Dear Hamid, Thanks for your Interest in Flexus.
The simulation speed that we measured on our Westmere cluster are around 1 MIPS (per core) for functional simulation of a CMP platform (including the memory hierarchy), and around 20 KIPS (per core) for detailed cycle-accurate simulation of out-of-order execution. These numbers can vary a bit depending on your hardware platform, the amount of memory your machine has and the type of storage you have, the exact workload you use, and the complexity of the additional components you may want to simulate. The good news is that our rigorous sampling methodology speeds up the cycle-accurate simulation by several orders of magnitude. After performing a round of functional simulation, you can finish your cycle-accurate simulation within a couple of minutes, assuming that you have several machines to run your simulations on. Let us know if you have more questions. Regards, Djordje ________________________________________ From: Hamid Reza Khaleghzadeh [[email protected]] Sent: Saturday, April 19, 2014 7:21 PM To: [email protected] Subject: Simulation Speed Hi all, I am working on CMP scheduling. My runs must be long, so I need a simulator which it's simulation speed is high. By the way, memory hierarchy must be simulated. Therefore, I want to know how many instructions per second can Flexus simulate? Thanks
