[Most Recent Entries]
Below are the 7 most recent journal entries recorded in
Al Gilman's LiveJournal:
|Thursday, May 23rd, 2013|
|The first time we ran the VHDL Simulator, it taught us something
[Bragging lines: The VHSIC Hardware Description Language was a second project I got to contribute to in a technical leadership role where the product escaped from the hothouse of military contracting and took on a life of its own in the commercial marketplace. By the time we delivered VHDL to the Air Force, I had taken on the role of Deputy Program Manager for Intermetrics, who were prime contractor for this development.']
In the 1980's the DoD ran a big capacity-building program called Very High Speed Integrated Circuits (VHSIC) to save the US computer chip industry from foreign competition, which had already pretty much taken over the memory chip market. Within this budget was a (10%?) set-aside for the development of design tools. The centerpiece of this effort was to be a "hardware description language" to capture the logical organization of computer chips.
The DoD at that time was heavily committed to the Ada programming language, and the staff at the Air Force laboratories at Wright-Patterson AFB, notably Dr. John Hines, got enthusiastic about the idea of a text declarative description of chip architecture. They had enough influence in the DoD brain trust of the VHSIC program that the RFP went out for VHDL, the VHSIC Hardware Description Language. The requirements had a strong Ada flavor, but also required a capacity to simulate the behavior of the circuit so described.
The closest thing in the field to what the DoD asked for was a simulation technology from Texas Instruments. They had a clear lead in that regard. On the other hand, their position was so clear throughout the industry, that like Lavrenti Beria, the chief of secret police at the time Stalin died, they had too many natural enemies in the industry for their selection to do the new 'standard' to be accepted. The program manager at Texas Instruments could see this, and recruited IBM to team with them on this program, because if there were anyone that the industry would trust not to be pushed around by TI, it was IBM. Once that partnership was agreed, they went shopping for an Ada house to front the job, because of the clear Ada flavor of the requirements.
Intermetrics was a clear candidate as a high-tech software specialist with strong Ada credentials. Intermetrics came in second in the design competition for Ada, having submitted a design that met the stated requirements, unlike the winning design from Jean Ichbiah. Now at this time, in the Betheda office where I had been shipped when kicked out of my management role in Cambridge, Roger Lipsett spearheaded a brown-bag lunch seminar series out of sheer intellectual curiosity on the problem of "silicon compilation," -- if you tried to use compiler technology to synthesize chip designs, how would it work? So the IBM/TI team visited our competitors in the Ada market and the competitors responded "a text language to describe hardware -- that's an interesting idea." When they came to Intermetrics, Bethesda the reaction was "What do you do about this?" And when TI and IBM answered that question, there were more: "But, what do you do about that? ..." There was no comparison. So Intermetrics, Bethesda was selected to be the "white lab coat" prime contractor leading the team, with IBM and TI as subcontractors. Like the little ballerina riding two elephants in the circus.
I was knocking around the Bethesda office doing odd jobs, having burnt my bridges in the area of my expertise -- Navigation and Guidance. They needed troops to write the proposal, so Roger gave me the job to write the section on the System Specification for the support software. This seemed my best use as something of a Systems Engineer. Once again, my main talent in this task was in perceiving what the government engineers were after as they wrote the technical requirements in the Statement of Work. And I had been through an Air Force software procurement on GPS, writing specifications to the Air Force's preferred MIL-STD recipes. Something worked. Whether it was the capability shown by TI and IBM, the Ada credibility of our company, or the system spec section I wrote, the Intermetrics team won the contract, and we set about to develop a specification of the language and some software that would parse, semantically analyze, and simulate the descriptions.
Once again, they needed hands to take on jobs, so Roger turned to me with the opportunity: "You wrote this section of the proposal (the System Spec.), do you want to do that part of the job?" And I went to work. I fought with the young lieutenant who was the project engineer for the Air Force. He wanted interface descriptions nailed down to the bit level between the segments of the system: Analyzer, Design Library, and Simulator. I was a firm believer in progressive refinement across the board, and kept the System Specification at a high level. In the proposal, the division of labor was that intermetrics was to do the language specification and the Analyzer tool that parsed and performed static semantic analysis on the language, and left an intermediate form in the Design Library. Texas Instruments contributed language and simulation requirements from a chip-design expertise, and was to develop the simulator that read the Design Library and produced time-wise I/O histories of the circuit. IBM provided language and simulation requirements from a system design expertise base.
