Chapter 40

STEVE — Tuesday, February 20, 1962

When I listened to the news last night, they said that the Mercury launch, which had been put on hold three times because of weather, was set for sometime after nine AM today. Unlike the first two manned Mercury flights, this one was to be at least one full earth orbit, with John Glenn as the astronaut.

Now it was close to the projected launch time. I knew that Isaac had a radio hidden in his desk, and stopped by the CDC 1604 room. As I suspected, he was listening to the broadcast from Cape Canaveral. I joined him, and eventually we heard the announcement that they had liftoff. Shortly thereafter, we heard John Glenn saying, "Oh, that view is tremendous!" as the spacecraft settled down in its orbit.

I knew I couldn't spend all day listening to Isaac's radio, and went back to my office to work on another small project that Dr. Edwards had handed me yesterday. However, I managed to get my lunch and eat it in the 1604 room, and was back more than two hours later when it was time for re-entry and recovery of the spacecraft and the astronaut. Needless to say, we were all greatly relieved when this critical stage was successfully completed.

When I got home, I called Sandy, and asked, "How much of the Mercury flight were you able to listen to?"

"Not much, although I did catch the liftoff and the recovery."

"You really want to watch the news tonight; they should have a good recap of the whole flight. Three orbits, and it looks like everything went very smoothly."

"I'll definitely do that. How's the machine coming along?"

"I saw Miguel today, and he said that they were almost done with all the tape drives and were about to rerun the final diagnostics again. He's also going to run my test cases once more."

"That's good. Let me know how things go."

The next day, Murray told me that the Reaper machine room would be fully secure by the end of the day, and that we could start running our programs on Thursday. I called Lou and Vlad into my office, and said, "Tomorrow is the big day. We'll be running the real program against a carefully selected set of input data, and I have a pretty good idea what results we should be getting. I assume you guys want to be there to watch."

"We wouldn't miss it for the world!"

Lou said, "Tell us about the input data you are going to be using, and what you expect for results."

"Wait a minute. I think Sam — Samantha — should hear this also." I called her extension, and she was with us in a few minutes. I then spent two hours going over the input data I had selected, and reminded Sam that once the program was debugged, she would be coordinating the data for the daily runs. I then suggested that they take the section of code where they had found a bug and build a set of test data and assemble all three versions: Lou's, Vlad's, and the combined approach that I had suggested.

"In addition to finding out which approach really works best, this will be your first chance to assemble and run programs on the real machine. Plan to get back with me at four this afternoon, so I can look at the data and your code."

Thursday morning, we all headed in to the Reaper room. Sam was astounded when she saw the size of the machine.

"I'm going to run the entire application first. Once that's done, I want you to assemble and run your programs while I check the output from mine. I'll give you any help you need, but you should be able to do it all pretty much on your own."

I showed Sam how to load up the cartridges in the fast tape jukebox, and explained to her how the first pass worked, in this case, copying the data from regular tapes to the cartridges. "We'll have an operator doing this eventually, but when we do this for real, most of the data will be given to us on several different kinds of media, as I explained to you before. Have you decided on the procedures for coordinating all the input?"

"Yes, I'd like to go over them with you when you have some time."

"I'll do that as soon as I get this program debugged and ready to go for real. Stick around as long as you want, but if you have other things you should be doing, there's not too much point in your watching the rest of the process."

Finally, Sam had the cartridges loaded. I said, "OK, everyone, here goes!"

I loaded the first phase of the program, set the flags to generate a debugging log file, and started to run it. It ran through all the input tapes, and apparently wrote the data to the cartridges, but then went into a tight loop instead of loading the next phase.

I killed the program, took a memory dump, and printed out my debug log.

"This is a Stretch program; there's no Reaper code yet. The log file looks good, but we should take a tape dump of one of the cartridges to confirm that the correct data was written."

I studied the source code for a moment, and showed Lou and Vlad where we were on the basis of the log data. Vlad was the first to spot the problem, and said, "Is this it? It looks like you're looping through the list of messages, but it never gets past the branch at the end of the loop. The branch should fail when we get to the end of the data."

I looked at it for a minute, and realized that I had miscoded the comparison instruction.

"Do either of you see what makes it do that?"

They thought for a moment, and referred to their Stretch manuals. After a moment, Lou cried out, "Steve, I think I have it. Isn't that Count and Branch instruction using the wrong index register?"

"That's the way I see it, Lou. Let's fix that and rerun the program. Did either of you have a chance to study the Stretch assembler manual yet, including the part about updating a source tape?"

"No, we hadn't gotten to that."

"OK, watch and I'll show you."

Explaining what I was doing as I went along, I punched up a card with the corrected instruction and the sequence number shown on the source listing, along with a couple of control cards. I then loaded my source tape and the assembler, put my three cards in the card reader, and ran the assembler in update mode. It generated a new source tape, a new object file, and a new source listing. I then ran the linker to write a new program tape with the new executable for the first phase.

"This may look complex, but it's actually fairly simple once you have done it a few times. Sarah wrote a good description of how to do it; I'll find it and make copies for you."

We reran the first phase, and it completed normally and the second phase also ran through to apparently successful completion. I paused the process before starting phase three, and printed the debug log.

"This looks almost right, although there's one set of data that should have been successfully decoded that came out as garbage. I'll look through this later, but I suspect that it's an error on my part in generating the test data. Let's go on to phase three, which, with this set of data, should run fairly quickly."

It did, and when I printed the debug log, it looked good also.

