Nicron #31 24 апреля 1997

Программирование     Бейсик

Programming - BASIC Calculator (continued).

WLODEK BLACK


      
CALCULATOR BASIC  

(Continued)

Work with real numbers in machine code, of course,
more complicated than in BASIC, but not so much that it
becomes incomprehensible. When using the calculator of the codes
we must first clearly separate the data on the constants and
changing along the way the data (do not want to use the term
"Variables"). Constants makes sense to calculate in advance and
recorded in the body of the program as a finished product - a 
5-byte internal representation. Determine the value of the 
constants in the 5-byte representation easily directly in 
BASIC. Type the following line: 

10 PRINT 2.7182818

(This is the number of "e"). After seeing the contents of 
memory at address 23872 (If TR DOS initialized), we find a

code sequence:


F5 32 2E 37 31 38 32 38 31 38 0E 82 2D F8 54 3A

P 2. 7 1 8 2 8 1 8
R
I
N
T


Marker # 0E shares ASCII-record numbers, and an internal 5-byte
format. Writing off somewhere, and the values ​​of 5-byte stand 
after # 0E, we remember the constant value. Data after # 0E

intended to register in the following order:


0E 82 2D F8 54 3A


     A E D C B


Zanesya them in the processor registers in this order and 
calling then the procedure STK-STORE (# 2AB2), we obtain a stack

Calculator value desired number. It is clear that by writing
after the "PRINT" any number of viewing and then the memory can 
be learn 5-byte representation of any real value in

range permitted by BASIC.

Attention - glitch! And apologize. In the previous issue has 
crept inaccuracy. Here is the correct list of procedures 
STK-STORE: 

# 2AB1 - entry into the stack string parameters (A reset);
# 2AB2 - entry into the stack 5-byte value numbers;
# 2ABB - entry in the register stack A, E, D, C, B without 
control. 

There is the so-called "packaged" form of numbers,
requires only 4 bytes to store values. To be frank,
never tried to use it. In one clever books
mentioned "glitch" packaged forms: 0.5 <> 1 / 2. Oh well ...
If the calculations or stage I / O will be found
any incorrectness, BASIC conversion to occur through RST # 08
and a stack of return by mistake. In other words, if the system 
BASIC was not modified (corrupted), for any traditional

BASIC error would lead to "escape" of the user program in
Normal mode for entering commands from the BASIC error message 
at the bottom screen. If this is unacceptable, it is necessary

reorganize the stack return by mistake (ERR_SP). Honestly
speaking, I do not know whether we should now tell you how it's 
done. Let's proceed like this: you write to me or to the 
address of the editorial board, If a story is needed, and it 
will appear. 

If in the course of the implementation program will require 
mashinnokodovoy enter numeric data from the outside, then at 
least with the keyboard it is wise to do as a normal character 
figures, as we enter the data to a request for INPUT. INPUT 
simulate the codes are not costs, better ensure that you enter 
a character string in any known to you the way and further 
transform introduced sequence of characters in the internal 
5-byte representation of numbers. So we have: symbolic image

numbers, should contact: the internal representation of 
numbers. Very easily this is done by evaluating the expression 
VAL "string". Parameters of the input string (starting address 
and length) are recorded on stack calculator, and then executes 
the function VAL: 

    LD DE, ASC_A; address of the start line
    LD BC, ASC_B-ASC_A; length of the string
    CALL # 2AB1; STK-STORE for string parameters
    LD B, # 1D; function code VAL
    RST # 28; call calculator
    DEFB # 3B; code of the function fp-calc-2

; At this point in the stack is an internal representation of

    DEFB ....... ; Codes of other operations calculator

    DEFB # 38, the output from the calculator
    ...........
ASC_A DEFM "2.7182818"; character string representing the 
number of ASC_B


Function VAL in his work is a calculator for recursively
perform multiple recalculations of values ​​ASC-codes and other
intermediate values. Because of this, it can not be included in 
the overall List DEFB opcode calculator, as in this case

other data are lost, and there is some mistake
BASIC. To call a recursive function is provided
special operation "fp-calc-2" - secondary call calculator.
When using fp-calc-2 operation code is put into register B,
and the body put the code DEFB # 3B - appeal to the fp-calc-2. 
After # 3B codes may be located downstream operations 
calculator. The same should apply to all functions, evaluates 
the expression including the string - VAL $. A string of 
characters defining number may not be an image of itself - in 
line You can write any expression, even with variable

BASIC, and it will be evaluated! Not hard to imagine that
this kind of work involved and the BASIC interpreter. If the 
function VAL (or VAL $) is only called to convert the

expression, and then immediately carried out from the calculator
(Say, for the preparation and input other data), access to VAL
You can podsokratit, causing it not through the fp-calc-2, and 
directly CALL through the well-known address: ... bummer! The 
handbook ochepyatka ... Address VAL $ - # 35DE, and VAL -? 
Well, okay, VAL $, so VAL $. We show the use of VAL $. Here is 
a fragment of some small utilities that I use to process server

disk processing "history", and similar concerns sysop:

BASIC:

INPUT "Source file:"; a $: INPUT "Destination file:"; b $: 
RANDOMIZE USR ....


Codes:

    LD DE, VARLIN; address of string "and $"
    LD BC, 2, string length
    CALL # 2AB1; STK-STORE
    CALL # 35DE; VAL $
    CALL # 2BF1; STK-FETCH
; At this place in DE - the address of a character variable a $,
; In BC - the length of a variable $
    LD A, C; check no more than 8 characters in it?
    CP 9
    JR C, VL8; if not more than 8 characters
    LD C, 8; otherwise take only the first 8 characters
VL8 LD B, 0; it just in case - maybe someone else
; 264 characters introduce? :-)
    LD HL, FILNAM; buffer address file name
    EX DE, HL
    LDIR; file name of the variable a $ forwarded to clipboard

    .............

VARLIN DEFM "A $"; "image" of a character expression,
            consisting only of one variable a $
    .............
FILNAM DEFM ""; buffer file name

And so, without much difficulty, and, most importantly, very 
quickly is a name to a file in a buffer, from which then

file name is used in accordance with the algorithm of the 
program. CALL # 35DE replaces the chain:


    LD B, # 18; opcode VAL $
    RST # 28
    DEFB # 3B, # 38,

that is 3 bytes instead of 5, plus an increase in speed of 
execution. Of course, if after DEFB # 3B followed by other 
opcodes calculator, it is wiser to call a VAL $ fp-calc-2.


Through the VAL and VAL $ can be easily calculated various 
useful functions type RND, INKEY $, etc:

RND = VAL CHR $ 165;
INKEY $ = VAL $ CHR $ 166.
Moreover, VAL and VAL $ from a single simple value (not the
expression) can be calculated directly without fp-calc-2. But 
still better not to risk it!