The Future of Computing (According to Me)

[PabloMack]

I just ordered a book that tells the story about the IBM AS/400 and its /34~38 ancestors.
It is a very strange story and a very strange architecture that was developed in isolation
from other parts of IBM including their mainframes. The location was Rochester Minnesota.
This was a business system that, apparently, did virtual mapping of all of the data that it
would ever access. Once the data was written, it could never be changed (or presumably
deleted). This is ostensibly so that data could not be altered to ensure integrity.

Back in 1988, I was working for a large corporation that had an IS department that had
been using a System /34 upgraded to a /38 and was being upgraded to an AS/400
(because nobody got fired for choosing IBM). The department I worked for was called
Advanced Process Control. We used a lot of DEC PDP/11 and VAXes as well as PCs but IS
was all IBM. My boss had to deal with the IS department for many reasons. After IS had
been using the new AS/400 for a while, my boss said that they could never figure out how
to do a backup so they just had to keep adding memory and hard drives. I never knew
how much RAM and disk storage they ended up having but it must have been huge.
They had a row of big racks full of just RAM and hard drives.

It will be interesting reading about it in more detail. There is still a lot of secrecy around
this mysterious line of computers called the iSeries which was eventually replaced with
the POWER series. I am not clear about how this relates to the PowerPC microprocessor
architecture but I have one or more books about that as well. It is also strange.

https://www.amazon.com/gp/product/1882419669/ref=ppx_yo_dt_b_asin_title_o00_s00?ie=UTF8&psc=1

I now have the basic programmer's model worked out for ϕEngine and a working assembler.
Today I started writing a document that will define how the HW works internally.

[masonspappy]

Interesting comments about the AS/400.  I worked for IBM for 25 years (1979-2002) and spent 3 years as a member of their mid-range support group that included support for AS/400s (which was later renamed to "i-Series".) After IBM I worked with an IBM Business Partner as a Project Manager  until 2007 .

Not sure why backups couldn't be done - that  was pretty  routine in my customer accounts.

AS/400s were the darlings of data processing in my part of the country (which was predominately industrial manufacturing in remote/rural  areas)  because they could be installed in less-than-pristine environments and they simply ran. And Ran. And Ran.  The Operating System was written specifically to the unique physical and logical architecture of the AS/400.  Much of my AS/400 work work was related to various attempts to keep the box relevant by installing a special board containing an Intel chipset and using some of the AS/400's memory and other resources as its own. Basically, a portion of the AS/400 was turned into a PC server.  It was a cool concept but never really caught on. 

I've been to the plant in Rochester several times prior to 2007.  It's a lovely area and the people were nice, although often a bit odd because they seemed so isolated from the "outside world" (ie, any computing platform that was not AS/400.)   And I believe that this isolation contributed to AS400/iSeries decline and end.

There were a lot of things right & wonderfull about the AS/400, but there were shortcomings as well.  And customers who loved their AS400/iSeries were very vocal about those shortcomings.  But the changes didn't happen fast enough and often enough, and customers believed  IBM simply wasn't paying attention.  Once IBM realized that customers were increasingly deserting the platform there were some frantic attempts to staunch the bleeding (that's where I came in - I was a PC server/Network/Project manager) , but by then it was too late.  IBM's strategy was to move the AS400/iSeries  to the PowerPC architecture,  which allowed it to run on the same hardware as RS/6000 (once upon a time AIX).  This in turn allowed the two OS's to be consolidated into the "IBM Power systems" which allowed customers more latitude in what OS they chose to use.   

By this time there were a lot of former AS400/iSeries support folks looking for new jobs.

I had pursued a Project Manager ticket (through IBM) around 1994, so I had another horse to ride long before leaving IBM in 2002 (and I rode that horse hard until my retirement 5 months ago.  But the downward spiral pattern I observed is repeated constantly in this industry and other industries as well:
- Company releases a great, innovated product that does its job well but has a few shortcomings
- Customers repeatedly point out shortcomings but increasingly feel that needed changes don't happen quickly enough or are only given lip service.
- Frustrated customers increasingly desert the product in favor of another, competing product.
- A bit of panic as the Company realizes they have are loosing/have already lost a sizeable chunk of their customer base.
- Company attempts to shore up the product but frequently it's "too little too late".
- {sound of bugle playing 'Taps"}


Old joke told at iSeries marketing meeting:

Young woman sits down with her husband-to-be the night before their wedding and explains that she's been married 3 time before but none of the marriages were ever consummated.
"How is that even possible?" exclaims the bewildered man.
The woman explains: "My first husband was pulling wheelies on his motorcycle during the wedding reception and flipped the bike and killed himself.  My second husband was this really old guy who died of a heart attack as he carried me over the threshold at our honeymoon hotel. And the third guy was an IBM iSeries sales rep who just sat on the edge of the bed night after night telling me how great it was going to be, and I shot him out of sheer frustration."

[PabloMack]

I worked for IBM for 25 years (1979-2002) and spent 3 years as a member of their mid-range support group that included support for AS/400s...