My job turned out to be to sit on the heads of the computer science smarties in our company who thought they understood what they needed to do, and make the listen to the chip designers from TI and the computer designers from IBM until they actually understood the problem. Early in the contract, we had a big "come, let us reason together" conference of all the VHSIC contractors under contract to the DoD anywhere. We had to give a briefing on the project that we had recently won. Roger Lipsett was unable to attend to give the briefing, and I stepped in to give his presentation. We got challenging questions from the lead technical people of the losing bidders, and I fielded them adroitly. I was dubbed the "designated spear catcher" from then on and given the title of Deputy Program Manager.
The simulator work at TI was not under the control of their simulator people, but Computer Science types who believed in the Ada voodo of the DoD. I exaggerate. Lots of the ideas, taken one at a time, were good ideas. But the TI team came up with a development strategy for the simulator that involved a seven-step bootstrap process of each language compiling the compiler for the next stage. The simulator never happened. The Air Force got sufficiently impatient that we had to submit a contract change to develop the simulator ourselves. So the whole support software bundle was developed at Intermetrics, Bethesda from Analyzer to Simulator. The Analyzer went through null compiles and non-null compiles and we got to the point where we were ready to try to run the simulator. So we needed a hardware description. I volunteered to write it, everyone else was busy, and they accepted. I put together a description of a four-bit counter hooked up out of flip-flops. So we ran the simulator, and the simulation gave us our report, but the counter didn't count. What was wrong with the simulator? Nothing. I had one of the connections hooked up wrong in the counter. Roger was flabbergasted that I had pulled this circuit description our of my anus, not a textbook. I didn't have a digital circuits text, I had never taken digital circuits -- there was just this plug-board toy I had played with in instrumentation lab as an undergraduate, where you could make a counter out of relay-based logic. So Roger rushed back to his office, pulled a digital circuits text off the shelf, and there in his book was my counter, modulo the one bug I had programmed into the hardware description -- not the simulator. So Doug Dixon's simulator was smarter that the nut behind the wheel, even the first time we ever ran it over a VHDL description!
Epilogue: The Intermetrics / TI / IBM team delivered the VHDL language specification and support software to the DoD. DoD wanted it to be an industry standard, and the DoD favorite way to gain Standard status was through the IEEE. So Ron Waxman of IBM got the necessary group set up in the IEEE, and the follow-on progressed. The language designers from Intermetrics jumped ship to form a startup, thinking that if they controlled the Analyzer for the language, they could control the Design Library and make lots of money supporting this platform technology for digital microcircuit design. That didn't quite work out, because graphical interfaces and not a text language won out as the human input medium for design, even as the abstraction level of design migrated upward. But the language itself found a commercial niche as the neutral form between a competing collection of high-level design tools, the 'silicon compilers' of our dreams, and the low-level technology-bound design tools that took a design from a logic network to circuit layout and mask geometry. There were separate pools of competitors at the high and low levels of design, and this allowed the high-level design tool industry to take hold and flourish no longer captive of the polygon-thinking low-level tools. VHDL filled this role as the Moore's-law march of miniaturization transited the area around a chip scale one million gates. As the fabrication technology progressed still further, it became necessary to integrate the levels of design differently, the floor-planning becoming sensitive to more of the parametrics of the interconnection runs, that then had to be looked at as microwave guides rather than Kirchoff-law isopotential nodes.
|Wednesday, March 10th, 2010|
|go with the flow
I have had recurring problems with my electric bill payment being late. My habit has been to make a sweep through my monthly bills some time around the first of the month. This works well for bills where the grace period runs out some time between the fifth and tenth of the month, as is most common. But the electric bill is awkward. The late-after date floats around during the last week of the month, and their bill-paying facility is supplied by CheckFree, which introduces an annoying two-BUSINESS-day lag in the payments. I'm accustomed in other relationships to being able to rush in at the eleventh hour on the ultimate day and still get credit for making the deadline.
Today I finaly stopped beating my head against this particular stone wall. It finally dawned on me that this payment setup DOES let me schedule payments into the future. So I took the occasion of the "your bill is now ready" email as my trigger to go in and SCHEDULE a payment that would be on time. I even give them two extra days with the money more than I have to. This way I get as much float as I choose, but I don't run the risk of depending on my remembering the peregrine due date from month to month.
