Reply 80 of 81, by QBiN
- Rank
- Oldbie
wrote:So, do you guys think that maybe Intel meant for the Pentium 4 to be a stop-gap product, with an Itanium compatible desktop design coming up after a few years?
Yeah, I believe that's true as well. When I was a graduate student taking superscalar processor design, the talk was that the penalty for incorrect branch prediction was too great on the existing architecture (NetBurst) to move forward without other significant microarchitectural advancements to compensate for these pipeline flushes.
So the buzzword was "VLIW" or Very Long Intstruction Word. One of Itanium's goals was to make the processor simpler by outsourcing parallelism from the processor (implicit or derived parallelism) to the compiler (explicit parallelism). If the compiler could optimize the code and explicitly predetermine the parallelism, it would be able to pack the optimized instructions in these VLIW instructions. Simpler, faster processors... more intelligent compilers that optimize for exposed hardware parallelism at compile time instead of the processor at run-time.
This talk was happening in the late 90's before Willamette was even officially released yet. So they already knew and were working on Itanium at that point.
edit: typo