Feedback

A forum for readers to speak their mind on issues that are important to the design community

The customer's always right

To the Editor:

Jonah McLeod said in his September editorial ["Who's Got the Power?" p. 6], "I'd like to hear that I'm completely wrong in my assessment of the process by which EDA platforms and tools are sold and bought." I'd like to grant his wish.

As the manager of platform marketing at Cadence Design Systems, my job is to manage relationships with hardware vendors and define our product deployment strategy with regard to hardware platforms and operating systems. In his editorial, Jonah implied that Cadence doesn't consider the opinion of its customer base to determine if users want tools that run under Windows NT. That statement couldn't be further from reality. We at Cadence base our business on input from our end users. We conduct surveys of our system administrators and end users through the International Cadence User Group and the Cadence Technical Advisory Board. The feedback over the last two years regarding OS support has been consistent and has formed the basis of our platform strategy.

The vast majority of respondents have told us that new software functionality and enhancements for our IC tools and flows, which are primarily Unix-based, are their most critical needs. In the latest customer poll, 95 percent of the respondents indicated that Cadence's rate of NT deployment across both our IC and PCB product lines is in alignment with their requirements.

Cadence hasn't made blanket statements of NT availability because we frankly haven't seen demand across all segments of the EDA community. We've focused our NT efforts on the applications in demand--high-performance PCB and Verilog logic simulation--and have reaped dividends as a result. From the first quarter of 1997 to the first quarter of 1998, Cadence saw its revenues for Windows NT-based PCB products grow dramatically in that market.

There has been no pressure from Microsoft. Decisions regarding our platform strategy aren't made by Microsoft, nor by Sun, HP, Compaq, NEC, Dell, or IBM. Rather, decisions are based on customer demand and requirements, reflecting our wide range of users in large, multinational companies--as well as medium-sized to small organizations and two-person start-ups. In fact, Cadence has actually discontinued support for some platforms and operating systems because the customer demand wasn't there. A platform decision isn't something that can be taken lightly. Considerable market research goes into every one.

Today, many software, hardware, and platform vendors are working together to find market-driven solutions. In June, for example, Cadence announced a strategic relationship with IBM to service the needs of the EDA industry as it transitions to a mixed Unix and Windows NT development environment. In September, Compaq and Mentor Graphics announced an alliance to support designers on NT, and Sun Microsystems announced enhanced support and compatibility with NT in certain product lines. Expect the trend of collaboration to continue as companies try to accelerate solutions that are derived from customers' needs.

Philip Laidlaw
Manager of platform marketing
Cadence Design Systems, Inc.
San Jose, Calif.

Jonah McLeod replies:
I would like to thank Philip Laidlaw for taking the time to respond to my editorial, and I appreciate his information on Cadence's relationship with its users and its view on EDA platforms. It was my intent in the piece to provoke readers to reveal their views on their role in purchasing EDA tools and hardware platforms. Mr. Laidlaw's views on what designers use and want reflect what I'm hearing informally from a number of readers. Notably, simulation has migrated to Windows NT because designs are typically hierarchically disintegrated, with individual engineers responsible for simulating the smaller parts. However, I'm told that those parts come together for final full-chip simulation runs performed on high-end servers. I'm also told that layout tools are all Unix-based and will likely remain so until Intel and Microsoft develop 64-bit hardware and 64-bit Windows NT. Even then, layout tools may not rapidly shift to the PC platform.

Not a new OS--a better OS

To the editor:
First, my take on the Linux versus NT debate:

Hate Microshaft--it's interested in getting rich and not interested in quality. Hate Unix,--too user-bitchy. Linux--haven't used it yet but plan on getting it soon, just to try it out. There's no software, so it'll be more of a toy than anything else.

Now, on to your September editorial.

• Other departments look to EDA guys for computer advice: Well, where I work, wrong! We have a Cray for the super-duper number crunching; access to off-site mainframes for large data processing tasks; and a huge complex for storage, analysis of data, and maintenance of a worldwide database. Nobody asks EDA guys what to do. Our software requirements don't even apply to them--not even for those types of applications.
• Most designers lack the authority to sign for big-ticket purchases: Users make a requisition, but can't sign for it? Oops, not quite. Around here, the engineers initiate the purchase request and release it to the system; then it takes a bunch of people to sign off on it from various departments. Technically, the guy who controls the dollars is the only other person who needs to be convinced--no IT department involved, no big corporate standards. The engineer decides, and if he can make a good case, it gets done, period. Very simple.

