After a brief introduction, each chapter presents warriors by subject. I then pontificate on the merits of these various warriors and give some hints for successful implementation. I mention credits and give references to other warriors worth further investigation. Unless otherwise indicated, these warriors are archived in warrior10.tar in the redcode/ directory.
The presentation of each warrior follows roughly the same format. First, the parameters of the warrior are given. These include the name, author, attack speed, effective size, durability, and effectiveness, and score against the Pizza Hill. The effective size is the size of the executing code during the attack phase, taking into account regenerative code. Next, self-contained source code is given, followed by a brief description of the warrior. Finally, a detailed technical description of how the warrior runs is given.
I hope that this helps. If you have questions or comments, send them to morrell@math.utah.edu, where you can reach me until June,1994.
Steven Morrell
This chapter deals with imps, from the basic imp proposed by A.K.Dewdney in the original Scientific American articles to the modern-day imp-spiral we see as a component of many successful warriors.
Name: Wait Speed: None Size: 1 Durability: Strong Effectiveness: None Score: wait JMP wait end waitWait is the simplest warrior. Its small size makes it difficult to locate. However, it has no attack, so it only wins if the enemy program self-destructs. We shall be using this program for fodder.
Name: Imp Author: A.K.Dewdney Speed: 100% of c (sequential) Size: 1 Durability: Strong Effectiveness: Poor Score: imp MOV imp, imp+1 end impImp presents the enemy with a small, moving target that will not die without a direct hit. It ties a lot, and is vulnerable to the imp-gate. (See program 3)
HOW IT WORKS: When Imp is loaded and before it executes, it looks like this:
MOV 0,1 (1)(The (1) shows which instruction will execute on the first cycle.) When process (1) executes, it first copies its instruction to the next address and then moves to the next instruction:
MOV 0,1 ;This is the original. MOV 0,1 (2) ;This is the copy.Process (2) now executes. Since all addressing is relative, the process copies its instruction to the next address and advances.
MOV 0,1 MOV 0,1 MOV 0,1 (3) ;This is the second copy.And so it goes, overwriting anything in its path with MOV 0,1 instructions. So when it encounters enemy code, it replaces the enemy code with MOV 0,1 instructions, turning the enemy processes into imps. Note that although the enemy code is gone, the enemy processes live on, so imps do not win unless the enemy code self-destructs.
Name: Imp Gate Speed: None Size: 1 Durability: Strong Effectiveness: Excellent against imps, Extremely Poor against others Score: gate equ wait-10 wait JMP wait,<gate end waitImp Gate waits and destroys imps that happen to pass 10 instructions before it. It is seldom overrun by imps and its small size makes it difficult to locate. The imp gate is defensive by nature, and will not win against a stationary enemy unless this enemy self-destructs.
HOW IT WORKS: The process running at _wait_ jumps to the A-value of this command, i.e. back to _wait_. However, it also decrements the B-field of _gate_. Thus, the B-field of _gate_ is decremented every turn. When an enemy imp comes by this is what happens:
MOV 0,1 (x) ;here comes the imp DAT 0,-5 ;here is the gate The imp copies itself and advances onto the gate: MOV 0,1 MOV 0,1 (x+1) ;here is the gateThe gate decrements:
MOV 0,1 MOV 0,0 (x+1) ;here is the gateThe imp copies this instruction to itself (effectively doing nothing) and advances, falling off the end:
MOV 0,1 MOV 0,0 ;here is the gate (x+2)The gate decrements again (but the damage has already been done.)
MOV 0,1 MOV 0,-1 ;here is the gate (x+2)The enemy process executes an illegal instruction and dies.
Name: Worm Speed: 25% of c (linear) Size: 1.75 Durability: Very Strong Effectiveness: Poor Score: launch SPL b SPL ab aa JMP imp ab JMP imp+1 b SPL bb ba JMP imp+2 bb JMP imp+3 imp MOV imp,imp+1 end launchWorm is a symbiotic collection of imps. The only vulnerable parts of the worm is the tail instruction and the instruction about to execute, hence the effective size of 1.75 (25% of the time, the tail instruction is the instruction about to execute.) It is very difficult to kill, because each imp must be disposed of individually. However, it is still vulnerable to imp gates. As with Imp, Worm overwrites enemy code but preserves enemy processes.
