Simics Accelerator

Simics Accelerator allows the simulation's host-processing and host-memory resources to be easily extended in order to satisfy increasing model complexity and performance requirements. It does this by distributing the simulation workload across multiple processor cores on a single host, and across multiple host computeres. The net effect of Simics Accelerator is a dramatic increase in speed on large target configurations, often linear in the number of processor cores used to run the simulation.

Using Accelerator, Simics is the only commercial simulator capable of running massive virtual platform models containing scores of boards and hundreds of target processors - at a speed that is fast enough to satisfy the needs of the average systems engineer, software developer, or systems tester.

Multithreaded Simulation

Simics Accelerator makes use of multiple host processor cores to accelerate the simulation of large target systems. It does this by running parts of the simulated system in parallel, without impacting the determinism, synchronous system stop, and well-defined semantics provided by regular Simics simulations. Typically, the unit of parallelism are individual boards or networked machines, which are worthwhile to simulate in parallel since

Simics Accelerator Fully Leverages Multicore Platforms

The use of parallel cores can be controlled by a user to allow multiple simulations to share a large server.

Distributed Simulation

Simics Accelerator 2.0 can also use distribute the simulation across multiple hosts, to further increase the simulation power available for truly large simulations.

Simics Accelerator Uses Multiple Hosts to Speed Simulation

Page Sharing

Virtutech's Page Sharing technology, included in Simics Accelerator 2.0 improves simulation performance and simulation-host memory utilization by allowing target memory and target disk pages containing identical information, to be stored only once and then shared with each of the relevant components within the simulation.

Page sharing technology is especially effective in eliminating duplicate memory pages and reducing the model's memory footprint in simulations of multicore/multinode systems where multiple co-processors (or accelerators) run identical software loads. Such an architecture is quite common in many of today's multicore network-equipment designs.

Summary