  Of course, a manager could say:

  "You're an idiot. OK, a PC is cheaper than a Sun box, so we'll save, say, $15,000--but the software you want to buy costs $120,000, and you don't have a whole flow on it, so you still need the @*&%#! Unix workstation, and we've got plenty of Unix workstations and our rent-a-gurus to take care of them. But I've still gotta buy you Windows so you can read and write all the documents flying around, even if you dual-boot it. And, you dumb crud, if you boot Windows when I send you something, your two-day simulation will have to be stopped. Does your software support that? Uptime is no good if you have to reboot to read the #&*@! mail. So that means you'll need three machines: one running Unix for tools, another with Linux (because you're an idiot), and yet another with Windows to do regular office stuff and build our test systems. We tried going with Linux alone, but it didn't work--couldn't read and write all of our files. What's that? Linux scales better for those 3-million-gate designs? Shut up. Ninety-nine percent of our designs are less than 100,000 gates. You should spend more time designing than telling me how you can find some jerk somewhere on the Internet who can send you an operating system patch. Your job is to design hardware, not play with operating systems. Use Unix, Windoze, or Viewlogic 4.0--I don't *&$#$@! care; just finish!"

• As a result, vendors look to corporate MIS and management--the people who write the checks--rather than engineers: Never seen that happen. In fact, it's hard to get the annoying EDA vendors away from us. Get onto their Web site once, download a demo, and you're married for life. They're relentless--at least the ones who want to win. The managers? They're totally lost; when the vendors come in for a presentation, virtually all attendees are engineers. I don't see what you're saying happening.
• Vendors ignore engineers and sell their wares directly to the person that signs the PO: Here's what happens if the vendors do try to sell directly to managers: Before the manager writes a $500,000 check, he'll ask us, "Er, does this make any sense?" If an engineer replies, "No, you old geezer, a fresh-out would know that's stupid," then he ain't gonna buy it. Do bad decisions get made? Sure, but not for the reasons you mention.
• EDA decisions are being made without considering the ultimate user: BS. Synopsys has said repeatedly in public that there's no demand. When we had Exemplar in recently, I said, just to start a tad of trouble, "Er, what about Linux?" The room got quiet. Little anger lines spread on the sales critters' faces. You see, they got burned with Linux because they couldn't sell squat. The tool ported to Linux wasn't even that expensive, so it at least should have been an easy sell from a monetary point of view. But it got zippo response. I recently talked to an EDA software manager about future support for Linux. He said more or less that nope, everybody's happy with Unix and Windoze and added that Linux wouldn't help because it would drive up the need for support. He saw no interest in Linux at all. There's simply no sign of engineers being bypassed, at least from what I see. I went and asked a bunch of engineers myself: Unix, Linux, or PC? Most simply don't care. It's a hassle having multiple machines on the desk. Most people are concerned with the application and getting their jobs done.

  What craftsman in his right mind would design a set of big tools with which you needed 100 little tools lying around to make them work? Why, then, do Unix people tout the hundreds of utility programs as a feature? I want to spend my time designing, and I want other engineers spending their time designing--not dealing with tools all day, fussing with them, and learning all the little kludges. Rather than say, "We need Linux for cheap hardware and to preserve our Rube Goldberg mess of a system that only an idiot would design," we EDA tool users need to put pressure on vendors to make their stuff really plug together with no muss and no fuss. I get paid for output, not process. I don't care if it's Unix, Linux, Exec 8, Domain, Windoze, Java, blah, blah, blah. Most engineers don't, either.

  The guys arguing about the use of arcane tools to keep things working soundly are like the old IBM mainframe programmers who were masters at JCL, which was often harder than the program. Yuck! Ease of use isn't necessarily a bad thing. Sure, the zealots repeatedly say, "Gooey bad, too many clicks." But hey, that's better than "howthehelldoesthiscommandwork?" Also, most of the tools save the state, so that you don't have to reclick everything in. Oh, and all the tools we've looked at for Windoze do sport command line interfaces--but in manual operation, it's far easier to run with the gooey anyway. Who wants to remember all the options? or even to write the scripts? It's easier to type in settings from the gooey, pick what you want, and then hit "Save configuration," which saves everything to an ASCII file.

