:: sf ghoul :: corewar

Maezumo: The Corewar Evolver with a Twist

2009-07-08T09:11:00.003+01:00

2009 looks set to be the year of the evolver with 5 new evolvers to date. The latest offering is Christian 'Fizmo' Schmidt's Maezumo, the Corewar Evolver with a Twist.

After extracting the archive and changing the settings in maezumo.ini for nano I fired up Maezumo. The evolver produces a table of results in index.htm which refreshes every few minutes.

For the first 93 steps simple coreclears ruled the day. Warrior 93 was a simple paper and over the next 20 or so steps papers managed to take complete control. By step 130 most of the papers had disappeared to reappear again at step 174.

At this point I noticed a potential problem. Whether through my error or a bug in the program, the last line of every warrior is always dat.f $0, $0. So far I haven't figured how to fix this.

My first impressions of Maezumo are that it's effective, maintains diversity in the pool and the reports look fantastic. For a future version it would be nice to see more statistics in the reports, e.g. lineage.

A New Arms Race, the Tiny Evolvers

2009-07-06T10:54:00.002+01:00

A new Corewar arms race is in progress, the battle to create the smallest effective Corewar evolver. There are currently two contenders:

Terry Newton: Terry is the evolver guru having previously written three evolvers, Fizzle, REBS and Redmaker. Terry has three new tiny evolvers, JB, SEWAX and LCE. At least two of these have produced nano warriors that entered the hill.
John Metcalf: John is a hand-coder who recently got tempted by the dark side of Corewar! John has published two versions of his tiny Corewar evolver, TEV. Both TEV0 and TEV12 have created warriors which entered the nano hill.

Right now, I'm wondering if anyone will break the 1K barrier and if so, who :)

April Update: What's New in Corewar?

2009-05-02T21:02:00.001+01:00

Here's a quick summary of what's new during April:

March Update: What's New in Corewar?

2009-03-30T10:30:00.004+01:00

Here's a quick summary of the last couple of months:

Blog posts:

Articles and Software:

nano papers: ghoulish and Apis Mellifera

2009-03-13T07:53:00.004Z

Here are a couple of nano paper/clear I've been playing with recently using a mov.i #1,<1 anti-imp clear.

;redcode-nano
;name ghoulish
;author S.Fernandes
;strategy paper/clear
;assert CORESIZE == 80

pstep   equ     -39
cpos    equ     -18
dpos    equ     27

paper   spl     #-21        ,    <-32
        mov     }paper-1    ,    }pboot
        mov.i   #1          ,    <1
pboot   spl     <pstep      ,    cpos
        djn.f   paper+1     ,    <dpos
        end

;redcode-nano
;name Apis Mellifera
;author S.Fernandes
;strategy paper/clear
;assert CORESIZE == 80

pstep   equ     -22
cpos    equ     -18
dpos    equ     -8

paper   spl     #0          ,    <33
        mov     }paper      ,    }pboot
pboot   spl     <pstep      ,    <-16
        mov.i   #1          ,    <1
        djn     paper+1     ,    #dpos
        end

Redcode Forth

2009-03-01T08:11:00.002Z

Redcode Forth is an interpreter for a 36 word subset of the Forth programming language. I don't know much about Forth, but fortunately the interpreter is supplied with a summary of the instruction set and a step-by-step example.

The interpreter uses a new Redcode instruction to display text, sts. I wonder if John is also planning to support ARES which has a different system for i/o?

January Update: What's New in Corewar?

2009-01-30T10:38:00.005Z

Here are the latest Corewar updates for January:

The DynaHill at Corewar.Info

2009-01-16T12:34:00.002Z

Towards the end of last week, Christian Schmidt announced a new Corewar hill, hosted at corewar.info. The Corewar DynaHill is build around a different concept to KOTH and SAL.

The battles are scheduled at regular intervals, regardless of the number of submissions. When battle takes place, the warriors fight their neighbours and move up or down the ranks depending on the number of wins or losses.

The hill is unlimited in size. LP and '94 Standard warriors can be submitted to the same hill. The battle settings are decided from the warriors' ;redcode declarations.

Apparently, the system is similar to the rankings used in Japanese Sumo. To find out more, visit the DynaHill FAQ.

