Super Pac-Man

Some memories of my stint at Atari, brought to the surface by a cow-orker who, it turns out, did a lot of work on some Atari platforms in the mid and late 80s, and we got to talking a couple days ago…

The 400/800 Super Pac-Man Cartridge

The reason that Super Pac-Man is the best arcade conversion that I ever did is that nobody really cared about it. I knew from the very first that it was a “loser” title, an oddball member of the Pac-Man pantheon that no one really understood, and the likelyhood of the cartridge actually shipping was pretty low. During development it was an orphan in marketing, and I swear there were at least three back-to-back Mondays when a new marketroid appeared in my office, said “I’m responsible for this game, tell me about it,” then I’d never see them again. Since no one truly cared, I got a good solid nine months of development time on the game — when most titles were being shoved out the door, ready or not — with around six calendar months, and I think that the extra effort shows.

How it happened: After shipping Donkey Kong I wound up with a terrible manager. I was itching to work on another title as hot as ‘Kong had been, but everyone else seemed to be getting the interesting work. Weeks went by while I twiddled my thumbs [I wasn’t goofing off, I was reading some CS texts, mucking with the shiny new Vaxes our group had been given, and writing some development tools, but still…]. Finally I was told to look at “Kangaroo” and “Arabian,” two games that I thought were really bad. I wrote some nasty memos about how miserable those two games were.

Finally my boss said, “Okay, work on this Super Pac-Man thing.” It sure looked better than Kangaroo, or (shudder) Arabian.

The department bought an arcade machine that I could play, and people pretty much forgot about me. Nobody cared. Apparently someone in marketing would draw the occasional short straw and show up in my office, then they’d have their heads chopped off or something because I never got another visit. Atari was pretty turbulent back then. It was easy to get lost in the shuffles.

On Conversions

You’d think that the conversions group would have help from the companies who wrote the arcade games we were reimplementing — listings, design documents, the opportunity to talk to the original engineers, that kind of thing — but in fact Atari had negotiated nothing but a license to make a cartridge version of the game. Atari even had to buy their own copy of the arcade game. This was pretty classic negotiating for Atari: Negotiate a hot deal with no thought to the technical issues, then wonder why the engineering didn’t “just happen” in a week or two.

Things like listings might not have helped anyway. In the consumer games industry, commenting code was kind of a luxury; you shipped something that played okay and moved on to the next title. There was essentially no code re-use. The 400/800 Pac-Man cartridge TWO comments; I forget the first, but the second was something like “ha ha,” at the single point in the ROM’s copy protection where it stored to itself in a rather lame attempt at committing suicide.

What we did in the absence of help from the original authors of a game was: We’d play the arcade game, video tape the action and use that as a basis for the artwork and finer points of the game play. We had to get pretty good at the arcade version.

The best commented code I ever saw in the games industry was the stuff responsible for actually doing the “quarter sucking” in the arcade machines. Talk about tough! It had to deal with real-world timing, anti-theft issues, international currency, and things like the coin box getting full. If this code failed and swallowed the customers’ money, they might destroy the machine in revenge.

Machine Resource Allocation

On the 400/800 I noticed that people knee-jerked toward using the player-missile hardware way too early in the design process, and that these special resources tended to blind people to the box’s real power. It was like all the whizzy hardware was allocated in the first five minutes and creative alternatives were never considered — “Of course the thing that the user controls has to be a player object.” The result was sometimes bad game play and poor looking games.

For instance: Hardware collision detection sucks in a lot of instances, but people used it because it was there to be used. The player loses a life because one of his outlying pixels happened to hit a single enemy pixel — that’s really unforgiving, and games that used it irritated the heck out of me. You know the tension when a Pac-ghost is right on your tail and you whip around a corner and get that last dot just as the ghost is about to eat you? You can’t get that feeling with pixel-perfect collision detection, the game just bloody clobbers you, and it feels mechanical and unfair. It’s like playing a tax accountant, not a fun opponent. Games like Caverns of Mars were particularly bad; real space ship crashes are messy and probably involve big noises and fire and rolling bits of fuselage. Crashing in ‘Caverns was like having a pocket calculator say “E” at you for dividing by zero; one of your ship’s pixels overlapped one of the cavern’s pixels, whoopee, game over.

