ACNews #23 17 сентября 2003

Программирование

Coding - On yet another way to the fort-like coding.

■ ■ ■ ■ ■ ■ ■ ■ ■ ■

  ■ ■ ■ ■ ■ ■ ■ ■ ■ ■ ■


          On yet another way to the fort-like coding

 Echo35 (2:5029 / 35.26) ZX.SPECTRUM
 Msg: 447 of 547 -401 Snt Pvt Loc
 From: Dima Bystrov 2:5029 / 35.26 16 Jan 2003 17:44:07
 To: Nick Sevryukov 16 Jan 2003 18:41:12
 Subj: Re: FORTH

Hello Nick!

14 Jan 2003 12:58, Nick Sevryukov wrote to Valerij Kozhevnikoff:

 VK>> Here are just a stack to be looked after. And then recover
 VK>> if necessary.
 VK>> All depends on how often and how is it called and how
 VK>> must be
 VK>> stack all the time.
 NS> The main thing is that it makes all the prog Nitak wrong. 
You  NS> actually copied, only the last command jp (hl).

 NS> Read what I wrote earlier in the notes to the way it is
 NS> should work.
 NS> This piece is often enough to spoil and a stack of very
 NS> desirable! Simply, he (the stack) is used actively.
[Skip]
 NS> Or you can arrange for another, only to remember that
 NS> have an array of addresses, which is posedovatelno
 NS> transition and a pointer to be automatically
 NS> rise to the following address (see above).
there are more crazy idea:)
does not have an array of addresses, and with JP ...
a first approximation, it seems that there some losses, but it 
is not so! let the HL will be the current address (similar to 
(SP), but without treatment to the stack)

why it is profitable to HL, in the course text will be 
appreciated:) 

"Larry is as follows:

    inc hl, hl, hl

    ...

    jp (hl)
(+ JP = 18 +4 +10 = 32 clocks)
(For comparison, call + ret = 27 cycles)
best inc hl, hl, hl place at the beginning of subroutine
with jp (hl) - in the end.
if hl is vital, it can be
push hl
...
pop hl
or (simultaneously with the readout parameter from the stack)
ex (sp), hl
...
pop hl
or (in conjunction with zapihivaniem result in a stack)
push hl
...
ex (sp), hl
(And other variations)

(A conventional method, a call-ret, is difficult
use the data stack
inside a subroutine, call-ret loses about 40 cycles:
ex (sp), hl: ld (exit +1), hl: pop hl, de ... exit: jp 0)

between the jp-s can be inserted pieces of conventional 
mashkoda (plain code), but we must not forget to restore hl. 
for example: 

ld hl, $ +6
jp kudato +3
thus skipped inc hl, hl, hl.
or even
ld l, $ +5; h was equal to the old
jp kudato +3

if we assume that the run-time "shell" that
method and the classic is not just about concede to each other, 
then there ^ ^ ^ on this piece of plain code we win 8 or 11 
bars! (By length is obtained by losing 02.03 bytes) in this 
case of (those same team) can be written even design plain 
code; call; goto! (Won by about 18 cycles and an average nibble)


so looks like goto in the normal case:
jp metka
ld hl, kuda
jp (hl)
In fact, this record will never need, since case
where to goto is plain code, we have discussed, so we consider
case where the call is to goto:
ld hl, kuda; or ld l
jp metka +3
(Not quite a logical order, but to write such a compiler can)
then winning 8 beats (or 11) and, if lucky, a byte.

Now consider konktruktsiyu call with a numeric parameter
in ordinary life is ld de, param: call metka
but we will have
jp metka
dw param; so you can make a lot of options
...
metka

    inc hl, hl, hl

    ld c, (hl)

    inc hl

    ld b, (hl)

    inc hl
...
(Of course, the option to ld is also possible, but it is rarely 
beneficial) 

conditional branch can be written as:
jp jpZ
dw kuda
...
jpZ

    inc hl, hl, hl

    jr z, jpZyes

    inc hl, hl

    jp (hl)
jpZyes

    ld a, (hl)

    inc hl

    ld h, (hl)

    ld l, a

    jp (hl)
or, if speed is important,
jp z, kuda
ld hl, $ +3; or ld l, $ +2
...
kuda
ld hl, $ +3
(Then losing 6.5 bytes instead of 2)

loop with a precondition:
on an ordinary ACME writes:

