Hi,
Do the ARM Fast Models support native acceleration when running on an ARM server. I am running arm binaries of ARM Fast Models on a high-end arm server but the performance is about 6-7 times slower as compared to running on an x86_64 system. I have specifically compiled a SystemC Virtual Platform for Cortex-R82x1 present in one of the examples in ARM Tools Portfolio. Regards,
Mughees
Fast Models use dynamic binary translation to translate an Arm binary to the instruction set supported by the underlying hardware. It does not attempt to run natively and then trap and emulate illegal instructions. The translation backend for Arm is only a few years old and will improve in performance over time. 6-7 times slower does seem abnormal. Can you send the details of your experiment to suppor@arm.com so we can investigate? Just reference this Community thread within your email.
Thanks for the reply. Yes, we are in contact with ARM through other email channel to investigate. We have provided further example use cases as well.
Hi Mughees, another option to execute R82 binaries that DOES use Arm native acceleration is to use a Corellium model. We have R82 models up and running on our site today and even offer free trials. You can log in using your Arm ID as well. Our R82 models typically run faster than the silicon implementation while maintaining complete functional accuracy. Try it out here https://app.avh.corellium.com