Slashdot Mirror
Gene Amdahl, Pioneer of Mainframe Computing, Dies At 92 (nytimes.com)
An anonymous reader writes: The NY Times reports that Dr. Gene Amdahl, who played a crucial role in developing the IBM System/360 series mainframes and formulated Amdahl's law, has died at the age of 92. "The 360 series was not one computer but a family of compatible machines. Computers in the series used processors of different speeds and power, yet all understood a common language. This allowed customers to purchase a smaller system knowing they could migrate to a larger, more powerful machine if their needs grew, without reprogramming the application software. IBM's current mainframes can still run some System/360 applications. ... Dr. Amdahl is remembered at IBM as an intellectual leader who could get different strong-minded groups to reach agreement on technical issues."
You were quoted for decades during life and will be quoted after death for a long time. That is the everlasting life after death, with an almost infinite speedup.
And just how many systems are running today which are still compatible with the IBM System/360?
I can't count how many retail stores I've been in who still bring up an IBM terminal emulator to do the real work, and I've known several places which have had mainframes running since the 60s. Every now and then you swap out a part while it's running.
That's a body of work ... safe travels Dr. Amdahl.
Lost at C:>. Found at C.
I had a recurring nightmare in the '80s (after Amdahl and Cray had spun out on their own): That Gene Amdahl, Gordon Bell (DEC's PDP 5, PDP 8 instruction set, PDP 11), and Seymour Cray (CDC, Cray Computers) would get together, found an "ABC Computer Company", and spend the next decades having architectural arguments but never producing a product.
Two of them are gone now, and the world is poorer for it.
(Ever wonder why the cabinets of IBM computers in the mid 20th century were the size they were? One of Amdahl's ideas: After seeing a facility have to tear out a wall to install an early machine (univac?) with the spectre of having to do it again to get it out some day, he designed his machines in modules that each just fit through the door and into the car of a standard elevator (with a bit of clearance for padding and room for someone to push the floor buttons. That meant lots of expensive custom cables and connectors, but still far cheaper than tearing up buildings. B-) )
Bantam Dominique roosters crow a four-note song. Once you've heard it as "Happy BIRTHday" you can't NOT hear it that way
Amdahl also coined the term "FUD" meaning a company trash talking another.
https://en.wikipedia.org/wiki/Fear,_uncertainty_and_doubt#Definition
"...FUD was first defined with its specific current meaning by Gene Amdahl the same year, 1975, after he left IBM to found his own company, Amdahl Corp.: "FUD is the fear, uncertainty, and doubt that IBM sales people instill in the minds of potential customers who might be considering Amdahl products..."
The point is that people has always been trash talking others, but IBM was the first company that systematically employed FUD on every level in the company.
His imprint remains on my life to this day. I probably owe this guy more than a beer. Cheers Dr. Amdahl!
---Up Up Down Down Left Right Left Right B A START
edge first
so does this count as unscheduled down time??
Brings back memories. Learned computing on an Amdahl 470 V/6 (the second one made, I think) back in the days of punch cards and standing in line at the input window...
Hang gliding accident.
Anons need not reply. Questions end with a question mark.
That should be TWELVE edge first. Nine edge was for the old card sorters.
Oops. My memory played tricks on me - it's twelve edge OUT, not twelve edge first. My apologies.
The exact expiration date and time for each specimen of the current model of human remains unknown. He lived a nice long life and contributed to mankind's technological progress. Thanks.
jargon file has it as nine edge first (ref eighty column mind)
Space and military spending did push research on integrated circuits (IC's) because they were the biggest customers IC's, looking for weight savings above other factors. The IBM 360 used mostly stand-alone transistors and sometimes magnetic core memory (iron donuts on a wire grid).
IBM was skeptical of the reliability of both the hardware and supply chain of IC's at the time of the 360. IC factories were not yet ready for mass manufacturing. But IBM gradually introduced IC's in later models.
Table-ized A.I.
Correction: biggest customers of IC's
Table-ized A.I.
Just heard some sad news on talk radio - Computer designer Gene Amdahl was found dead in his Palo Alto home this morning. There weren't any more details. I'm sure everyone in the Slashdot community will miss him - even if you didn't enjoy his work, there's no denying his contributions to geek culture. Truly an American icon.
Welcome to the Panopticon. Used to be a prison, now it's your home.
Amdahl finally reached his limit.
Too soon? ... I'll let myself out.
Well, what a good innings and what a man who helped bring computing technology to where it is today. There must be some way that we can immortalise people like he and Seymour Cray. I worry that people will forget the legacy that mainframe computing architecture has left us; if we knew more about the history of computing architecture, we might stop re-inventing wheels.
... Now you're telling me a company called Internatioal "Business" Machines had some sort of weird 32 bit computer family in the 1960s?
I didn't read TFA, but I'm guessing the 360 was one of the first or second generation of 32-bit machines. Decades ago 32-bit wasn't the thing. A lot of the bigger older machines used 36 bit data types. That allowed integers up to 10 decimal digits long to be represented so that these new electronic computers could compete with the 10 digit mechanical computers that preceded them. Later, the advantages of using powers of two won out over that compatibility.