December Corewar Updates

2009-01-01T22:31:00.003Z

Here are the latest Corewar updates from December:

Corewar: Hostile Programming

2008-12-08T12:25:00.005Z

Corewar is a game from the 1980's, played between computer programs written in Redcode, a language similar to assembly. The programmers design their battle programs to remove opponents from the memory of the MARS virtual computer by any means possible.

Some of the simpler techniques include blindly overwriting memory, searching for the opponent or spawning off new processes. These are commonly known as stone, scissors, paper after the popular playground game. Stone usually wins against scissors, scissors normally defeat paper, and paper mostly beats stone.

Here's an example of a typical Corewar program:

     org   wipe

     step  equ 5
     first equ bomb-10

bomb:mov.i #1,       -1

ptr: sub   #step,    #first
wipe:jmz.f ptr,      @ptr

     mov   bomb,     >ptr
     djn.f wipe,     {ptr-5

     end

This simple example of scissors once held a 20 point lead over it's rivals. The first instruction is never executed, it's the bomb used to overwrite opponents. The next two instructions form a loop which looks through memory for an opponent, and the final two instructions actually overwrite it.

Corewar is still going strong, and celebrates it's 25th anniversary in 2009. If you'd like to discover more about Corewar, here are the top resources:

The Beginner's Guide to Redcode will teach you the language of Corewar

pMARS is a portable implementation of the Corewar virtual machine

Corewar Tutorials exist on virtually every aspect of the game

Koenigstuhl is an archive of thousands of published Corewar programs

SAL organises a number of on-going king of the hill tournaments

sfghoul and impomatic report the latest Corewar news on their blogs

#corewars is the official Corewar discussion channel, hosted by irc.freenode.net

What are your experiences with Corewar, have you ever had any success?

Corewar: Hostile Programming is a guest post written by John Metcalf.

nanoWarrior #03

2008-11-19T20:00:00.004Z

The latest issue of nanoWarrior has finally been completed, containing articles by John Metcalf, German Labarga and me. Check it out for yourself here:

http://corewar.co.uk/nw/nw03.txt

Neo's Mini Challenge - Roy's Idea

2008-11-12T06:56:00.004Z

Roy suggested a tricky extension to Neo's Mini Challenge in rec.games.corewar:

Lets try to find a solution for this gate:
djn.f #0,>-5
(I haven't got one yet)

It appears a one instruction solution isn't possible. However, 5 players found a solution in two instructions. Congratulations to Ilmari, John, Rashnok, Roy and me! Also, well done to German and Lukasz who also discovered solutions, albeit slightly longer.

I discovered a five line imp:

imp mov.i #4,   *1
    mov.i #4,   *1
    mov.i #4,   *1
    mov.i #4,   *1
    mov.i #4,   *1

By adding a loop, I managed to reduce the initial size to two instructions:

imp mov.i #4,   *1
    djn.a imp,  #0

John and Rashnok discovered the same 3 line imp and used a loop to reduce the size. Ilmari discovered a 2 line imp. Roy wrote a program to find 58 solutions by brute force!

Full details are on Neo's page for the Harmless Overrun Mini Challenge.

Thoughts on Corewar...

2008-11-09T08:20:00.003Z

Thoughts on Corewar is John Metcalf's Redcode programming blog, hosted at http://impomatic.blogspot.com. Since John's blog was last reviewed, both the name and design have changed.

The standard template has gone and the sidebar now contains a selection of useful rss feeds, including the latest posts from rec.games.corewar and the most recent additions to the infinite nano and tiny hills.

The most interesting recent posts have been about Semaphores in Redcode, which provide a means of implementing mutual exclusing in Corewar.

What I liked:

Covers advanced programming topics
The feeds make it easy to keep up-to-date

Room for improvement:

More comments on the code would be handy

tiny Twilight

2008-10-31T23:49:00.004Z

Here's the code for Twilight, my most successful tiny warrior so far. When this entered SAL's tiny hill in 24th place two years ago, I didn't expect it to last long!

