern parlance, after each
command, the operator GO TO! Zasun'te all in Pascal's pipe and 
smoke. 

   Mel loved the RPC-4000 because he could
optimize the code for it: he wrote
machine instructions on the drum so that as soon as one 
completes the instructions under the read head instantly 
provided Next, the processor could immediately be taken for the 
performance of this team. Was special program that all these 
things, and called it "optimizing assembler", but Mel refused 
to use it.


   "You never know what address
She writes the command, "- he explained
me - so I had to use
individual constants.

   It took a long time before I was able to
understand his words. Since Mel knew by heart the numerical 
value of each team and always appoints address the team on a 
drum, he could have each of his instructions to use as a 
numerical constant. For example, he could pull out from the 
memory of the old manual "add" and multiply it. It is not easy 
would the person who would even think of to change something in 
his program! 

   I compared the program debugged Mel manually, with the same 
programs that have gone through "optimizing assembler", and 
each time it appeared that the program Mela is faster. 

   This was because in those days was not yet invented a method
programming top-down "and therefore
Melo did not need to use it. First he wrote the most critical 
internal cycles, so in the future could choose the most 
appropriate address on the drum. No optimizing assembler could 
not do so. 

   Mel never inserted into their empty cycles for the delay, 
even when I had to write to the antediluvian

Flexowriter, which required that at
displaying symbols has always been a pause between
two characters, otherwise he was choking.
Mel so arranged his team in memory,
that every time they needed a delay
the next instruction is behind
Readhead exactly one position,
therefore, the drum had to make
full circle, to find the correct address. He
even invented a term for this procedure. We know that there are 
universal The term "optimal", in order to deny sometimes say, 
"suboptimal" or "not at all optimal "or even" not very optimal. 
"Mel called the maximum time delay to find the right address on 
the drum of "the most pessimal. 

   Shortly after the program was
put ("Even though the initialization is optimized," - he said 
proudly), from sales Request for change has come. For the cards

and select the next card from the deck in the program used an 
elegant (and also optimized) random number generator and some 
marketoidy decided it was too too be honest - the players and 
the car were equalized at odds, so the potential

buyers could be a loser!
They demanded that Melissa, that he bit
changed its prorammu, to be her
stopping and using a special console switch to choose from a 
deck appropriate map. Then the buyer can always win!


   Mel balked. He felt that
it will be apparent fraud and compromise him as a programmer, 
so he refused to do so.


   First Mel persuaded Chief Marketoid, then the cause took the 
main boss, and after that, as Mel was not scared and the owner, 
he made a deal to take those programmers, who went to the 
friends with Mel. Finally Mel surrendered. He rewrote the 
program and put it in a test that checks whether a switch 
occurred, and if the answer is yes, then the program knew how 
to cheat better than human when installed was a new switch,

it turned out that now wins only
program! Mel did not conceal delight
declared: "Now everything is honest and with a haughty air 
refused to make any corrections. 

   Of course, Mel retired. A week later
Big Boss called me into his office and ordered to investigate 
the program, and Mel try to detect the test. And in the case 
good luck I had to throw it. That do? Poupiravshis how could I

agreed - but in the end, only one
study of any program of Mel did not
worse than the most incredible adventure!

   I've always intuitively felt that
Programming is an art which assess truly capable of
Only he who himself is a master in
this case, what are the depths of the human
thought the essence of programming hidden away in long columns 
of numbers. You can learn a lot about a person just by reading 
written by the program, even a native. I realized that Mel was 
a genius. 

   At first I was shocked to discover that in
the main loop of the program there is no test. No test. No. 
Common sense suggests that there must be a cycle

which turns the program may
indefinitely. But nowhere is not transferred to the control of 
the cycle. Two weeks out of me trying to understand what was 
happening. 

   Machine PPC-4000 was progressive in
the times means, which is called
index register. This allows the programmer to organize the 
cycle with Indexed instructions: each time the contents of 
index register was added to operand address, and it allows you 
to access the following data elements in the array. Each time 
it was necessary only increase by one the contents of the index

register. But this is not Mel never used it.

   Instead, he pushed the instructions in
register, then adds one to its address and
hidden back in the memory. After that, he could
carry out this change of instructions directly
from the register. Cycle has been organized in such a
a way that took into account the additional
time spent to perform
this instruction - as soon as it finishes, the next is exactly 
under the read head drum, ready to make it read and execute. 
But in This cycle was no test.


   The clue was in a bit, which indicates the index register is 
used or not. It was located between the address operands and 
the operation code in a word instruction. It was installed, 
although I knew that Mel never works with an index register, 
and in it all the time should be zero. When I realized vision 
of Mel, he blinded me with its beauty and perfection. 

   All the data that he needed to
work, it is situated in the uppermost
memory addresses - the biggest addresses that were only 
available. And after processing the last item in the array of 
incrementing addresses instruction led to overflow: Bit 
transfer unit would add to the code of the operation, thereby 
altering its values, and it becomes the following in

instruction set, a jump instruction.
Naturally, the next instruction after it was recorded already 
at zero, so the program quietly jumped back.



   I never again saw Mel, so do not even know he was able to 
withstand of the current wave of innovations that make it great 
art in the collection of primitive shabby methods by which any 
more or less educated guy can cook up a working version. 

   I would love to believe that Mel
did not succumb to temptation.


   Being able to finally understand the structure
program has no difficulty, I found it unfortunate place to check
special switch. Of course, the Big
Boss that I said nothing, only reported that he could not 
understand the program; looks like it is not surprising.


   When I left that company, blackjack is still starting cheat, 
it was worth Right click on the console switch,

and I think it should be. I've never
could not forgive myself if I zahachil
Real Programmers code.