HOW IT WORKS: First, we launch the worm using a binary launch:
SPL 4,0 (1) SPL 2,0 JMP 5,0 JMP 5,0 SPL 2,0 JMP 4,0 JMP 4,0 MOV 0,1The first process splits into processes (2) and (3):
SPL 4,0 SPL 2,0 (2) JMP 5,0 JMP 5,0 SPL 2,0 (3) JMP 4,0 JMP 4,0 MOV 0,1Process (2) splits into processes (4) and (5):
SPL 4,0 SPL 2,0 JMP 5,0 (4) JMP 5,0 (5) SPL 2,0 (3) JMP 4,0 JMP 4,0 MOV 0,1Process (3) splits:
SPL 4,0 SPL 2,0 JMP 5,0 (4) JMP 5,0 (5) SPL 2,0 JMP 4,0 (6) JMP 4,0 (7) MOV 0,1Process (4) jumps:
SPL 4,0 SPL 2,0 JMP 5,0 JMP 5,0 (5) SPL 2,0 JMP 4,0 (6) JMP 4,0 (7) MOV 0,1 (8)Processes (5), (6) and (7) jump:
SPL 4,0 SPL 2,0 JMP 5,0 JMP 5,0 SPL 2,0 JMP 4,0 JMP 4,0 MOV 0,1 (8) (9) (10) (11)The worm will now start crawling though memory. Note that if processes (9), (10) or (11) executed right now, they would execute an illegal instruction and die. But process (8) executes, copying the MOV instruction to where process (9) is going to execute:
SPL 4,0 SPL 2,0 JMP 5,0 JMP 5,0 SPL 2,0 JMP 4,0 JMP 4,0 MOV 0,1 MOV 0,1 (9) (12) (10) (11)Now process (9) executes, copying the MOV instruction to process (10).
SPL 4,0 SPL 2,0 JMP 5,0 JMP 5,0 SPL 2,0 JMP 4,0 JMP 4,0 MOV 0,1 MOV 0,1 (12) MOV 0,1 (10) (13) (11)And after (10) and (11) have executed, the worm has crawled forward an instruction, leaving a slimy MOV 0,1 trail behind.
SPL 4,0 SPL 2,0 JMP 5,0 JMP 5,0 SPL 2,0 JMP 4,0 JMP 4,0 MOV 0,1 MOV 0,1 (12) MOV 0,1 (13) MOV 0,1 (14) MOV 0,1 (15)
Name: Ring Speed: 100% of c (mostly linear) Size: 1 Durability: Average Effectiveness: Fair Score: c JMP imp-2666 launch SPL c SPL imp+2667 imp MOV 0,2667 end launchRing is a symbiotic collection of three imps distributed through core. It has the capability to destroy enemy processes it overruns, if the enemy is running only one or two processes. This code will run correctly only in a coresize of 8000, although the constants may be tweaked to run in any coresize not divisible by 3. Ring is an example of a 3-pt imp.
HOW IT WORKS: The launching code is a very small binary startup:
JMP -2663, 0 SPL 0, 0 (1) SPL 2668, 0 MOV 0,2667The first process splits:
JMP -2663, 0 (3) SPL 0, 0 SPL 2668, 0 (2) MOV 0,2667The second process splits:
JMP -2663, 0 (3) SPL 0, 0 SPL 2668, 0 MOV 0,2667 (4) ... (5) ;this location is 2667 instructions after the impThe third process jumps:
MOV 0,2667 (4) ... (5) ;this location is 2667 instructions after the imp ... (6) ;this location is 2667 instructions after process (2)Now the fun begins. Process (4) executes, copying the imp instruction to process (5) and becoming process (7):
MOV 0,2667 (7) ... MOV 0,2667 (5) ... (6)(5) executes, copying the imp instruction to process (6):
MOV 0,2667 (7) ... MOV 0,2667 (8) ... MOV 0,2667 (6)And now (6) executes, copying the imp instruction back to process (7):
MOV 0,2667 MOV 0,2667 (7) ... MOV 0,2667 (8) ... MOV 0,2667 (9)The cycle starts all over again, and the ring creeps forward.
Let's see what happens when Ring fights Wait (Program 1). Wait executes JMP 0,0 until eventually Ring overwrites this instruction with MOV 0,2667.
MOV 0,2667 (1)Wait executes this instruction and advances:
MOV 0,2667 (2)Since Ring takes 3 cycles to move the next command into place, Wait's process now executes an illegal instruction and dies.
So Ring slowly advances through core, and if the enemy is running a single process, it falls off the end of the imp ring.
Name: Spiral Speed: 37.5% of c (mostly linear) Size: 1.875 Durability: Very Strong Effectiveness: Fair Score: step equ 2667 launch SPL 8 SPL 4 SPL 2 JMP imp JMP imp+step SPL 2 JMP imp+(step*2) JMP imp+(step*3) SPL 4 SPL 2 JMP imp+(step*4) JMP imp+(step*5) SPL 2 JMP imp+(step*6) JMP imp+(step*7) imp MOV 0,step end launchSpiral crosses the durability of a worm with the effectiveness of a ring. Spiral is resistant to most conventional attacks, and since it is an 8-process imp-ring, it kills any enemy it overwrites if the enemy has less than 8 processes running. The only vulnerable parts of the spiral are the tail and the process that is currently running. Spiral is vulnerable to imp gates, however.