So collisions in Super Pac-Man were done with cross-shaped bounding boxes on the player and the ghosts. You could overlap the ghosts slightly, get away with it and feel like you cheated the game a little.

I used character cell animation (in graphics mode 5, and that’s probably meaningful to a disturbing number of you out there) to get an extra color in the ghosts. The “super” Pac Man was seventeen pixels across rather than sixteen simply because the extra pixel made the circle a lot prettier, so it used two 8-pixel-wide players and a single extra pixel from a missile. I think I used the remaining missiles to add even more colors to the ghosts, maybe animate the eyes. (NB: A “player” is an 8-pixel wide mono-color vertical stripe the height of the screen, and a missile is a 2-pixel wide stripe. There are 4 players and 4 missiles, and they can be repositioned horizontally on a scanline-by-scanline basis).

Other touches: People complained that the cartridge versions of games didn’t have the “cartoons” (cut scenes) from the arcade. I wrote a little bytecode engine and stuck those in. I didn’t know how the ghosts “found” the Pac-Man, so there’s an N-ply recursive search algorithm in there and the ghosts search deeper in the maze as the game gets harder. I tried to duplicate some of the ghost movement patterns, but probably didn’t succeed.

Like Donkey Kong, at one point the game was about 20K of code (for a 16K cartridge), so things had to get smaller. I went for features first, then started cutting. The sound tables were the first to go on a diet. I forget most of the crunching process, but my roommate tells me that I used to come home and proudly say “I saved 18 bytes today.”

Easter Eggs

One fine day, the marketroid du jour walked into my office and asked me if I could jazz the game up a little. “Say, could you add some easter eggs?”

“Like what?”

“I don’t know. Some secret thing that people would have to find.”

I have to admit to a complete lack of creative interest at this point. Also, there were only about 200 bytes of ROM left over, so the easter egg couldn’t be terribly extravagant. I’m afraid I utterly failed at coming up with something interesting. Then again, Atari was in the newspaper every few days with something new falling off and making loud financial noises. Most of us were pretty beaten up, morale-wise.

Copy Protection

If you didn’t do anything to prevent it, 400/800 game cartridges could be copied to disk and then just re-loaded at the same address and run from RAM. Cartridges were starting to have some simple copy protection — mostly just code that pretty blatantly scribbled to the ROM space, trying to wipe out other code and cause a crash. In Super Pac-Man I had time to be more clever. There were several levels of protection, including some “bait” code that stomped the ROM, but if you modified the bait then game would notice that and collapse in more subtle ways.

The clever stuff involved decrypting some code in an interrupt handler over several minutes, which (when finally run) changed some constants around in the ROM that in turn caused a crash a while later. It took about two weeks to get all of this working and tested. I’d gotten my start in the games industry by cracking protection on a few disk-based titles, so this was fun to do.

A fairly prolific pirate of Atari games lived in my apartment complex. Now, the way that you got a ROM burned in the Home Computer Division of Atari was that you asked the duping lab to do it, and this lab leaked unreleased games like crazy. I was unsurprised when this pirate got hold of a fairly complete copy of the game. I asked him how long it had taken to crack. “That was a hard one,” he said, “It took us nearly three days.”

The version of Super Pac-Man that’s available for download on various web sites appears to be a late copy that leaked from the lab. There was a nasty bug that I fixed pretty late in the game, but I really don’t know if the images floating around out there have the fix or not, or even how many versions actually leaked.

I heard a rumor that Super Pac-Man was actually going to go to manufacturing, then a week or two later the Tramiels bought Atari and things got really different. I never touched 400/800 development again, except to write some development tools.

