This is one of the key points that makes TRIPS so special: Very much unlike Cell, or GPUs, or even multicore X86s, TRIPS runs fast with NORMAL, sequential, ordinary C/C++ code. Your old programs will magically get faster, just like the good old days when MHz was increasing. So in short, all your old libraries already work on TRIPS.
TRIPS is also designed to evolve much faster in terms of MHz and # cores, etc. The key point of academia is to be 10 years ahead and tell industry what works and what doesn't because industry cannot afford to fail, so they don't try anything new. Cell and GPUs run faster by taking out hardware that helps programmers - TRIPS finds a way to keep things the way they always were but run faster anyway. It's an "under the hood" architectural enhancement to an existing ISA. (Not that you could tell from the average reporter spin, but the main point is that TRIPS is an alternative to getting speed from multiple cores.)
BTW, I don't get all these "slow development" comments - Intel typically has a research prototype 3-5 years before a chip goes into production. The reason they seem to release a new chip every 2 years is they have several in development at once.
Slashdot Mirror
OK, you can wake up now - we do. :)
This is one of the key points that makes TRIPS so special: Very much unlike Cell, or GPUs, or even multicore X86s, TRIPS runs fast with NORMAL, sequential, ordinary C/C++ code. Your old programs will magically get faster, just like the good old days when MHz was increasing. So in short, all your old libraries already work on TRIPS.
TRIPS is also designed to evolve much faster in terms of MHz and # cores, etc. The key point of academia is to be 10 years ahead and tell industry what works and what doesn't because industry cannot afford to fail, so they don't try anything new. Cell and GPUs run faster by taking out hardware that helps programmers - TRIPS finds a way to keep things the way they always were but run faster anyway. It's an "under the hood" architectural enhancement to an existing ISA. (Not that you could tell from the average reporter spin, but the main point is that TRIPS is an alternative to getting speed from multiple cores.)
BTW, I don't get all these "slow development" comments - Intel typically has a research prototype 3-5 years before a chip goes into production. The reason they seem to release a new chip every 2 years is they have several in development at once.