The CAD Multi-Core Processor Problem

vinny wrote:


It was in 2001, dumass. The drawings were produced with CAD. The 'Print Room' archived and controlled the issue of paper copies of APPROVED drawings. Every manufacturing document ever produced was on file in their fireproof vault, and all but the earliest were created with electronics CAD programs. There were several used, over the years as companies went out of business, and there was no support for newer hardware.
You obviously know nothing of building space certified Telemetry and communications electronics at an ISO 9001 certified company. Company policy was that only A or B sized prints were to be avaible outside of engineering. That was fine when everything fit on B sheets, but VME based boards with 5,000+ components were added to our product line it was impossible to read A or even B sized prints. That policy was put in place in the early '70s, when it made sense. Your solidworks type mechanical CAD software was useless, except for case & moddule housing designs. the VME cards had 17 layers of copper pours, blind vias and had to meet the standard VME form factor.
Some of the RCB200 series VME boards had over $8,000 in componets before they were placed & reflowed. With over 20,000 solder joints it was difficult to get a run where every joint had the proper wetting & flow from the 80/20 paste solder.
From the other ignorant, electronics related crap you've posted, it doesn't suprise me.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

I've had to deal with that too. B sheet schematics for complex VME boards ran to over 30 sheets. It works if the draftsman understands the circuit, left-to-right signal flow and what to group together. We drew them as C size.
The most difficult schematics I've tried to decipher were of TTL random logic drawn in pencil. On one the RESET line made three complete loops.
jsw
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Jim Wilkins wrote:

At the end, the block diagrams were being drawn in CAD by the engineer, and finished by the drafting department. Nothing was in pencil after about 1970. Even test fixtures were cad drawings, if any existed. I can believe the triple loop. What's even worse is when they design for a real reset circuit, then build with a simple R/C reset that fails after a few years. I ran into that in some broadcast video editing interfaces (For our three Sony 1" R-R VTRs) at an Orlando TV station. To add insult to injury, the PC boards were laid out for 555 timers for reset, and they had just jumpered around them to save a couple cents. I removed the crap and added the 555s to make a permanent repair.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

So let me try to pice this together... Your some super genius hand picked by the international space admin to create space age circuits and in 2001 you don't even get supplied with a pc? Not even a terminal?
The drawings were produced with CAD. The

But you know if I did know some super specific specialized crap like that I'd try to bulldoze people with technical dogma to hide the fact I couldn't count to ten without a slide rule.
Company

Your solidworks type

How ridiculous. A bunch of wires on layers or individual files representing layers. Scary stuff man.

pffft. They print crap like that a million times a second now.

Why ya on disability? Circuit board fall on ya?
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Vinny, I don't have a dog in this fight. But let me remind you that a lot of design shops don't allow individuals to use up consumables. If you want a print, you go to "the print shop".
A two or three person business might allow anyone to print E-sized plots any time they opt to do so. A "regulated" shop might have different rules.
Hell... I worked as a computer technician for a large accounting firm once. I had my own PC. I had all the softs and all the tools. I couldn't get _anything_ on paper unless it went through approval, and the "print shop" did it. Partly, it was for the purpose of document control. Every request for hard copy went through an approval process (brief) that checked to make sure the requestor wasn't asking for something they didn't WANT on paper. The other thing was the deliberately restricted physical resources (to control the above reason).
The only printer (for employee requested documents) in the place was a monstrous Xerox rig that filled up half a room and collated, stapled, sorted, and automatically jogged books into a pile. You sent your job into their cue, and they printed it when they got to it.
I don't know what circumstances he worked under, but I know that such circumstances as he described do exist, because I worked in that environment.
LLoyd
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
"Lloyd E. Sponenburgh" wrote:

The only thing I could print was datasheets I downloaded. My 'printer' was a huge Xerox network printer a couple hundred feet from my workbench. It printed double sided, and could punch a printout for three ring binders. I had to watermark them as 'For Reference Only', or they would be removed from the production & test area and destroyed.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
"Lloyd E. Sponenburgh" <lloydspinsidemindspring.com> wrote in message

At MITRE I ran the department's CAD network, electronics lab and machne shop. I could plot prints any size up through E for internal use within the department , but to release them externally they had to be A or B sheets, since they would be the appendix to an A size printed document.
In practice sheets larger than B didn't fit on a crowded lab bench, or violated the ESD separation rules. I could Xerox a marked-up lab copy B sheet much more easily than a larger one, partly because folded, wrinkled, dogeared drawings didn't feed well through the large copier whereas all the secretarys' copiers took bound/stapled B size flat on the glass.
I learned to adapt to B schematics although I had been drawing D sizes for machines built out on the shop floor where there was room to spread them out on a table, and E sizes for cable harness assembly boards.
jsw
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Jim Wilkins wrote:

I kept the large prints on a shelf, and only pulled what I needed to troubleshoot a board. They were used on a shop cart with a particle board top that was slightly larger than a D print.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
vinny wrote:

Pice all you want.

