Use 2 FPGA chips... first(CONTROL) controls proccess timing and the second(PROCESSING) does the processing.
CONTROL looks at the execution stream of the current process, then reconfigs the gates of PROCCESSING to match the *optimized* %^) HW arch. for the process, then PROCESSING executes it while CONTROL looks the the next stream.
But, I talk from out my annal... total speculation... and as far as I know (which isn't much &^) a round-robin version of this idea, without the HW optimization, is how normal SMP works.
I think you are totally right on... That was my first thought as well...
As to SA, I think the first article I read about FPGA's was in SA about 2 years ago... I'm diggin' out my paranoid bone here, but I find it odd that it made such a small splash when it came out... my first thought was "this will rock the UNIVERESE!!!" hhheeemmmm...
I'm also wondering if Transmeta has bought a little talked-about NIC card company that was listed in the Linux Journal a few months back(forgive, I don't remeber which month).
Slashdot Mirror
CONTROL looks at the execution stream of the current process, then reconfigs the gates of PROCCESSING to match the *optimized* %^) HW arch. for the process, then PROCESSING executes it while CONTROL looks the the next stream.
But, I talk from out my annal... total speculation... and as far as I know (which isn't much &^) a round-robin version of this idea, without the HW optimization, is how normal SMP works.
mia $.02
--shift8
As to SA, I think the first article I read about FPGA's was in SA about 2 years ago... I'm diggin' out my paranoid bone here, but I find it odd that it made such a small splash when it came out... my first thought was "this will rock the UNIVERESE!!!" hhheeemmmm...
I'm also wondering if Transmeta has bought a little talked-about NIC card company that was listed in the Linux Journal a few months back(forgive, I don't remeber which month).
wha-hurd,
--shift8