does there exist a good writeup of the Itanium story?
-
@regehr Have you read Bob Colwell's oral history?
> The Intel iTanium project started with the Multiflow compiler. They gradually
rewrote pieces of it to make it go faster.
https://www.sigmicro.org/media/oralhistories/colwell.pdf
ARF from recent times: https://www.abortretry.fail/p/the-itanic-saga@sdbbp no! but I will, thanks!
-
-
Intel has a tradition that every ten years they build an architecture that it is impossible to write a compiler for. The iAPX432, i860, and Itanium were each iterations of this. The next one was a GPU architecture with a two-dimensional register file (amazing for hand-coded assembly kernels, not great for anything that needed to merge lowering of vector permutes with register allocation). I didn’t pay attention to what they did most recently, I presume they made some AI accelerator that has amazingly high FLOPS numbers on paper and is impossible to target from MLIR.
-
@koakuma @regehr @rygorous now that reminds me of Nuon! @llamasoft_ox
-
Intel has a tradition that every ten years they build an architecture that it is impossible to write a compiler for. The iAPX432, i860, and Itanium were each iterations of this. The next one was a GPU architecture with a two-dimensional register file (amazing for hand-coded assembly kernels, not great for anything that needed to merge lowering of vector permutes with register allocation). I didn’t pay attention to what they did most recently, I presume they made some AI accelerator that has amazingly high FLOPS numbers on paper and is impossible to target from MLIR.
@david_chisnall @koakuma @regehr The iAPX 432 from what I can tell was not at all impossible to write compilers for, but as per Bob Colwell who wrote a bunch of papers dissecting the 432 for his PhD, the compiler team did not get along with the HW team and explicitly didn't want the effort to succeed. https://www.sigmicro.org/media/oralhistories/colwell.pdf p. 51
-
@david_chisnall @koakuma @regehr The iAPX 432 from what I can tell was not at all impossible to write compilers for, but as per Bob Colwell who wrote a bunch of papers dissecting the 432 for his PhD, the compiler team did not get along with the HW team and explicitly didn't want the effort to succeed. https://www.sigmicro.org/media/oralhistories/colwell.pdf p. 51
@david_chisnall @koakuma @regehr incidentally, Bob Colwell was at Multiflow working on (and shipping) commercial VLIW HW and then later went to Intel becoming Chief Architect of the PPro. (What he thought about Itanium given his background shows up later in the same PDF.)
-
@david_chisnall @koakuma @regehr incidentally, Bob Colwell was at Multiflow working on (and shipping) commercial VLIW HW and then later went to Intel becoming Chief Architect of the PPro. (What he thought about Itanium given his background shows up later in the same PDF.)
@david_chisnall @koakuma @regehr Mind, there's no question that the 432 was still bad in many ways, but the issue was less "good compilers are impossible to write for this" and more "there were many unforced mistakes that the compiler people tried to raise awareness about and got shut down, so they stopped caring"
-
@david_chisnall @koakuma @regehr Mind, there's no question that the 432 was still bad in many ways, but the issue was less "good compilers are impossible to write for this" and more "there were many unforced mistakes that the compiler people tried to raise awareness about and got shut down, so they stopped caring"
@david_chisnall @koakuma @regehr (I will say that during the relatively short window I contracted for Intel, this tradition of organizational dysfunction with different teams ostensibly on the same effort actively trying to sabotage each other was alive and well)
-
@david_chisnall @koakuma @regehr (I will say that during the relatively short window I contracted for Intel, this tradition of organizational dysfunction with different teams ostensibly on the same effort actively trying to sabotage each other was alive and well)
I saw some of that when I was at Microsoft. The pattern for Intel folks was:
- Pitch something to the Windows team, lying outright about expected performance if that helped.
- Get the Windows team to say they wanted it.
- Put it in the ISA and get it made plan of record for the next core.
- Get promoted.
- Move to a different company in a more senior role on the basis of that promotion.
- Be long gone by the time everyone notices how much of a disaster it is.
-
R relay@relay.infosec.exchange shared this topic
