Hello, I'm working on methods for simpler and faster cache modeling by batching functional requests locally and sending them to Ruby as needed (L1 misses). I have recently found that purely sending functional requests in many cases has yielded slower simulation time compared to sending purely timing requests into Ruby. Are functional requests inherently slow? Further, I was hoping to find out if there is a way to batch functional requests for speed up? More specifically, is there a way to add multiple read requests or write requests on to a single packet?
For more background, I am hacking the gem5 simulator and feeding traces into the memory hierarchy. I have tested sending all functional requests as opposed to timing requests as to test the simulation time speed of a functional request. I have used a few splash2 benchmarks (LU, water-spatial, and barnes) and found that the simulation speed with only functional requests was roughly 2x slower than the simulation speed with timing requests. Thanks, Paco
_______________________________________________ gem5-users mailing list [email protected] http://m5sim.org/cgi-bin/mailman/listinfo/gem5-users