I'm surprised to see it's still on the tiny hill age 79. Currently it's in 10th place, the best position to date :)

;redcode-tiny
;name Twilight
;author S.Fernandes
;strategy oneshot v1

;assert CORESIZE == 800

; entered the hill on 12 Mar 2006 in 24th place
; still on the hill on 31 Oct 2008, age 79, 10th place

        org    scan
step    equ    24
gap     equ    6
first   equ    loop+step*5

ptr     dat    first+gap-1     ,  first-1

        dat    0     ,  0

x       dat    1     ,  9
clr     spl    #350  ,  12
        mov    *x    ,  >ptr
        mov    *x    ,  >ptr
        djn.f  clr+1 ,  }clr

        for    2
        dat    0     ,  0
        rof

loop    add    inc   ,  ptr
scan    sne    *ptr  ,  @ptr
        djn.f  loop  ,  *ptr
        djn.f  clr   ,  ptr
inc     dat    step  ,  step
        end

Neo's Mini Challenge - The Results

2008-10-22T05:18:00.003+01:00

Neo's Mini Challenge is over and the results are in. Seven players entered, 6 of who discovered an optimal solution - 1 line, 1 process, 100% ties.

Neo and I entered the following code:

mov }0, }1

Ilmari and John entered this:

mov.i #1, *1

While Roy and Rashnok used a variation on the above:

mov.i #1, }1

Finally, mjp entered code which expands to a 6 line imp:

mov 2, @1
djn -1, #6
mov 0, >6

Roy suggested an interesting alternative which Neo will hopefully adopt as the next challenge.

Neo's Mini Challenge

2008-10-19T09:08:00.004+01:00

After a break of four years, Neo has relaunched the Mini Challenges, starting with Harmless Overrun. To take part, you need to write an imp capable of passing through an imp gate:

gate: jmp 0,<-5
end gate

The aim is to tie 100% of the time with the shortest possible code. Good luck to everyone taking part.

Miz's Corewar Page

2008-07-12T22:56:00.004+01:00

A few days ago Mizcu moved his webpage to http://users.evtek.fi/~mikaos/corewar/, so now seems like an ideal opportunity to take a closer look.

The highlight of Mizcu's site is Monowire's story, the tale of a beginner's warrior. Mizcu details the development of his oneshot from a beginner warrior to king of the KOTH.org '94nop hill.

Also in the Redcode section, Mizcu publishes some of his Corewar projects, including a generic table scan, Stoney the second and Lady Imperactive. Huffpaper includes an interesting variation to the standard silk concept.

Finally, there's an article about decoys which Mizcu is working on. This covers the basics, active decoys, invisibility and will eventually include decoy makers.

What I liked:

Monowire's story

Room for improvement:

A few comments would make the code easier to understand

Will Varfar's Corewar Page

2008-07-02T18:34:00.000+01:00

Will Varfar has a webpage devoted to Corewar at http://redcoder.sourceforge.net and shares some of his experiences implementing and testing Redcode simulators and evolvers.

Will's Which MARS report gives a thorough examination of different simulators, including some observations on their inner workings. The report concentrates on 6 open source MARS. It would be interesting to see the closed source simulators included in the comparison.

Species is Will's Redcode evolver, which attempts to improve on past evolvers with the introduction of species and chromosomes. Source code and a Windows .exe are available to download.

Some interesting statistics are published on Will's page. Will has gathered data about the frequency of opcode execution during a round robin of over 400k battles. Taking this one step further, graphs have been generated showing how the opcode frequency changes during a battle.

This data has been used to automatically divide the warriors into groups based on their execution fingerprint. It might be interesting to compare the results to those of Joonas's self organising maps.

Finally, there are a couple of article in which Will discusses the development of his entries for Neo's Mini Challenge #3 and Redcoder's Frenzy round 9.

What I liked:

In my opinion, Will has the most visually attractive Corewar site
100% original content

Room for improvement:

the 3d graph of Optimax benchmark stats is a bit tricky to interpret

Welcome sh0newars

2008-06-26T21:16:00.004+01:00

A few days ago N. Ristic aka sh0ne joined the Corewar enthusiasts in the #corewars IRC channel and not only started learning the ropes, but created a website which will hopefully document his rise to Corewar fame. sh0ne's page is at http://sites.google.com/site/sh0newars and is definitely worth keeping an eye on over the next few weeks.

Łukasz Grabuń's Book of Stones

2008-06-17T12:20:00.000+01:00

Łukasz Grabuń has a small Corewar site at http://grabun.com/corewar. Apart from a selection of useful links, Łukasz's site contains the Book of Stones which he describes as "a modest continuation to the second chapter of Steven Morell's book".