HOW IT WORKS: After a binary launch, the processes are arranged as follows:
MOV 0,2667 (16) (19) ;this process is 2667 instructions after process (18) (22) ... (17) ;this process is 2667 instructions after process (16) (20) (23) ... (18) ;this process is 2667 instructions after process (17) (21)Now the spiral worms along: (16) copies the imp to (17), which copies it to (18), and so on. All the processes advance 1 instruction as this happens, and then the imp-passing instructions begin again.
A step-by step analysis of how imp gates destroy spirals would be lengthy and unnecessarily complicated. The key idea is this: The imp gate is constantly being modified. As the imp overruns the imp gate, no imp instructions are left intact to copy to the next processes' location. This next process executes an illegal instruction and dies. This scenario repeats until the entire spiral moves through the imp gate and disintegrates.
Name: Gate Crashing Spiral Speed: 12.5% of c (mostly linear) Size: 5.875 Durability: Very Strong Effectiveness: Good Score: step1 equ 2667 step2 equ 2668 start SPL lnch1 SPL lnch3 lnch2 SPL 8 SPL 4 SPL 2 JMP imp2+(step2*0) JMP imp2+(step2*1) SPL 2 JMP imp2+(step2*2) JMP imp2+(step2*3) SPL 4 SPL 2 JMP imp2+(step2*4) JMP imp2+(step2*5) SPL 2 JMP imp2+(step2*6) JMP imp2+(step2*7) lnch3 SPL 8 SPL 4 SPL 2 JMP imp3+(step2*0) JMP imp3+(step2*1) SPL 2 JMP imp3+(step2*2) JMP imp3+(step2*3) SPL 4 SPL 2 JMP imp3+(step2*4) JMP imp3+(step2*5) SPL 2 JMP imp3+(step2*6) JMP imp3+(step2*7) lnch1 SPL 8 SPL 4 SPL 2 JMP imp1+(step1*0) JMP imp1+(step1*1) SPL 2 JMP imp1+(step1*2) JMP imp1+(step1*3) SPL 4 SPL 2 JMP imp1+(step1*4) JMP imp1+(step1*5) SPL 2 JMP imp1+(step1*6) JMP imp1+(step1*7) imp1 MOV 0,step1 DAT #0 DAT #0 DAT #0 imp2 MOV 0,step2 MOV 0,step2 imp3 MOV 0,step2 MOV 0,step2 end startGate Crashing Spiral is a collection of three spirals that work together to kill imp gates. The first is a standard imp spiral and the other two are slightly modified, interleaved for greater protection against split bombs. The large size of its launch code makes it vulnerable to fast attacks.
HOW IT WORKS: Each spiral has its own binary launch. The first spiral launches first and crawls forward an instruction by the time the other two spirals have launched. Core then looks like this (after resetting the counter for clearer exposition):
MOV 0,2667 ;This is label imp1 | MOV 0,2667 | MOV 0,2667 MOV 0,2667 (17) | MOV 0,2667 (18) | MOV 0,2667 (19) MOV 0,2667 (20) | MOV 0,2667 (21) | MOV 0,2667 (22) DAT #0,#0 (23) | (24) | MOV 0,2668 (1) ;This is label imp2 | | MOV 0,2668 | (2) | MOV 0,2668 (9) ;This is label imp3 | | (3) MOV 0,2668 | (10) | (4) | | (11) | (5) | (12) | | (6) | (13) | (7) | | (14) | (8) | (15) | | | (16) |The imps then move forward via the usual instruction juggling.
When a gate crashing spiral overruns a gate, the second or third spirals hit first:
MOV 0,2668 (x) ;imp gate hereThe gate decrements:
MOV 0,2667 (x)The wounded spiral copies this instruction 2667 ahead:
MOV 0,2667 (x+24) ... MOV 0,2667The second and third spirals now fall off the end and die, and then the first spiral hits the gate:
MOV 0,2667 (y) ;imp gate here ... MOV 0,2667 (y+1)The gate decrements:
MOV 0,2666 (y) ... MOV 0,2667 (y+1)Process (y) executes, and can't copy the imp to process (y+1), but this is okay, because process (y+1) executes the imp instruction from the two spirals gone before. The spiral crawls through the gate and goes on to kill the enemy processes.
