Best X11 Graphics: David Applegate Guy Jacobson School of Computer Science AT&T Bell Laboratories Carnegie Mellon University 600 Mountain Avenue Pittsburgh PA 15213 Murray Hill NJ 07974 USA USA Judges' comments: Make and run as follows: make davidguy davidguy ip_address:server.screen where 'ip_address' in an IP address of an X server. For example, try: davidguy 127.0.0.1:0.0 Of course, may need to supply a more reasonable address. :-) Also try: cp davidguy guydavid guydavid 127.0.0.1:0.0 Can you determine why this makes a difference. We permitted this type of entry to win because: - it did not require X to be on the host system - both BSD and SVR4 have sockets Now that X11 and sockets have become 'standard', we plan to open up the contest to programs that use them. Future rules will explain this further. We are pleased that this entry has helped bring new 'life' into the contest. NOTE: The orignal program that won may be found in davidguy.orig.c. The author supplied a second version of the program, which we have placed in the default location davidguy.c. Selected notes from the authors: The program is a fully-functional X client. It talks directly to the X server through a socket without using Xlib, Xt or any other wussie toolkit junk. With no arguments, it will try to run locally using the unix-domain socket (only on BSD :-( ) or you can give it a command line argument identifying a display host address (in dotted octet notation, like 127.0.0.1:0.0) where the X server is running. It handles color and monochrome displays, and performs whatever byte-swapping and bit-reversal is needed (so that the program works even when the server and the client have different architectures). It buffers communication to the X server for efficiency. [[ NOTE: The original source in davidguy.c used only an IP address argument (assumed :0.0) and assumed the color of black and white. ]] The program doesn't require any user interaction; you can just sit back and watch it run. All the action takes place in the root window of the display. We think the program is interesting for these reasons: 1) It belongs to a class of programs that programmers make the following complaints about: they require a mountain of code to implement even the simplest program; and they must be linked with giant libraries of hard-to-understand functions. In contrast to these complaints (made by wimps, in our humble opinion), our program fits in under 1536 bytes, doesn't need any external functions except for a few system calls, and doesn't need to include any header files. 2) It implements an extremely efficient algorithm for the computation it performs. One cool feature of the algorithm (at least as far as this contest is concerned) is that its obscurity is fundamentally necessary for its efficiency. In other words, even if we took steps to present the algorithm clearly, (we haven't) the relation of the computation performed to the specification of what the program was supposed to do would still be hard to understand. Quasi-spoiler on internals: The program plays Conway's game of Life in the root window's background pixmap. It starts by setting the background to random bits, and then plays Life, with one Life cell for each pixel of the screen. In the game of Life, a cell survives to the next round if 2 or 3 of its 8 neighbors are also alive, and a cell is born if exactly 3 of its neighbors are alive. Otherwise, a cell dies of exposure if it has only 0 or 1 neighbors, and dies of overcrowding if it has 4 or more neighbors. The algorithm used to compute the next generation is based on the observation that a cell's state in the next generation is a boolean function of its current state, and the states of its 8 neighbors (ie, a boolean function from 9 bits to 1 bit). This function can be computed by a boolean circuit. In addition, intermediate values computed by the circuit can be shared between neighboring cells, reducing the number of gates per cell required. These ideas have been used before, to compute the next generation through a series of bit blits. Instead of doing this, we map values in the circuit to bits in registers, so that the next generation can be computed efficiently within registers, minimizing memory accesses. As a result, the computation of the next generation is performed with about 1.6 instructions per life cell, consisting of .125 memory accesses, .17 shifts, and 1.3 logic operations. The net result is that the time to transfer the bits to the X server, and for the X server to draw them on the screen, dominates the time to compute the next generation. Copyright (c) 1991, Landon Curt Noll & Larry Bassel. All Rights Reserved. Permission for personal, educational or non-profit use is granted provided this this copyright and notice are included in its entirety and remains unaltered. All other uses must receive prior permission in writing from both Landon Curt Noll and Larry Bassel.