search

neurosim / DNN_NeuroSim_V1.4

Benchmark framework of compute-in-memory based accelerators for deep neural network (inference engine focused)
53 stars 12 forks source link

SRAM onoffratio setting suggestion #5

Open alfa871212 opened 10 months ago

alfa871212 commented 10 months ago

As the manual says, the SRAM onoffratio might have issues. Do you have any suggestion for the setting?

neurosim commented 10 months ago

We do not recommend using NeuroSim as-is to determine the inference accuracy using analog SRAM compute-in-memory. We currently only support the inference accuracy estimations for memories that follow Ohm's law and Kirchoff's current law to compute the MAC operation (e.g. RRAM, FeFET). Analog SRAM CIM uses a different mechanism for the MAC operation and thus are not currently supported in our inference estimation framework. This is why in the manual we set the on/off ratio to infinity for the SRAM benchmarks.

alfa871212 commented 10 months ago

So, what is the setting to re-generate the table in the manual (SRAM)? Or we just use the SRAM to compare with NVM CIM?

neurosim commented 10 months ago

You can use --onoff float('inf') to generate the results from the table. Using a sufficiently high number (e.g. 1M) should also have the same result.