http://hardware.slashdot.org/article.pl ... 24/2021243
Who besides me remembers Doug Burger's talk at CMU?
At first glance, I would say he has significant competition from existing ISAs that do the dataflow in software library rather than embedded in the compiler.
Doug Burger
-
- Grand Pooh-Bah
- Posts: 6722
- Joined: Tue Sep 19, 2006 8:45 pm
- Location: Portland, OR
- Contact:
Doug Burger
Disclaimer: The postings on this site are my own and don't necessarily represent Intel's positions, strategies, or opinions.
-
- Tenth Dan Procrastinator
- Posts: 4891
- Joined: Fri Jul 18, 2003 3:09 am
- Location: San Jose, CA
Wasn't he at Wisconsin or Michigan or something at the time of the talk? From what I remember, this was still going to run into similar issues as superscalar processors in the end since it wasn't really that different overall. Other than the fact that it issues more instructions at once and has more execution units much of the rest seemed pretty much the same. You're right, they're relying heavily on the compiler to be able to handle building the frames (like reordering in hardware but with a much larger instruction window to work with), which is not much different than other software dataflow optimizers. Still, controlling the hardware and the compiler should give them some performance benefit over people just writing optimized libraries that can't change the hardware at all.
I think the most interesting part of this is that it's CS people working on this instead of CE. This is just another instance showing that hardware is becoming commoditized while the real breakthroughs are gonna come from software. It's gonna take a major shift in the fundamental construction of a transistor to swing the pendulum the other way and that's not gonna happen for at least 10-15 more years at the earliest. Being a CE right now seems a bit boring really because it's just more of the same.
I think the most interesting part of this is that it's CS people working on this instead of CE. This is just another instance showing that hardware is becoming commoditized while the real breakthroughs are gonna come from software. It's gonna take a major shift in the fundamental construction of a transistor to swing the pendulum the other way and that's not gonna happen for at least 10-15 more years at the earliest. Being a CE right now seems a bit boring really because it's just more of the same.
-
- Grand Pooh-Bah
- Posts: 6722
- Joined: Tue Sep 19, 2006 8:45 pm
- Location: Portland, OR
- Contact:
Re: Doug Burger
Now he's at Microsoft Research. Interesting place to do architecture research...
-
- Tenth Dan Procrastinator
- Posts: 4891
- Joined: Fri Jul 18, 2003 3:09 am
- Location: San Jose, CA
Re: Doug Burger
I hope he's not working on IA64 type stuff... Really, he's probably the hardware expert for all the software dudes attacking the same problem everyone's starting to take real notice of which is how the heck are we going to we program efficiently for massively parallel systems.
-
- Grand Pooh-Bah
- Posts: 6722
- Joined: Tue Sep 19, 2006 8:45 pm
- Location: Portland, OR
- Contact:
Re: Doug Burger
He is working a next-gen EDGE architecture, basically the next step from what he did at Austin.