"Now we get to the hardest part. Phase four is where we apply all the cryptanalysis techniques, and also do repeated subsetting of the data for the particular source. There's some fairly complex Reaper code in here, and I'll be amazed if the bug that you guys found last week is the only one in this section."

"I'm sure neither of us fully understands the logic or the code that you use for the subsetting."

"When you have a chance, review the streaming mode again, particularly the Stream Search. I've set up a tree structure for the messages. I use it to keep track of all the subsets, and to pass over subsets that haven't changed since the last time the program was run. I'll show you the layout of the tree when we have a chance."

I started the final phase of the program, and set the debug log to print directly as the program ran so that we could observe the progress. After a while, I said, "I see some problems that I think I can fix easily, but even after that, it's still not right. This would be a good time for you to get your own programs ready to assemble and run."

I pulled out my source listing for the Reaper code, along with my notes on the tree structure. I noted two Reaper code sections that were causing most of the problems, along with the code that I thought would fix them. When the program finally completed, I turned the machine over to Lou and Vlad to assemble and run their programs. I knew that I would need several hours to study the debug log in more detail, and I also wanted to code up some additional debugging hooks that would allow me to get periodic snapshot dumps of the message tree.

SANDY — Friday, February 23, 1962

An all-day New Processor Line (NPL) architecture meeting had been scheduled for next Wednesday, and Paul asked me to attend. We were at a point where there was not much more we could do on defining the control points for the machine until most of the architecture issues were resolved, so I used the time to complete my work on the Stretch training course. I had my final meeting with their education manager this afternoon, and, after looking at what I had suggested, told me that it looked very good, thanked me for my efforts, and said that they would also tell Paul that I had done a very good job.

Feeling good about that, I went back to my office, put my stuff away, and drove to Mom and Walt's for dinner. I was still sleeping at the old house, and making my breakfasts, but by now Mom was fixing most of my evening meals. I got there, and sat down in the kitchen with Mom, who was working on dinner.

"Mom, I think that next week I should do most of the cooking. If I don't, I'm going to be completely out of practice by the time I get married."

"Don't worry about it, dear. You'll do well in the kitchen, I'm sure, and you'll have good help if I remember right about the meals that you and Steve prepared. What you need to work on is thinking and planning ahead; you need to put together a tentative meal plan for the week ahead, and then organize your shopping so that you know you have everything you'll need. After dinner, let's plan the meals for next week, and then we can do the grocery shopping together tomorrow. We can share the cooking also."

"We can do that, but I'm going to let you take care of Wednesday. I have an all-day meeting that could easily run late, and in any event, I'll still be trying to absorb everything that happened and won't be much good in the kitchen. Right now, I want to call Steve."

I went to the living room, picked up the phone, and placed the call to Steve's apartment. He answered, sounding slightly discouraged.

"What's wrong, love?"

"Nothing really, but I'm still trying to figure out why my program isn't running right."

"Dear Tigger! Please don't worry about it. Tomorrow is Saturday. Take the weekend off, get outdoors if the weather is still nice, and by Monday, you'll have a fresh approach to the problem. I'm sure you'll have it fixed in a few days."

"I hope so. How's the work on the new machine coming?"

"We're sort of stalled right now; there's a big all-day architecture meeting Wednesday that should resolve a lot of issues. I took the time to finish off that Stretch training material course that I told you about. They seemed happy with what I had done."

"That's good to hear. I'll try to call you by Wednesday and let you know how much progress I've made. I think I'm also going to have to take some time to write some better debugging tools; what we are using isn't as good as what Sarah had put in the simulator. I had made some notes about a better tool last spring, but never got back to it."

Just then Mom told me that she was about to put dinner on the table. I told Steve, "Mom told me that dinner is almost ready. I'll talk to you in a few days."

"OK. It's so good to talk to you, dearest love!"

Mom asked me, "How is Steve?"

"He's fine — rather busy right now. They just got the machine going and he's working on his project. I told him that he needed a break, and to get outside this weekend if the weather was decent. It's usually pretty mild at this time of year. I'm sure I won't miss the winters here, although Steve tells me that if there's any snow, nobody knows how to drive in it."

STEVE — Wednesday, February 28, 1962

I had taken Sandy's advice and gotten outdoors both Saturday and Sunday, and thought that the time off would help me find the problems with my program. True, it was easy for me to fix the test data that caused the only problem with phase two, and the minor problems with phase four. The Reaper code to manage the message tree structure was still not working right, although my debugging hooks and snapshots of the tree data helped a great deal. Finally, just before lunchtime today, I tried one more fix to the code, and it finally ran correctly. When it was done, Lou said, "That was a tough one, wasn't it?"

"I know. Look, you two have been watching me debug this, and I know you've learned a lot in the process. Why don't you sit down together and come up with what you'd like to see in the way of debugging tools for the Reaper code, other than programming in hooks and specific dumps of the data? I made some notes last spring, but never had a chance to go any further with it. We can get together tomorrow and compare notes, and then plan the work. I need to tell Murray and Sam Morris that we think it's now working, and then get with Samantha to start with the first production run."

Vlad said, "Yes, I have some ideas on that score, and I did some work on debugging tools as part of my summer job with Honeywell a year and a half ago."

"Good. Let's all go to lunch now."

We talked about our summer job experiences as we ate our lunch, and then, after reporting our status to Murray and Sam, went to the tape library. Samantha was there, and I told her, "I think we are ready to try our first production run. Do you have the real data for yesterday ready?"