Very interesting background. Thank you for sharing.

There will always be enough smart people to write compilers for difficult-to-understand CPUs (not just in multi-billion dollar companies, but also independents) so that the rest of us don't have to understand them...

Apparently, the complexity of writing compilers for Itanium even broke the backs of Intel and HP.

https://softwareengineering.stackexchange.com/questions/279334/why-was-the-itanium-processor-difficult-to-write-a-compiler-for

[WAS]

Interesting comments about the AS/400.  I worked for IBM for 25 years (1979-2002) and spent 3 years as a member of their mid-range support group that included support for AS/400s (which was later renamed to "i-Series".) After IBM I worked with an IBM Business Partner as a Project Manager  until 2007 .

Not sure why backups couldn't be done - that  was pretty  routine in my customer accounts.

AS/400s were the darlings of data processing in my part of the country (which was predominately industrial manufacturing in remote/rural  areas)  because they could be installed in less-than-pristine environments and they simply ran. And Ran. And Ran.  The Operating System was written specifically to the unique physical and logical architecture of the AS/400.  Much of my AS/400 work work was related to various attempts to keep the box relevant by installing a special board containing an Intel chipset and using some of the AS/400's memory and other resources as its own. Basically, a portion of the AS/400 was turned into a PC server.  It was a cool concept but never really caught on. 

I've been to the plant in Rochester several times prior to 2007.  It's a lovely area and the people were nice, although often a bit odd because they seemed so isolated from the "outside world" (ie, any computing platform that was not AS/400.)   And I believe that this isolation contributed to AS400/iSeries decline and end.

There were a lot of things right & wonderfull about the AS/400, but there were shortcomings as well.  And customers who loved their AS400/iSeries were very vocal about those shortcomings.  But the changes didn't happen fast enough and often enough, and customers believed  IBM simply wasn't paying attention.  Once IBM realized that customers were increasingly deserting the platform there were some frantic attempts to staunch the bleeding (that's where I came in - I was a PC server/Network/Project manager) , but by then it was too late.  IBM's strategy was to move the AS400/iSeries  to the PowerPC architecture,  which allowed it to run on the same hardware as RS/6000 (once upon a time AIX).  This in turn allowed the two OS's to be consolidated into the "IBM Power systems" which allowed customers more latitude in what OS they chose to use.   

By this time there were a lot of former AS400/iSeries support folks looking for new jobs.

I had pursued a Project Manager ticket (through IBM) around 1994, so I had another horse to ride long before leaving IBM in 2002 (and I rode that horse hard until my retirement 5 months ago.  But the downward spiral pattern I observed is repeated constantly in this industry and other industries as well:
- Company releases a great, innovated product that does its job well but has a few shortcomings
- Customers repeatedly point out shortcomings but increasingly feel that needed changes don't happen quickly enough or are only given lip service.
- Frustrated customers increasingly desert the product in favor of another, competing product.
- A bit of panic as the Company realizes they have are loosing/have already lost a sizeable chunk of their customer base.
- Company attempts to shore up the product but frequently it's "too little too late".
- {sound of bugle playing 'Taps"}


Old joke told at iSeries marketing meeting:

Young woman sits down with her husband-to-be the night before their wedding and explains that she's been married 3 time before but none of the marriages were ever consummated.
"How is that even possible?" exclaims the bewildered man.
The woman explains: "My first husband was pulling wheelies on his motorcycle during the wedding reception and flipped the bike and killed himself.  My second husband was this really old guy who died of a heart attack as he carried me over the threshold at our honeymoon hotel. And the third guy was an IBM iSeries sales rep who just sat on the edge of the bed night after night telling me how great it was going to be, and I shot him out of sheer frustration."

Wow interesting background you have. Thanks for sharing your history with us.

Are those old machines worth anything? My best friend has this as400 E series desktop tower in his garage.

[masonspappy]

Are those old machines worth anything? My best friend has this as400 E series desktop tower in his garage.

Probably only as scrap or museum piece.   

[PabloMack]

Are those old machines worth anything? My best friend has this as400 E series desktop tower in his garage.

I think they would be worth more if their internal architecture had been made public. But IBM has a habit of keeping their internals very secret. I remember that we looked the /3X IBM mini that was being replaced. We took out the boards and all of the chips had their labels blacked out with a magic marker to hide what they were. I'm sure that IBM uses a lot of 3rd-party components but they don't like the labels their manufacturers put on them. These computers rely on a pile of software from IBM and the computers would be pretty useless without them. They were business machines so its not like their is any other kind of application available for them.

[WAS]

We did some digging online and seems this specific tower goes for 600-1000. He's pulling it apart now and testing components.

Here is one listed right now for 800: https://www.ebay.com/itm/IBM-AS-400-Computer/312566218959?hash=item48c66600cf:g:PeMAAOSwaplcnYap

Considering what it is, and the fact is has basically no use today, no PC-type utilities like browsers etc, that is very expensive.

[PabloMack]

We did some digging online and seems this specific tower goes for 600-1000.