More later. I’ve got some Atari ST stuff outlined . . .

Forthcoming SF

Uncle Hugo’s list of forthcoming SF. This is more accurate and updated more frequently than the list on Locus. Link.

Highlights: A new Vernor Vinge book (set in the “bobble” universe, not the “Depths of Thought” — Amazon is incorrect), Varley’s sequel to Red Thunder, a couple books by Charles Stross. Robert Charles Wilson may have a new book out (his previous one, Spin, is very good). Altogether, kind of a dry year again.

A technical book you might not want to miss: Robert P. Colwell’s The Pentium Chronicles, about the development of the Pentium VI processor. It’s not Soul of a New Machine — it’s mostly about managing a large on-the-edge engineering project — but it does have its moments. It’s basically anecdotes about the people he managed or worked with, ranging from really smart folks to the security wacko who (at 3am) told him to stop listening to music on his headphones because the noise was disturbing the people around him.

Transactions, again

(Geek warning)

Again? Maybe the first time I’ve mentioned transactions here. This is a transcript of a nice talk on the implementation of the ext3 file system. Link.

(I did a bunch of work on the Newton’s transactional object store, and this takes me way back. We did a lot of the same tricks. I especially like the compatibility with the ext2 file system, once the ext3 fs has been recovered after a crash, that’s really cute).

There’s a bunch of stuff on the implementation of the BeOS file system in the book Practical File System Design with the Be File System by Dominic Giampaolo. It’s a nice book if you’re into file systems, transactions and meta-data.


My parent’s home is next to a big cemetary in Port Townsend. It’s a nice place to take a walk. Some of the headstones are . . . interesting –


Look carefully at the dates . . . oh no! She’s in there, and she’s not dead yet!


An unambiguous sentiment from beyond. Thanks, but we’re pretty happy here.


One hopes that the Bircher campground in the sky has less trash strewn around than their current terrestrial one. One reason that I refused to stay in a KOA.

The Enemy of Anemone in an Anime

We’re playing “Can you say…?”

“Trader Joe’s”

“Drajha Jos”

“Good. How about . . . ‘Cracker’?”



(loud screech. supposed to be ‘meow’, but okay…)

“Cool! Now . . . ‘Bush’?”


“Close enough.”

Antec story

A year ago my four year old PC died, and I chose to do a near entire refresh of it. I bought a fairly mundane system with a relatively quiet case, an Antec Sonata, and things have been fine.

A few weeks there was an electrical burning smell in the office that I couldn’t track down. This morning the power supply fan on the new machine was going full tilt and the case temperature was nearly 60C. I cracked the machine open and saw that the case fan was valiantly trying to move, but was only twitching.

I was going to call Antec, but decided to try their online support. I put in a warranty request for a dead fan, and frankly didn’t expect much from the process.

Within about an hour, a support tech responded in email with a request for my home address, and twenty minutes after that he said “The fan is on the way.” No request for a copy of my receipt, no run-around, just “Here you go.”

I like that kind of customer service. A cynical person might say, “Well, they probably have to replace a lot of fans, so they’re used to it,” and they’d probably be right. I still like the way this was handled.

the parental class

World of Warcraft Buffs in the Real World

Remote Attention. Allows you to remotely sense what a toddler (up to level 24) is doing. Range: 1 yard per level.

Mystic Comprehension. Understand toddler speach (level 15 or greater) for two minutes. Does not allow you to speak in Toddler.

Ghostly Hands. Manipulate a child without actually touching it. Can be used to change up to level 8 diapers.

Chaotic Banishment. Cleans an area equal to your level in square yards, putting away toys, sweeping floors and vacuuming rugs. Does not clean liquids, does not repair broken toys or furniture.

Arj’quns Attitude Adjustment. Chance to calm an upset child, 1% per level. Cannot be used to stop crying.

Guardian of the Ancients. Summons a grandparent, permitting you to do something else for 30 minutes. Two day cooldown.