Hacker News Reader

Comment by eismcc | Hacker News Reader

I worked in the windows kernel team and my favorite story about DC is when he basically made x64 happen because he hated Itanium architecture so much. He worked with AMD and basically made it happen while cranking in his corner office.

View on HN

What's intriguing to me is that Itanic was the "fetch" that Intel tried to make happen not once, not twice, but three times: once as iAPX 432, once as i860, and once as Itanium. The whole idea of "pack multiple (often variable bit-length) instructions into a word and rely on compiler writers to determine optimal instruction packings" to make implementation simpler and therefore hopefully faster, but woe betide thee if thy compiler dost not find such optimal instruction packings; for then thy CPU will actually run code more slowly than a conventional design like x86! Someone very high up at Intel must've been very in love with this idea, as it was tried and flopped thrice throughout Intel's entire history as a major microprocessor vendor.

I didn't think AMD64 was a good idea at the time because it was more of the same. But good on Dave for helping eliminate the cruft that pretty much killed the RISC revolution.

Funny in contrast to all the "I want the cpu to do what I tell it and nothing else" comments in every thread about speculative execution.

Thinking about instruction packing and if this is something a JIT could have helped with, if we'd had state-of-the-art JIT's at the time.

Unfortunately it can't work well because memory latency is so high and variable.

It has been used in GPUs though.

One of my favorite technical papers covers the AMD x64 inception -- it's gold if you're into low-level OS internals: https://github.com/tpn/pdfs/blob/master/A%20History%20of%20M...

There’s a little segment where he talks about that! It blew me away because he so casually mentions how it happened as a side project. I hope someday I can manage to put something that impressive together in my free time.

I was an SDET on the Windows team in the 2000/XP era and heard that the internal code name for the x64 port was “Sundown” because Microsoft was hoping it would take down Sun in the server market. Unfortunately for Microsoft, that prize went to Linux on x86-64.

Microsoft also wanted to "destroy Sony" in gaming and so codenamed Xbox and DirectX after a WW2 battle that Japan lost and the US atomic bomb project.

Presumably the resulting bad karma is why the Xbox never sold well in Japan.

Reply on news.ycombinator.com