  To reiterate the bottom line: Who cares what the OS is? Here's what this engineer wants: fewer bugs in the software from both the EDA vendors and Microshaft. The software quality, frankly, sucks. Go read the damn errata lists and ask yourself if you would bet your life that the applications work correctly. I see no correlation between software price and quality, so where the hell does the maintenance funding go? Recall that a decade ago, Unix's reliability was a total joke. Good old, messy DOS would beat it hands down. But those issues are pretty much fixed. Microshaft should get its OS nice and stable and not worry about browsers, puh-lease!. And easier-to-use tools should operate in a reasonable way by default and be tailored as needed. That would lower the learning curves--which are far too steep--for the more advanced tools. You've gotta hit a nice balance among degree of control, steepness of the learning curve, and ease of use.

Name withheld by request

Neither Linux nor NT?

To the Editor:
First, I wish you had modified the heading saying "Fired for Choosing Linux" in the listing of related articles ["Linux vs. NT, Part 1: Readers Speak Out," July, p. 18], because it was a false and misleading headline in the first place. The article didn't report anything about anybody being fired for choosing Linux. By listing it as is, and in bold, you're propagating the unfair tactics the author of that article, a pro-Microsoft "journalist," unethically used.

Second, a major problem for me with Linux is the lack of a single consistent GUI. It seems ironic to me that the Linux supporters trumpet the availability of four GUIs, when in actuality, that's one of the things holding Linux back. I don't want to have to learn and remember four different GUI styles, and I don't think anyone else does either. If all application developers for Linux would get together and choose one, then all the apps would have a consistent look and feel--which, by the way, is one of NT's major current advantages.

The lack of EDA apps is definitely the single major drawback, as others mention, and I don't see the major EDA providers ever fixing it. Changing offerings would work against their desire to keep their prices so high. If possible, Red Hat and its peers should get together and fund a version of Linux, or a "layer" for Linux, that makes it look just like a Unix flavor that the EDA tools do run on. Don't wait for the vendors to fix the tools; fix the OS instead. That strategy has the other obvious advantage of allowing everything--not just the EDA tools--from the "major" Unix OS to run on Linux.

The EDA software that I've used on NT, by the way, has been buggy as hell. Are the crashes caused by the application or by NT? I don't know. All I know is that it crashes all the time, without any indication of how to fix it, and without any kind of repeatability to report to "support."

Some claim that NT's context-sensitive help screens make configuration much easier. Give me a break! Have they ever actually read those screens? I've never seen such poor excuses. They rarely tell you what to do to fix the problem; thus I call it "half help." However, I must say that configuring Linux doesn't look like fun either. I've just spent a half hour on my first visit to the Red Hat Web site, and what I saw for configuration was a dozen text files full of programmer gobbledygook. Again, Red Hat is seriously hurting itself by failing to address a major ease-of-use issue. I believe it's continuing to look at its product from a programmer's perspective. That's fine--if it doesn't care about Linux's success. But if Red Hat wants the other 99 percent of the user universe to embrace Linux too, then it really ought to do a crash development of a complete window-based installation and configuration based on "choice boxes," rather than the current complex syntax typing.

But don't let NT off the hook! Have you ever tried to understand or edit an NT registry file? Holy smokes! It's fine for C programmers, but a nightmare for the rest of us who only want to lay out PC boards. It would be fair to say that the configuration of NT is every bit as arcane and complex-syntax driven, but it's just buried where the user doesn't see it. That's an advantage during installation, but a huge disadvantage when something "breaks."

Finally, why whine about no single company supporting Linux? Microsoft's support stinks. It's the most arrogant, poorly managed company I've ever had the misfortune to deal with.

Richard Hager
Chief executive officer
Precision CNC Machine Tool Controls/
Ah-ha! Design Group, Inc.
St. Paul, Minn.

The title of the original ZDnet Anchor-desk story (www.zdnet.com/anchordesk/story/story_1774.html) was a question, and the article alerts its readers to the possibility that choosing Linux might get them fired under certain circumstances. We erred in omitting the question mark.