The first chapter has an analysis of Carbonite by Ian Sutton. After describing Carbonite's bombing run and the switch to a clear, the interaction with an opponent D-Clear is examined in detail. Unfortunately, the advantage of the remote bomb and using add.ab {0,}0 in preference to add.ab -1,-1 are not mentioned.

Behemot by Michal Janeczek is discussed in the second (and final) chapter. The article begins by introducing the airBag and incendiary bomb techniques before moving on to Behemot's inner workings and interaction with dat bombers.

What I liked:

The Book of Stones

Room for improvement:

The chapter about Carbonite doesn't mention a couple of important points

Tales from the Digital Wilderness

2008-06-12T12:02:00.002+01:00

Tales from the Digital Wilderness is John Metcalf's Redcode programming blog, hosted at http://impomatic.blogspot.com. Emphasis is given to implementing real algorithms in Redcode such as sorting and seaching.

John uses a standard Blogger template, with a post archive and a selection of useful Corewar links in the sidebar. Usually there are 3 to 4 posts per month.

There are two Corewar related posts. Parallel Processes describes a useful pMARS macro and An Improved '88 Quick-scanner shows how Paul Kline's technique can be used to create a 48 instruction '88 qscan which scans 36 locations. The remainder of the blog is devoted to algorithms.

Some posts I found particularly interesting are Comb sort, Stooge sort and Gnome sort which I hadn't previously heard of. Apart from sorting, there are implementations of seaching and number theory algorithms. A post comparing the performance of the different techniques would be an interesting addition.

What I liked:

A unique concept
The code is well written

Room for improvement:

For the more complex code, a detailed description would be nice
A few comments would make the code easier to understand

Philip Thorne's Corewar Introduction

2008-06-08T13:26:00.003+01:00

Philip Thorne's Corewar Introduction can be found at http://home.pipeline.com/~corewar and has been reporting the latest tournament news since 2002. The homepage has a handy selection of links for Corewar beginners.

What really make's Phil's site stand out is the unique content. There are two useful tools, RSS feeds for the KOTH hills and a Redcode syntax definition for VIM. The RSS feeds are updated daily.

Phil's Stone Analysis article has a thorough examination of 13 stones, comparing everything from speed and coverage to the type of core-clear. Unfortunately, I didn't understand all of the figures in the table - a key would be helpful.

Finally, Phil has published a selection of warriors. The layout here is almost perfect! The warriors are presented in an interactive tree structure and the warrior files are displayed with syntax highlighting.

What I liked:

The Stone Analysis article
The presentation of warriors

Room for improvement:

A key to help interpret the Stone Analysis table

Neo's World of Corewar

2008-06-04T19:35:00.007+01:00

Germán Labarga completely overhauled his website a few months back with a new URL, new design and updated content. The new location is http://labarga.atspace.com. The new layout is clutter free and easy to navigate.

Neo used to pose a programming challenge in Redcode every couple of weeks, and the results from the Mini Challenges are hosted here. There's also a selection of Neo's published warriors. This section could be improved a little by including some short comments.

The highlight of the site has to be the article documenting P-Key's Development. P-Key is an adaptive switcher which won the Smart Switching tournament round.

What I liked:

The results from the Mini Challenges
The article about P-Key

Room for improvement:

Some brief comments on the warriors page would be nice
The tools page needs more links (it currently has only one)

ripples in space-time

2007-09-11T14:47:00.000+01:00

While looking through the breakdown of nano warriors by strategy I
noticed the absence of a decent 4 line paper, despite the success of
5-liners like Fluffy's Black Sun III. I quickly hacked something up
and added a quick-bomb to the beginning.

I tested with some hand-picked constants against Metcalf's nano
benchmark. It scored 139.2, or 129.8 with the quick-bomb removed.


