Friday, January 29, 2021

HP, Apple, and thoughts



 So, here's a sample of what's wrong with American Education


Gurman works for Bloomberg; the San Jose Mercury-News just ran the story.   But the SJ Merc added the headline, which is demonstrably stupid and wrong.   Graduate of a local Bay Area school?

Moving past that faux pas, I was in a conversation with noted MIT researcher Charles Leiserson this week, and we were discussing his role in the CM-5, the first supercomputer to ever head the Famous Fifty list, cited in the last post re Rattner's key machines of the 1996 era.   The CM-5 debuted in 1992, architected by Danny Hillis and Leiserson.   It was the computer used for the movie Jurassic Park, even though the book cited Seymour Cray's X-MP.

Here is a Web quote re the CM-5
The surprise for me was when Leiserson held up his i-Phone 10, and said "this little baby outperforms that CM-5 in every dimension except size, weight, and cost.   The first CM-5 cost $45 million, he averred, but the performance metrics of the $1,000 iPhone 10 exceed the CM-5 'significantly.  WOW!

And I woke up this morning thinking about the combinatorics here:
1. The HP 35 handheld calculator was the "first mobile computer" of consequence, thanks to Hewlett
2. Wozniak worked for HP (and handhelds) and we struggled to see that this PC idea was valuable
3. My Logic State Analyzer team built the tools for HP Computer design, but also provided 50 of our top-end HP 1610As to Danny Hillis for the CM-1.   It was our largest order to date, in 1978. $500K.
4. Rattner (an old HP guy) worked with Hillis and the CM-5, in designing the RED machine for NSF .
5. And now HP, with notable attempts and little success, in mobile computing platforms (e.g phones) is about to deliver a competitive supercomputer for $40 million, shades of the CM-5 in its day

And that reminded me of an Intel friend, Nathan Zeldes, who visited our horse ranch last year, and met my sister-in-law, who asked "what did you do at Intel?"   He pulled out an iPhone, and said "ever seen one of these?"   She was, like, "duh, of course" to which he said, "I helped design the first Intel chips for these things, and this computer in my hand has more compute power than the nation of Israel when I went to work for Intel.   And now, with 900 million of these marvels in the world, we use them to argue with strangers and show pictures of cats."

This, some assert, is amazing progress







What goes 'round, comes 'round . . .

 The San Jose Mercury-News yesterday had two headlines that caused me to scratch my head . . . .

The first said "I-Phones top $100 Billion" but the story didn't support the headline.   We'll discuss that in a subsequent post.

The second said "HP building new supercomputer"       WHAT???

Here's a picture of the supercomputer headline and the first paragraph, along with a picture of "the machine" (must be a mock-up), followed by the citation


https://www.klove.com/news/tech-science/hp-building-new-supercomputer-to-be-one-of-world-s-fastest-19686 


Well, this is a fine "How Do You Do"?   Who'd have thunk that HP, venerable tired phlegmatic HP, would be vying to build a supercomputer, not only such a machine, but a real live competitive machine.

The 'deja vu' of this announcement was, for me, incredibly auspicious.   During my nearly 30 years at HP, of which some 23 were in Colorado, I spent two years working for NCAR because of my role with the Colorado Air Pollution Commission, circa 1970.    That role led to discovery, the first discovery I hasten to note, that emphysema (now lumped into COPD) is NOT just a smoker's disease, nor even primarily a smoker's disease, despite fifty years of the American Medical Association, the American Lung Association, and the recently maligned Center for Disease Control (CDC) insistence that it accounts for 85-90% of all COPD victims.   Pure bullshit, as can be demonstrated with any decent geospatial longitudinal study.   And this fraud is still perpetrated on an unaware America (and world).

The 'deja vu' part is that NCAR and the Colorado Dept of Health had practically no compute power in 1970.   One IBM 360/40 if I recall correctly (though it might have been the scientific configuartion of the 360/44) and access to a CDC 6600, courtesy of Kaman Nuclear next door to our HP Colorado Springs facility.   So when they handed me a stack of Z-fold paper with 2,754 names and 17 fields of data for deceased Colorado residents who were said to have died from emphysema, the question was "what do you do with this stack of stuff?"