If you're working on data bus width or data element width, I think you can count S/360 as 32 bit since some models implemented a 32-bit memory data bus(except it did support 64-bit floating point) but it was always limited in address space to less than that.
I suspect that one class of software the AC was referring to was the various shops' own (whether they developed it themselves or purchased it and then customized it, which was very common in those days) [business] applications software. In that case, the assertion is certainly true and, the larger the company, the more often that will be the case, simply because there was a (then) huge investment in COBOL or PL/I (or even Assembler) programs in the System/360 era, which was replaced, but not that quickly, with the System/370 machines (particularly the Models 155, 165 and 158 and 168). But there's another class of System/360 software: all the lines of code that constituted OS/360 itself and all of its mostly free subsystems. There were some licensed program products that IBM had just started to charge for, such as IMS and CICS, that did run on OS/360, so I will include them as well. It is my estimate that more than 25% of all the lines of code that were ever included in OS/360 and its subsystems still exist in z/OS or the current generation of those subsystems. A lot of OS/360 code was of course replaced, rewritten, or discarded as it became OS/VS2 SVS and then OS/VS2 MVS, MVS/XA, MVS/ESA, OS/390, and finally z/OS. But much of that code still remains [in the source] as can be seen by anyone who can look and knows what to look for and where to look. Two major OS/360 subsystems, ASP and HASP, now called JES3 and JES2 respectively, still contain more than 25% of their original lines of code that once ran on a System/360 running OS/360. I still have object decks of PL/I programs compiled in 1967 that will still bind (link) and run on z/OS Version 2 Release 2 -- that's 48 year-old code. Not only that, but at one of my old shops, they still run Assembler and a small number of COBOL programs that were last compiled (and last linked, even) in 1973! The programs work, and there has never been a need to change them. That is rare, of course, but not unheard of and I suspect more common that many insiders would ordinarily think.
The 360 was the first machine to use an 8-bit byte as the unit of memory addressability. In that single byte you could store either 2 BCD digits, a text character, or a binary number. Prior to the 360 machine were more special-purpose; they were good for either decimal math (financials), or text, or binary math, but not all three at the same time. The register size of the 360 was 4 bytes. The address size was 24 bits, allowing 16MB of addressing.
Decimal math (BCD) is still heavily used in financial applications.
The USA and the USSR were in a "space race". The USSR, typically for the Russian mindset, had rocket engines with pure brute force that could easily shove heavy payloads into space.
The USA hadn't invested as much work into powerful engines, so to compensate, they made a virtue of micro-miniaturization. It probably helped that a lot of the groundwork technology (such as the transistor) had been developed using US resources (AT&T and the feds go way back).
The computer you are using today isn't some fuse-blowing vacuum tube behemoth in large part because this self-same miniaturization found its way back into the commercial sphere. I have an old flip-flop IC that I bought from Radio Shack that says "nasa" right on the case, so in some cases, apparently quite literally so.
No, the space program didn't bring us velcro or even Tang, but it did affect computer technology development even though relatively few computers of the day were actually used in the space program.
| The IBM 360 used mostly stand-alone transistors and sometimes magnetic core memory (iron donuts on a wire grid).
Sometimes? Nope. Essentially exclusively. All but two (individual machines) in the IBM System/360 model line used magnetic ferrite-core memory exclusively. The singular exception was the Model 95 [360/95] processor. IBM built only two, for NASA: one was installed at the Institute for Space Studies in New York City, and the other at the Goddard Space Flight Center at Greenbelt, Maryland. Even so, the The Model 95 still used a hybrid main memory technology: each had only 1MB of thin-film memory, but also 4MB of conventional ferrite-core memory.
R.I.P
In GOD we trust, all others we monitor.
A few years a go, I read a few books about computer history. I suspect that I saw his name, but I don't remember any mention of his being from South Dakota. Being a SD native, this is nice. I see he also served in the US Navy during WWII.
Another name to add alongside Ernest Lawrence and Joe Foss.
I've been at various parts of the Bell System and AT&T for some decades now. There were years when it was much easier to get leftover 23" telco racks for my lab than buy new 19" racks for computers, so we had a lot of extra rails for conversion. The more serious problem with that in recent years is that computer racks are generally deeper than telco racks, so not everything would fit in the cabinets (or we'd have to take the back door off.)
Bill Stewart
New Fast-Compression-only CPR http://preview.tinyurl.com/dy575ks
I studied your law enough to build on it. I never met you however you influenced me and I appreciate the inspiration. May you have ample CPU time for your journey and not suffer cpu latency getting there.
My ism, it's full of beliefs.
Okay, I was wrong about individual transistors. But I'm not sure those qualify as "integrated circuits". But let's not get into a terminology game, as space and military spending on IC technology was a big driver of IC R&D. I remember Intel stating that somewhere, but cannot currently find the source.
Making small circuits was more expensive than "medium" circuits at the time, like SLT, such that it was more economical for most commercial computers to use the "medium" circuits as they were cheaper and more plentiful than the smaller ones of the time, being the smaller ones were cutting edge with cutting edge problems to be worked out. Aerospace, on the other hand, paid a premium for reduced size. I imagine they had to throw away a lot of attempts to get one "good chip".
Table-ized A.I.