    <...>

    jp while
loop

     while


    

    jp cc, loop

    <...>

but we will have (if we assume that <...> was call metka, and
 was call metka2):

    ld hl, while; or ld l

    jp metka +3
loop

    ld hl, $ +6; or ld l, $ +5

    jp metka2 +3

    <More body>
while

    

    jp cc, loop

    ld hl, $ +6; or ld l, $ +5

    [Call metka3 +3]
(Gain in speed of approximately 9 cycles per iteration of the 
loop, loss of $ 5.4 B)


intermediate result calculations shall be in the de

For example, the expression with the addition of:

push de

pop hl
add hl, de
ex de, hl
ld hl, $ +3

or subtraction:

push de

pop hl
or a
sbc hl, de
ex de, hl
ld hl, $ +3

or shorter, but more slowly (so you can record and any other
operation):

push de

jp subpp
...
subpp

    inc hl, hl, hl

    ex (sp), hl

    or a

    sbc hl, de

    ex de, hl

    pop hl

    jp (hl)

maybe 2 spetsvyzova - jp (ix) and jp (iy), for something very
frequent

In general, it is less a fort, and the "alternate coding";)

 - Alone Coder [ZX-Guide] [Inferno] [AlCoNews] [ACEdit] [PT]
[Chip] [STS] [ANSI] [8col] [UNRAR]
--- GoldED/386
 * Origin: The queue is in the store, "said T (c) S,-and this 
is SHOT 

:::::::::::::::::::::::::::::::::::::::::::::::::: 
:::::::::::::: 

--------* List Honored Neispolzovateley AC Edit *-------
1) Pawel / Real_soft Paul Kislyak 2:454 / 11.25
2) Kurleson / HorrorSoft Alexey Kravchenko 2:5068 / 6
3) SirX Sergei G. Shelginsky sirx@atnet.ru
 Arkhangelsk Region., Severodvinsk, ul.Zh / Dorozhnaya 35-51
 (818-4) 55-10-89, 8-921-670-02-85
4) Alex_Xor / PSV Alexander D. Shushkov axor@mail.ru
 (8172) 25-28-71, 160035 Vologda POBox 136
5) Diver/4d Alex Golubtsov orangefox@mail.ru
6) Noviset / Brutal_Creators Igor Shakhov [2:5050 / 109.52]

----------* List Honored Members AC Edit *---------

                  Honorary and eternal:
1) Nikphe Aleksey Emelyanov
 391420 Ryazan region., Chuchkovsky rn, p.Avangard, Ruleznaya 12
2) Mr.Event Alexander Lesovskaya
 391420 Ryazan region., Chuchkovo, Kalinin, 26 (8-238) 
07/10/1973 3) CyberDemon Maxim Mit'kin 2:5029 / 35.18

 Ryazan, ul.Novoselov, 50k2-197
4) Basil Basil E. Loginov, 390035 Ryazan, Poletaeva ,15-5
5) John_Silver Denis E. Loginov there:)
6) TS Sergey Tyurin triod@aport2000.ru
 390029 Ryazan, High-,16-53 (0912) 75-40-74
 8-903-640-01-49
7) U # 74 Sergey Machines u_74@mail.ru
 443065 Samara, Pugachevskiy_trakt d.66 kv.3
8) Shaitan / SoK Maxim Shashkin 2:5057 / 21.36 (8462) 34-82-92
9) Dissonator Constantine Shaev
 390048 Ryazan, ul.Zubkovoy ,19-106 (0912) 41-03-50
10) Way of the Samurai, Alexander Golovanov [aiki@mail.nnov.ru]
11) Dark Raven Semen Tsehmeyster
 187500, Leningrad Region. g.Tihvin Mashinostroiteley 44, kv.151
12) Disabler / DPL Andrey Gorelov
 344045 Rostov-on-Don ul.Mironova 3 / 3, kv.120
13) Max / Cyberax / BDA Maxim Arnold [amv@kuzbass.stack.ru]


                       Other:
1914) Stormbringer / Cobra_Soft Valera Selev 2:462 / 162.7
 79022 Ukraine, Lviv-22, PO Box 798
15) AmoNik Nikolai Amosov 2:5030 / 978.14, 500:812 / 2.26 
(SPbZXnet) 16) Macros Sergei Gulyaev zxmsf@mail.ru

17) Jerri / AF Yuri Potapov, 2:5057 / 21.89 (8462) 92-73-26
18) Ldir / mp., I8 Jaroslav Turbin (0912) 77-40-37 
last_128@mail.ru  390000 Ryazan Pervomajskij Ave 40k.1, kv.18

19) Chapay / mp. Viktor Spiridonov (0912) 76-07-36 
mir_tut@mail.ru  390029 Ryazan ul.Stroiteley 1, kv.10

1920) Voodoo / Delirium_Tremens Dmitry Voo vooha@ua.fm
21) Risk / O