You wouldn't be picked by anyone who read your illiterate rambling.
I worked for a long time manufacturer of Aerospace Telemetry equipment. NASA would simply tell us what they needed, and we would supply it. We also sold equipment to NOAA to control their weather satellites, and two earth stations for the European Space Agency. One was fixed site, the other was mobile in two European military style trailers.
Sigh. I had several networked computers on my bench, but you can't make notes on 'controlled document' computer files. The $80,000 Telemetry receiver had an Ethernet port as well. The screens were too small to display what was needed to follow a circuit, when the print was a 'D'
There were multiple networks in the facility, and some were not connected to the outside world.
Serial terminals were used on the MRP system to track contracts, parts and assemblies. Do you know what MRP means?

You'd never make it past the interview, and every design went through a design review. HR pre screened all applicants, then engineering interviewed each candidate before they were hired. You had to have a good track record in the industry before you got past HR. I had repaired some of their equipment in the field when I worked in Cable TV & broadcast engineering. The head of my department was amazed when I told him the units I had repaired in the field, without factory support. Hell, I helped a scientist in Antarctica repair a telemetry receiver that was damaged aboard the supply ship, via the internet so they wouldn't be without it for a year.
I've never claimed to be a genius, but I tested out of a three year engineering course while in Basic traning in the US Army. I received the highest score on record, and was the only one to ever test out of that course at Ft. Monmoth. The average score was 20% correct. My score was 93% correct.

It would be, to an idiot like you. There were two man years invested in that design. Proper trace routing is critical in high speed digital circuits.

Sure, they do. Tell me who does $80,000,000,000 in the Telemetry business each second.

I am a disabled Veteran. Do you have a problem with that? I was offered a discharge while on active duty, and told them to stuff it. I finished my time in service, then found technical work I could do. I managed to keep working till I was close to retirement, and will soon be on SS.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

so...youv'e basically never worked in private industry, yet you "assume" all the real talent is at nasa or other bloated tax payer carried organisations? What I see is people too wrapped up in themselves to even begin to see what's going on out there.
How the hell do you know what others have or havn't done, all you focus on is yourselves. For the past 20 years, every significant thing has come from private industry. WTF has nasa done? The Hubble? pffft, private industry did all the hard work.
Sorry if I'm not impressed.
Now go ahead and spell check and grammer check my rabble if it satifies some kind of schoolteacher urge.
And I still say this multicore single chip pc design sux. The standard motherboard design is the bottleneck.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

You continue to be sidetracked rather than focusing on understand what makes history based CAD such a pain in the ass to use and what the real problems are.
The bottlenecks are listed here but as per usual you're too lazy to do the reading:
http://www.cloud-invent.com/Vision.aspx
**** wtf? lol
dood, if your going to continually paste the same damn thing, at least have the decency to fix....nevermind.
History based cad. Is it good or bad? It's really good if your the one who modeled it and now have to make changes. But you have to model a certain way for that to be true. Direct modeling is an awesome thing when it's not your model. Every cam system needs direct modeling. To go to a non history based modeler with the same options for editing requires a lot more cpu power. Maybe if you redesign the modern pc motherboard and it might be worth it. At least in the pc market.
I wish I had direct modeling in the esprit version im running. Pretty hard to adjust stock on dumb models with no history.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

"Is it good or bad?"
Get off your lazy azzzzzzzzzzzzzz and start reading the links I post. When you finally make the effort you won't be asking this type of question.
"It's really good if your the one who modeled it and now have to make changes."
Wrong. Long regeneration times in history based CAD are a problem even if you're the one who modeled it. Depending on the amount of features even very simple parts can take too long. Ask me for an example video that very clearly shows this if you don't believe me.
"Direct modeling is an awesome thing when it's not your model."
It's an awesome thing even when it is your model. Why limited the tools you can use?
"Every cam system needs direct modeling."
I've said this for many, many years here and elsewhere.
"To go to a non history based modeler with the same options for editing requires a lot more cpu power."
It requires a lot more than just CPU power. Get off your lazy azzzzzzzzzzzz and start reading. The bottlenecks are listed and explained.
"I wish I had direct modeling in the esprit version im running. Pretty hard to adjust stock on dumb models with no history."
If you had the newest version of Esprit you would have the same problem.
******* All your bitching about mastercam....esprit is worse man. imagine you bring in a model for wire edm, select some faces, and post. problem is there's no way to adjust those faces for stock. Your stuck with what you have. HOW FRIGGEN STUPID! You need to adjust different things, like press fit vs. slip fit. What you end up doing is breaking the program up into groups, and adjusting size at the machine. Direct modeling would fix this, or something as simple as an offset value applied to the toolpaths. There's nothing.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
vinny wrote:

Do you even read what you reply to? It was 'private industry' who sold some of their equipment to various government agencies.

