>If we as field can succeed at this amazingly difficult challenge, the future looks
>good. If not, then performance increases we have relied upon for decades will come
>to an abrupt halt, likely dimishing the future of the IT industry.
>

Part of the problem is the lack of access to basic hardware primitives from the software side. Too much emphasis is placed on the value of software transparency - which is a real obstacle for software people to do anything really interesting. Just look at anything that Intel does. How can you handle more than a few cores effectively when anything related to communication is off limit to software?!

IBM is better at exposing hw capabilities but they don't seem to want software to run on their platforms - other than what they can control, say linux. Take a look at cell, the most interesting part is the ring bus, but they kept the spider south bridge secret. The message is clear - don't even think about doing anything before talking to the gate keepers. sigh.

It's understandable that architecture people are frustrated. If having eight pipelines doesn't work, spliting them into separate cores won't work either. There seems to be nothing really going on that can stop the commoditization.

How about just open up a little, and let software people do their job - which they will. 2c from a software guy.