I seem to recall that the IS people were offering to give this larger than a washing machine
system to anyone who wanted it. Considering the size, lack of documentation etc. I turned
them down. If I had been a salesman I probably could have found a buyer. But I'm not much
of a salesman.

[WAS]

We did some digging online and seems this specific tower goes for 600-1000.

I seem to recall that the IS people were offering to give this larger than a washing machine
system to anyone who wanted it. Considering the size, lack of documentation etc. I turned
them down. If I had been a salesman I probably could have found a buyer. But I'm not much
of a salesman.

They definitely seem to be selling as novelty pieces based purely in nostalgia.

[PabloMack]

Just an update. CPU architectures have many trade-offs between a large number of things such as number of registers, addressing modes, kinds of operations that are available etc. In trying to maintain high levels of such things as coding efficiency, simplicity and  capability, generally you have to give up some of one thing in order to get more of another. Designing a CPU architecture is a complicated problem. Today I decided to make a change to the standard addressing modes that are used with most of the general-purpose instructions in ϕEngine.

The x86 was given a new lease on life when AMD defined the AMD64. But it was somewhat of a stop-gap solution and I admire what they accomplished. However, even though the x86 has a pretty full set of 64-bit data operations, its addressing is still pretty weak and little improved. The architecture was basically upgraded to give some breathing room in its address space. In this regard, it is more like a 32-bit architecture with a 64-bit address space which can accommodate multiple 32-bit images. Programs are still limited to 2 G-Bytes unless you do a lot of back-flips in order to address more than this. The ModRM byte has only a 2-bit address mode field which accommodates offsets of either 8-bits or 32-bits. That is it! To reach farther than that requires a lot of manual address calculations. You'll get little help from the linker or link-loader. Jumps have the same limitations.

ϕEngine has double the number of address modes in its standard operations. What I did today replaced two auto-increment address modes with indirect with offsets that can reach the whole address space for pretty much the entire instruction set.

During the 1970's and 80's, Digital Equipment Corp. was a major player in the minicomputer market. Their operating system (VMS) was pretty powerful and was used by a lot of industrial companies. When DEC was acquired by HP, different pieces of the company went in different directions. Intel ended up with rights to the Alpha but VMS was spun off and became OpenVMS which is still in use today. As is the case with Microsoft, they are at the mercy of CPU hardware manufacturers and have found themselves like a polar bear on a melting ice berg. Support for the x86 is now in development. For the computer-philes among you, I found this talk so interesting that I watched it twice.

https://www.youtube.com/watch?v=FZN6LjuEgdw

[PabloMack]

Another update. I have been extensively using a program I wrote (I call "CPU") to manage the instruction set of ϕEngine. It spits out source code for the assembler to make that job a lot easier. I have reworked the ISA many times and now feel it is in pretty good shape and it lays out an excellent plan. Friday I officially started work on the hardware design. I anticipate that I will be spending the lion's share of my time during the next several months using a simulator that I purchased a license for a few days ago. It will be cool to finally be using tools that produce true parallelism which programming languages do not since they are inherently sequential. I am also planning to expand "CPU" to automate code generation for the hardware design. This is interdisciplinary co-design on steroids. I feel like I was born to do this.

[PabloMack]

Another Update. Just before Russia's invasion of Ukraine, I started work on the next generation parser. I went from there to a start on the code generator. I also did a rework of the syntax and this meant I had to go back to the assembler for some weeks. I reworked the vector instructions and they are much better than they were. So now I am working on HDL again and CPU is expanding to fill some needs in anticipation of execution. CPU spits out an instruction decoder in Verilog. It compiles without error. Seeing if it will simulate is another matter.

[PabloMack]

Update. I'm really getting into the hardware design of ϕEngine now. I can't remember when I've had so much fun. Things are moving along smoothly so far. This first implementation will be micro-coded. I bought an FPGA-based SBC from Digilent. It will run a soft-core version of ϕEngine. Hopefully the FPGA will have enough logic and memory for it. It will be a new experience to have a computer with a user-definable processor. Even if it only has the performance of a 68030/80386, I will be ecstatic.

[WAS]

Update. I'm really getting into the hardware design of ϕEngine now. I can't remember when I've had so much fun. Things are moving along smoothly so far. This first implementation will be micro-coded. I bought an FPGA-based SBC from Digilent. It will run a soft-core version of ϕEngine. Hopefully the FPGA will have enough logic and memory for it. It will be a new experience to have a computer with a user-definable processor. Even if it only has the performance of a 68030/80386, I will be ecstatic.

That sounds awesome. Even if it's performance is small by today's standards, that still a huge step in heading in the right direction. I'd argue there is still a lot you could do with that sort of power anyhow. I'm pretty sure the original DART mission's central CPU was a 030.

At heart, the system consists of one Motorola 68030 CPU, two 68302 I/O processors, 2MB of SRAM, and 1MB of Flash RAM

which was capable of real-time "Advanced Video Guidance Sensor"

[WinterLight]

That's pretty close to the processing power of the computers in the original space shuttle.