What would that be? I've worked in most areas of electronics, starting at age 13. My only 'Government work' was six years in the military.

What the hell have you ever done? A lot of modern electronic is the spin-off technology of NASA and the military's need for better electronics.

It's obvious that you're sorry.

Technical writer, not school teacher, even though I've done both. No need to point out every mistake, since you're so proud of them.

Typical ignorant remark of the clueless. If that's 'the bottleneck', tell everyone the solution. The board design is RF layout, at current processor speeds. That problem became obvious in the '70s with large PCB backplanes. That was when they had to revise the board layout methods, to make identical trace lengths for the data & I/O bus. The 'hobby' computer market had problems with the S100 bus, which was the first 'standard' when physics kicked their ass when faster processors were tried on the existing layouts. Faster was > 1 MHz.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
would that be? I've worked in most areas of electronics,

I was waiting for the working since 13 gimmic. Actually I thought you would say 12.

Not a whole lot. i'm just a humble toolmaker.

Oh jeeze. A technical writer, well then, there ya go.
even though I've done both.
of course you have, youve done everything, invented everything. lol
No

ok, i agree, It cannot be improved upon. We are at the pinacle of development.
technical writer...oh man this is rich!
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
vinny wrote:

Working is a gimmick? I was repairing electronics for spending money, and selling used TVs as a sideline. It was a hell of a lot better than mowing lawns, and paid about five times as much.

Humble? In what universe? You make claims about fields you have no clues about.

I've never made that claim. I did teach 'Electronics One' during the school day and 'Small Appliance Repair' for night adult education at Lemon-Monroe High School in the late '60s. The technical writing was done on test procedures for Microdyne/L3-Com. They were written at a level that even you might have been able to read them. I was tasked with walking new technicans through every step of a jib they have never performed.

You don't understand the physics of making semiconductors, or that the multilayer PC boards that you damn. The conductors in ICs are already approaching a few atoms wide Are you claiming you can split silicon atoms, to narrow their width? Maybe you are going to orient them till they all stand on edge?
Those boards allow the construction of high speed, impedance controlled data bus. Without them, you would have to use silver plated subminiature semi rigid coax, like Amdahl did on some of their mainframes.
If you think you have the answers go to news:sci.electronics.design and tell them what fools they are. You'll find IC designers, and several engineers who worked for IBM, developing state of the art CPUs.

This, from someone who doesn't capitalize the start of every sentence.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

If your so more intelligent than the rest, why is it so easy to push your buttons?
appliance repair, this keeps getting better....
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
vinny wrote:

Yes. I was 16, and teaching adults. Unfortunately, you're unteacahable.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
In article
[snip]

The PowerPoint "Cheetah Presentation" briefing "Beyond Parametric" seems to have the core ideas.
Interesting, and plausible-sounding, and O(n) scaling is a holy grail in many fields.
But the algorithms are proprietary and held close, so it's hard to know what to believe. The default in such cases is to reject the claim as probable hype. Big claims require big proof.
Are there any independent validations of the claims when the product is used on full-scale problems?
Joe Gwinn
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
In article

I did see these examples, which are pretty simple compared to the drawings of say an airplane engine, but are well crafted to cause history-based solvers to crumple.
Anyway, the cloud-invent website is very clear that this technology is at the early prototype stage, and is nowhere near to a real product. Their stated business model is to provide technology to the current major CAD vendors.
There were some interviews of the principals in blogs, from which I gleaned that the ~O(n) scaling is achieved by using banded sparse matrix methods, and the key is how one formulates the problem to ensure that one always gets a suitable matrix, one that can be solved in O(n) or so. Now this is standard outside the CAD world, and you can bet that the majors are now all looking into this. There is no shortage of Mathematicians, so much will come of this. For more information, google on "sparse banded matrix solution" or the like.
Joe Gwinn
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
In article

That's trees for you. But they will always be with us - too much embedded base.
I'm pretty sure that Recompute does the entire tree from the top, rather than incrementally updating the tree based on what last changed.

I don't know. People are well aware that one can easily crush a CAD program, so I don't know that any conspiracy can work.
There is no reason that the current tree description cannot be expressed in a banded matrix or block banded matrix, and the principals implied that this could be done, but for unknown reasons was not.
By the way, the cloud-innovate approach also directly supports parallel solution, which will also allow multicore solutions to be practical.
So the outcome in the market is far from clear.
Joe Gwinn
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.