Name: Nimbus Spiral Speed: 50% of c (somewhat linear) Size: 1.992 Durability: Very Strong Effectiveness: Fair Score: step equ 127 imp MOV 0,step launch SPL 1 ;1 process SPL 1 ;2 processes SPL 1 ;4 processes SPL 1 ;8 processes SPL 1 ;16 processes MOV -1,0 ;32 processes SPL 1 ;63 processes SPL 2 ;126 processes spread JMP @spread,imp ADD #step,spread end launchNimbus Spiral launches a 63-point spiral with two processes per point. Because a binary launch would exceed the 100-instruction limit, Nimbus Spiral uses what is called a Nimbus-type launch. The code for this type of launch is obviously smaller, but the time it takes to launch spirals is roughly doubled.
HOW IT WORKS: Each SPL 1 command doubles the number of processes acting in tandem at the next instruction. The first process that executes the MOV -1,0 command does not split, but all subsequent processes execute a SPL 1 command. Hence, before execution of the SPL 2 command, core looks like this (with counter reset):
MOV 0,127 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 2,0 (1)-(126) JMP @0,-9 ADD #127,-1After execution of the SPL 2 command:
MOV 0,127 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 2,0 JMP @0,-9 Odd processes ADD #127,-1 Even processesWe reset the processes again. Process (1) now executes, jumping to the location of the B-operand of the JMP instruction:
MOV 0,127 (253) ;this came from process (1) SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 2,0 JMP @0,-9 Odd processes greater than 1 ADD #127,-1 Even processesProcess (2) now executes, adding 127 to the B-operand of the JMP instruction:
MOV 0,127 (253) SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 2,0 JMP @0,118 Odd processes greater than 1 ADD #127,-1 Even processes greater than 2 (254) ;this came from process (2)And it continues. Process (3) jumps to a new location. The even processes modify the jump vector, and the odd processes do all of the jumping. By the time process (127) is ready to execute, we have the following situation:
MOV 0,127 (253) SPL 1,0 (379) SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 1,0 SPL 2,0 JMP @0,-134 ADD #127,-1 Even processes ... Odd processes broadcast throughout coreThe odd processes form an imp spiral and the even processes execute illegal instructions and die, leaving just the spiral to crawl through memory.
Most of today's fighters have some resistance to imps, so pure imp programs seldom are successful. But imps are easy to add to code that has multiple processes running, like today's stones, vampires, or paper. The most successful imp warriors use most of their process time in a more conventional attack, and rely on the imp-ring as a backup. Whether an imp is a good idea in your program depends on the program; you may lose less, but you may win less. About the only thing you can be sure of is tying more. But testing your warrior always helps.
Killing imps is difficult, but not impossible. Imp gates work well against most imps, but should only be executed after the rest of your code has done its stuff. Imp gates of the form
SPL 0,<gate DAT <gate,<gatecan sometimes kill even gate-crashing imps. Fast bombing programs can occasionally catch the launching code before it has completed, especially with fancier imps. Code with a long enough bombing run (e.g. Charon v8.1) can hit and destroy all the imp instructions if it is done right. Dropping a single MOV 0,<1 bomb on the tail (or vulnerable instruction soon after the tail) of an imp-ring will kill the entire ring off. Dropping a MOV <2667,<5334 instruction on a 3-point imp ring can kill as many as 9 imp instructions, and is extremely effective in a stream (which is sequential bombing of memory). Some programs use an imp trap tailor-made for stunning imp-rings by dropping SPL 0 bombs on the imp-ring using a step size of 2667, so that the ring is attacked from the tail forward.
An enhancement to the imp-launching routines is to add decrement statements to all the b-fields of the SPL and JMP commands. If you have a large binary launch, for example, you could decrement 63 instructions throughout core for free. Most of the original code I have based this chapter on has such b-fields.
Here is a list of imp-style programs worth investigating. Unless otherwise noted, they can be found in warrior10.tar in the 88 directory. Imp-stone combos will be listed in the back of chapter 2.
"The IMPire strikes back" by Anders Ivner (impire)
"Trident" by Anders
Ivner (trident)
"Nimbus 1.2" by Alex MacAulay (nimbus12)
"Imps! Imps!
Imps!" by Steven Morrell (contact morrell@math.utah.edu)
Program 3, Imp Gate, was suggested in its current form by B.Thomsen, and is often called a wimp in the literature.
Program 5, Ring, was stolen and modified from a _Push Off_ article from P.Kline, but it looks suspiciously like A.Ivner's "Trident."
Program 7, Gate Crashing Spiral, was stolen and modified from P.Kline's "Cannonade."
Program 8, Nimbus Spiral, was stolen and modified from A.MacAulay's "Nimbus 1.2."