Jul 182003

Itanium is orders of magnitude more difficult to gen for than AMD 64. That means Borland product management will need to see orders of magnitude more customers wielding pitchforks and cash demanding native Itanium codegen.

I would love to take a crack at AMD-64. ‘Twould be a piece of cake and a pleasure compared to Itanium. AMD-64’s architecture feels like “a good hack” in the old noble sense of the word. Itanium’s architecture reeks of committies. I sincerely hope AMD is successful in positioning the AMD-64 as a gamer’s dream machine. They’ll eat Itanium alive. (Disclaimer: I own stock in AMD and in Intel. Though at current prices, a helluva lot more AMD than Intel)

Disclaimer: My bias against the Itanium architecture is largely due to my sucesses with and fondness for the Alpha AXP RISC architecture. Damn, that was a beautiful architecture. Clean. Simple. And FAST! It’s a shame Compaq killed it in the DEC acquisition.

Can you in some way give us an indication if the estimated effort would be acceptable to Product Management?

Product Management needs to see dollar signs: sales numbers, installed base, sales projections for dev tools. I can’t supply that information to them. They also have to weigh the opportunity cost: the cost of not doing Y because we’re doing X.

I can only estimate the work required for the compiler. A compiler alone does not make a compelling product. You need a debugger, an IDE, libraries, etc. It’s also probably a good idea to have an OS that runs on your target hardware.

Compared to the Java, Win32/x86, and .NET markets, the Itanium and AMD-64 native code tools markets are each so small they easily fall into the “niche” bucket. For Borland to take an interest in a market, we need to see a fair chance of making a fabulous return on investment, or at least a high probability for a certain minimum return. That minimum is measured in millions of dollars per quarter, tens of thousands of units. Are there tens of thousands of people lining up to buy tools for Itanium? Doubtful. There are hardly even tens of thousands of Itanium machines installed. (and most of those are in the Microsoft .NET dev labs…)

Also, the fact that code compiled with Delphi for .NET runs at native execution speed on the 64 bit implementations of .NET currently in development at Microsoft means that Borland management needs to see an even more compelling argument to invest in building specific 64-bit native Delphi products.

Your best bet to get Borland’s attention is direct solicitation. Inform the Borland execs of your interest in certain technologies. Better yet, inform them of market stats that show where the money is. Dale Fuller, Simon Thornhill, Michael Swindell. They do read their email, and they can be persuaded when rationally informed.