NT stands for 'Natural Tendency'

To the Editor:
Your September editorial made me laugh! I want to emphasize that having engineers use the same operating system (Windows) that nonengineers use lowers costs. Companies can standardize on software so that they have to train support people on only one package. They can buy inexpensive Windows software instead of expensive Unix or Linux software. And please don't say Linux software is free and Windows software is expensive--there's plenty of Windows shareware and freeware available for people comfortable with low or no support software. Engineers who prefer Unix and Linux do so because it's what they're used to, not because it's inherently better than NT. That's just human nature. After switching to NT they'll fight switching to another operating system, even back to Unix, just as ferociously as they're fighting the switch to NT now. Is it really reasonable to ask companies to spend thousands of dollars per year per engineer--indefinitely--so they don't have to accustom themselves to a new operating system? I would be surprised to find that any company knowingly did that.

Name withheld on request

Windows EDA addition

To the Editor:
I would like to bring to your attention that Time-Rover was not included in October's Windows EDA Focus Report [p. 48]. Our product, the Temporal Rover, is a specification-based verification tool for C, C++, and Verilog.

Doron Drusinsky
President
Time-Rover, Inc.
Cupertino, Calif.

We regret the omission. Details of Temporal Rover are now available in October's on-line edition at www.isdmag.com/EEdesign/focusreport9810.html, under the "Simulation & Related Tools" category.

Visas unlimited

To the Editor:

I would like to respond to two points made in your July editorial ["On the Politics of High-Tech Workers," p. 6]:

1. High-tech workers are indentured. That will happen only if the current quota isn't increased--as there will be fewer H-1 visas available per year, thus preventing a foreign engineer from switching jobs. Two factors limit the ability of an unhappy engineer to dump his or her employer: the number of jobs and the number of visas. Presumably jobs are in good supply, so visa restrictions artificially force an engineer to become indentured. Otherwise a foreign engineer, just like a native engineer, can hire and fire an employer. The solution: Don't limit the visa numbers to encourage indenturing of foreign engineers.
2. Cheaper engineers from overseas will take jobs from able U.S. engineers. If the United States doesn't produce a critical mass of engineers in time to dominate a new technology, then it won't gain a lead over other countries. Thus a U.S. company must either open offshore development centers or let the business slip to its competitors. In any case, the existing base of U.S. engineers, well versed in older technologies, doesn't benefit. The key is continuous training and open competition rather than the use of legal means to curb the supply.

This is a global economy, and using immigration laws to prop up salaries will at best provide a temporary solution, because companies are driven by the profit motive. I'm not speaking for Intel, but I've been here for over 10 years after joining Intel on an H-1 visa. During this period, I not only completed a Ph.D., but also obtained U.S. citizenship with Intel's sponsorship. We've seen other closed-minded societies, such as Japan and Russia, suffer because of laws that restrict supply in their markets. In contrast, countries like China and Hong Kong are flourishing because of their open market policies.

Name withheld on request

To Jupiter and beyond . . .

To the Editor:
In his October editorial ["Man vs. Machine: The IC Designer's Dilemma," p. 8], Jonah McLeod mentioned that in 2001: A Space Odyssey the human eventually conquers the machine. True, but Ku-brick's subtlety obscures the real subtext: Without HAL to control reentry into suspended animation, astronaut Bowman may not be able to survive until a rescue ship arrives. Bowman's best choice is to venture into the unknown and ultimately to be transformed.

On another subject, I've understood that the term "tape-out" comes from a somewhat different source. Originally, rubyliths were photographically reduced to create plates. The release of such artwork might have been called "delivery of artwork" or "artwork-out." The term "tape-out" came into use when the electronic generation of photomasks was developed. In this process, coordinates of polygons on the rubylith were captured on digitizers. The digital information was sent out to the mask shop on magnetic tape--thus the term.

Brad Martin
Managing partner
North Shore Circuit Design LLP
Austin, Texas

Jonah McLeod replies:

I stand corrected about "tape-out." What I'd heard must have been erroneous folklore.

To voice an opinion on this or any Integrated System Design article, please email your message to miker@isdmag.com.

integrated system design  December 1998