Would you take this approach?
I had a conversation this evening with a customer and an interesting topic came up. For early software development, especially for firmware, do you really need a Instruction Set Simulator? Could you use discrete-event simulation model with the code compiled to the x86 be sufficient?
There are some advantages to this approach:
1. If the entire code is not available, the unavailable parts can be modeled. This will help in maintaining the full functionality at all times.
2. You can split the code up and make them separate component in the platform. Now if the code is time-dependent or dependent on the correct arrival of the sensor data, the platform could easily modify the intermediate data values. You could also delays between pieces of code.
3. You could create failure scenarios and test the robustness of the code when the code hacked or buffer values are modified or output values from individual code blocks are wrong because of incorrect calculations.
4. You could reuse a high-level architecture model that supports separation of behavior and hardware.
5. You do not pay royalty to the Instruction Set Simulator vendor?
What do you folks think?
#hardware #data #architecture #event #softwaredevelopment #architecture #onARm #riscv
Do we need this complexity?