Multicore Speedup / Utilization

[AndrewSchmidt]

I have been able to run dual and quad core implementations of various processor designs, both bare metal and with Linux. In Linux I am adding OpenMP and trying to run with multiple threads, but I only ever see 1-core being utilized. I think this is a result of how OVP simulates multiple cores?

I am willing to sacrifice situation time, so if adjusting the quantum would work, is there a preferred mechanism to do so? I have tried to use icmSetSimulationTimeSlice, but I am not noticing any differences.

Any suggestions or comments would be great, thanks!

[LarryL]

Hi Andrew,

I think what you need is likely the Imperas QuantumLeap product, which allows the multiple processor cores in the target virtual platform to utilize the multiple processor cores in the host machine. This works with the Imperas simulation products (www.imperas.com/products), however, not with OVPsim.

As I think you have access to the Imperas simulation products through the formal Imperas University Program, this should not be an issue. Please contact univ [at] imperas [dot] com for more information.

We also published a paper on QuantumLeap at the Embedded World 2015 conference. I'll send that to you.

Larry
_________________
Larry Lapides
Imperas