Multiprogrammed workloads

From gem5
Jump to: navigation, search

Running

In SE mode, simply create a system with multiple CPUs and assign a different workload object to each CPU's workload parameter. If you're using the O3 model, you can also assign a vector of workload objects to one CPU, in which case the CPU will run all of the workloads concurrently in SMT mode. Note that SE mode has no thread scheduling; if you need a scheduler, run in FS mode and use the fine scheduler built into the Linux kernel.

Terminating

There are several options when deciding how to stop your workloads:

  1. Terminate as soon as any thread reaches a particular maximum number of instructions. This is equivalent to max_insts_any_thread. The problem here is that multithreaded programs are non-determinstic, and there is no way to determine how many instructions the other threads will have executed. The amount of work done per instruction could change as well if more or less time is spent waiting for other threads. The benefit of this approach is that all threads are running fully until the simulation terminates.
  2. Terminate once all threads have reached a maximum number of instructions. This is equivalent to max_insts_all_threads. With this option you can be sure all threads do at least a certain amount of work, but threads that reach the maximum continue executing and there's no way to know how much extra work these other threads will do.
  3. Terminate each thread individually after it executes a particular number of instructions. This option is not currently supported. All threads will do the same amount of work which avoids some of the problems mentioned above, but now the threads may not all be running for the entire simulation. After some threads have finished, the remaining threads will have less competition for resources, and the compute resources will have less work to do.
  4. Terminate each thread individually after it reached a particular, per thread number of instructions. This is basically the same as the previous option but allows tuning the number of instructions executed per thread. It is also not currently supported.

Merging Single Workload Checkpoints