Living in Colorado Springs, I had access to one of the first HP 9100A desktop calculators, and also to Lionel Baldwin's distance learning modules from Colorado State University (CSU) in Fort Collins, so I could take an advanced statistics course (I never liked that during my college days, but now I needed it)

With these tools, with virtually no one in the world possessed, except for the few hundred folk who had an HP9100 at the time, I was able to construct incredible graphs that demonstrated a different causal agent (in fact, several) that related to sub-micron particulate matter in a toxic gaseous environment--e.g. high mountain valleys with a local cement plant or lumber mill burning slash or with trapped auto exhausts.   One particularly interesting finding--closed-cab air conditioned tractors with smoking drivers or alfalfa dehydrating plants were deadly on the Colorado plains.  See, for example, https://www.researchgate.net/publication/236865558_Confronting_COPD 

The response, I expected, would be "Wow, that is fascinating" but instead it was "Well, how could you do something like that--our computers don't show that."   It was easy to ignore, and in fact, still is.  People, very often among the smartest, really resist being shown a new way that contradicts their earlier skillset and learning.   Didn't Thomas Kuhn write compellingly about such human behavior?

So, WHAT will NCAR do with this beast?   I think of the week prior to the Katrina hurricane, when NOAA computers predicted the onslaught on New Orleans 96 hours in advance, but NASA (who was the 'climate' agency) forbid them to publish it.    NASA's own computers did not predict the issue until 28 hours in advance, which was too late to get effective evacuations.    NOAA, as it turned out, had better supercomputer technology at the time (NASA was loathe to discuss that, let me assure you), built by Intel's advanced projects team, led by Justin Rattner who had led NSF's ASCI RED supercomputer effort earlier.   Rattner brought both groups to an Intel Fellows meeting, where we heard the story.  The story, as you may surmise, never made the public press.

His success, though, did get noticed.   In December 1996, Rattner was featured as Person of the Week by ABC World News for his visionary work on the Department of Energy ASCI Red System, the first computer to sustain one trillion operations per second (one teraFLOPS) and the fastest computer in the world between 1996 and 2000. In 1997, Rattner was honored as one of the Computing 200, the 200 individuals having the greatest impact on the U.S. computer industry today, and subsequently profiled in Wizards and Their Wonders (ACM Press). See https://www.huffpost.com/author/justin-r-rattner 

I cannot help but add two postscripts here:
1. Rattner worked for HP Computing in the early days (and PARC) before joining Intel in 1973.   
2. And three people from HP in 1997 were honored in that ACM book, Wizards, for the "200 in America who made computing".   Packard, Hewlett, and House (yup, me, imagine).  Should have had Birnbaum and Worley for sure.   John Cocke, who did the RISC work for IBM working for Joel, is in.
See https://www.amazon.com/Wizards-Their-Wonders-Portraits-Computing/dp/0897919602  



Saturday, January 9, 2021

Fault tracing

 Years ago, maybe even eons ago, I and a few others at HP Colorado Springs spent time trying to 'Chase Glitches"    The idea was that in the newfangled digital circuits of the day, the gremlins causing the 'real problems' were glitches, transient spikes of energy usually attributed to mis-designed logical paths, or to "race conditions" that had long bedeviled IBM mainframe designers of synchronous circuitry abetted by careful attention to wire cable lengths and timing diagrams for fast logic circuitry.

Bill Farnbach, a brilliant designer in our sampling 'scope labs  understood this problem from a different standpoint, which was that we needed to ignore these spikes and look ONLY where conditions were stable, after a switching event had 'settled down' -- this of course became the basis for HP's valuable Logic State Analyzer approach, and a couple of generations of logic designers got 'relief' from the worst of these transient behaviors.

But this is not a universal panacea, and HP Colorado Springs has continued to be on the forefront of dealing with these issues, even supplanting the 'old standard' Tektronix in this arena.   And a couple of weeks ago, at Greg Peters' urging, I made contact with Brad Doerr, the site manager for HP Colorado Springs, who graciously took me through a tour of their amazing new "Fault Tracer"

Talk about a versatile solution.  This is essentially a fully tailorable 'trigger system' for a 'scope (analog) display, or a logic analyzer (state) display.   The trigger shape, pattern, magnitude, directionality, and duration can all be controlled by the operator -- what we in the olden days would have called a "do-all" trigger function.   I was mesmerized.   The world has definitely evolved from the simple one I knew.

I am attaching the first page of a once-confidential document that describes this for your reading interest.   And I thank Greg and Brad for their patience with my questions about it.