Off to get into another unproductive rut. Like blogging...
|Wednesday, October 21st, 2009|
|All Saints Concert Nov. 1
On Sunday evening, November 1 at 7:00 p.m., Immanuel Church On-the-Hill will present a concert of sacred music, in the Zabriskie Chapel
. The music will range from plainchant and Southern Harmony Americana to modern works by John Tavener. I will be in the choir for this concert.
|Brudieu Requiem Nov. 2
Hallowe'en is the eve of All Saints' Day. All Souls is the day after (Nov 2). On All Souls the Church remembers everyone who has died, no matter how sainted. All Souls Memorial
Church (by the Zoo) in DC will observe All Souls Day with a High Mass at 7:00 p.m. The Brudieu Officium Defunctorum
will be the service music for this service, showing off the mastery of early music of the All Souls Choir. I will be in the choir for this service.
|Duruflé delights Nov. 6
On Friday, Nov. 6 there will be a concert at Grace Church, Alexandria
. The program includes the Requiem, Op. 9
of Maurice Duruflé and also his Suite Pour Orgue. Thom Robertson will conduct the Grace Church Choral Society, Nathan Laube will be the organist, and Jonathan Hoffman and Debby Wenner will sing solos. I'll be in the chorus. The Requiem is one of the masterpieces of 20th Century music. The whole program is designed to show off the new organ, which will be dedicated Oct. 29.
|Friday, October 9th, 2009|
|Winning at football
[for those not of the Pompton Lakes High School class of 1959, you should understand that PLHS was a football-mad school and I was a nerd who was lucky not to get cut from the track team.]
In high school, I didn't play football. My parents would not sign the parent consent form, on account of relatives with life-long football injuries.
But coming from Pompton Lakes, you will appreciate how much it meant to me to, once in my life, be the _quarterback_ of a victorious football team. Here's the story of how I quarterbacked our side to victory, once in my life.
You have to understand that it was a pick-up game. The occasion was the METCO picnic in Lexington, MA. METCO was a voluntary busing program in Lexington which brought Boston students to Lexington schools. One of the features of the program was that each out-of-town student had a host family to serve as alternate/emergency contact during the school day. To make this a more meaningful relationship, the commuter students spent one afternoon a month with their host family. We were such a host family. An additional bonding opportunity was the annual picnic, alternating between Lexington and Boston locations, where both home and host families were invited and all had a general good cookout and time. At one of these events we decided to have a pickup football game, and somehow early in the game I managed to wangle my way into the quarterback position.
Here's how we won. First off: talent. We had the Principal on our side. He served as my blocking back. Nobody really wanted to get rude with him, so I had time to set up in the pocket. Secondly, we had a colleague of mine who, although not supremely athletic, was a head and shoulders taller than anyone else in the game. This meant that we had a "long bomb" threat. I could sent him far down the field and a hail-mary pass would connect, because he could get under it as it came down and get to the ball before the opposition could just by reaching up. (I think the operating rule was that two connected forward passes gained you a first down.)
But this was too obvious. After two such go-deep plays, the entire opposing team would became the secondary, and my receiver was drowning in defenders. This is where democracy, feminism, or what-you-may-call-it came to our rescue. I saw that one of the mothers had progressed over the line of scrimmage and was totally free in the flat. I lobbed a soft pass to her. She was amazed, juggled the ball for a bit, and finally found the handle and hung on. Voilà! we were un-stoppable. They had to defend everyone, everywhere; and the long bomb was once again there when we wanted to use it. So we rolled on to victory.
The reason this is reunion news has to do with why I could put the passes where they needed to be for the receivers to catch them. I lay this up to the afternoons spent playing sandlot football in Chet Allen's back yard with our brothers Art and Bobby as receivers. I think that's where I got a grip on how to throw a football.
Al Current Mood: nostalgic
|Friday, November 14th, 2003|
|Score one for Java
At SC2003 next week we are going to repeat the "simultaneous transcription service" gig that Trace did last year.http://trace.wisc.edu/handouts/sc2002/index.htm
To use the Applet that implements the corrector's client, I had to borrow a more modern WinTel box that my beloved Win98SE relic, because the screen formatting was hosed when I tried to run the Applet.
Another year, another Java. With the new Java runtime, it works with the OLD Windows. I can play from the comfort of my own iron.
We fossils appreciate stuff like this.