;redcode-nano
;name hand coded paper
;author S.Fernandes
;strategy paper
;assert CORESIZE == 80

pstep   equ     15
ipos    equ     40
cpos    equ     -5
dpos    equ     35

        mov     >60         ,    10
paper   spl     #0          ,    <pstep
        mov     }paper      ,    >paper
        mov     >ipos       ,    <cpos
        djn.f   *pstep-2    ,    <dpos
        end

Optimization

In the first round of optimization 400 warriors are generated with
all constants randomized. A small script is run over the source
below replacing each exclamation mark with a random number. The
quick-bomb will be disabled until the final stage.


        org     paper

pstep   equ     (4+!%73) ; range 4-76
ipos    equ     (2+!%70) ; range 2-71
cpos    equ     (8+!%70) ; range 8-77
dpos    equ     (7+!%70) ; range 7-76

        mov     >27         ,    54
paper   spl     #0          ,    <pstep
        mov     }paper      ,    >paper
        mov     >ipos       ,    <cpos
        djn.f   *pstep-2    ,    <dpos
        end

I've use modulo to restrict the range of random values to prevent
self-bombing. This reduces the search space by approximately 40%.
While this may stop some interesting self-mutation or interaction
being found, it's really the author's job not the optimizer's to
discover these.

The peak score is 139.5. Now we take a closer look at the top 10
results to compare the paper step. -17 occurs four times, -18 twice
and -15, -22, -25 and 33 once.

For stage 2 we generate 100 warriors for each of the above paper
steps with all other constants randomized. The peak score is 145.8
with pstep -22 followed by 145.0 with pstep -17.

The top two warriors from stage 2 are taken and their cpos
optimized in stage 3. All 70 values are tested for each warrior.
The peak score for the warrior with pstep -22 remains at
145.8. The score for the warrior with pstep -17 improves
to 145.7.

In stage 4 we take the top warrior for each pstep from
stage 3 and optimize dpos. The scores increase to 146.2
(pstep -22) and 145.8 (pstep -17).

In stage 5 we optimize ipos. The scores go up to 146.8
(pstep -22) and 149.2 (pstep -17).

Finally the quick-bomb is enabled and its two constants optimized.
400 random pair of quick-bomb constants are generated. The final
benchmark scores are 158.3 (pstep -22) and 159.2
(pstep -17).

Submission to SAL's Nano Hill

Although the benchmark scores are impressive, the hill contains 50
completely different warriors. If the benchmark reflects the hill
well, the scores should be within 10%.

The paper with pstep -22 scores 146.2 and takes 6th
place. With pstep -17 however the score is 152.3 and 4th
place.


;redcode-nano
;name ripples in space-time
;author S.Fernandes
;strategy paper
;assert CORESIZE == 80

pstep   equ     -17
ipos    equ     10
cpos    equ     -33
dpos    equ     36

        mov     >-39        ,    25
paper   spl     #0          ,    <pstep
        mov     }paper      ,    >paper
        mov     >ipos       ,    <cpos
        djn.f   *pstep-2    ,    <dpos
        end


;redcode-nano
;name ripples 22
;author S.Fernandes
;strategy paper
;assert CORESIZE == 80

pstep   equ     -22
ipos    equ     -34
cpos    equ     17
dpos    equ     -29

        mov     >-38        ,    25
paper   spl     #0          ,    <pstep
        mov     }paper      ,    >paper
        mov     >ipos       ,    <cpos
        djn.f   *pstep-2    ,    <dpos
        end

nanoWarrior #02 (finally!)

2007-07-19T19:00:00.000+01:00

Finally, the latest issue of nanoWarrior is complete! Check it out for yourself here:

http://corewar.co.uk/nw02.txt

:: sf ghoul :: corewar

Maezumo: The Corewar Evolver with a Twist

A New Arms Race, the Tiny Evolvers

April Update: What's New in Corewar?

March Update: What's New in Corewar?

nano papers: ghoulish and Apis Mellifera

Redcode Forth

January Update: What's New in Corewar?

The DynaHill at Corewar.Info

December Corewar Updates

Corewar: Hostile Programming

nanoWarrior #03

Neo's Mini Challenge - Roy's Idea

Thoughts on Corewar...

tiny Twilight

Neo's Mini Challenge - The Results

Neo's Mini Challenge

Miz's Corewar Page

Will Varfar's Corewar Page

Welcome sh0newars

Łukasz Grabuń's Book of Stones

Tales from the Digital Wilderness

Philip Thorne's Corewar Introduction

Neo's World of Corewar

ripples in space-time

Optimization

Submission to SAL's Nano Hill

top 10 links for Corewar Newbies

nanoWarrior #02 (finally!)