4 ms·

Cycle-accurate simulation as a software development vehicle while you wait for the real hardware to become available. Get your software debugged now and you wil

• wyldfire 8y ago
  I think cycle-accurate simulation is critical for developing a RISC-V chip and/or toolchain -- being able to get real feedback from benchmarks, etc. For functionality testing (to port your software to RISC-V) you probably don't care quite as much about the cycle accurate simulation. An inaccurate simulator would probably just behave the way that a target with a much faster or slower clock speed would behave.
  • dbcurtis 8y ago
    Sure. First step is "make it work". A functional simulator is sufficient for that. If you are working on a compiler back-end, then the ability to see actual pipeline stalls, to run benchmarks, to investigate the impact of various text and data section layouts on cache performance, then a more accurate simulator is useful. Or if you are working on time critical features of the OS, like a context swap, then you definitely benefit from the ability to do a cycle-accurate instruction trace. What the cycle-accurate simulator buys in end is schedule acceleration for when a mature RISC-V software stack lands. So from an overall RISC-V program-management perspective, this is smart thing. Granted, the audience is small, but the resulting leverage is huge. Imagine being the PR person having to respond to press that says: "I tried this basic thing on your new chip and it is slow." and having to say: "Just wait until we turn on the optimizer in the C compiler! It will be 35% faster!". Reporter: "Yeah, whatevs." This simulator is all about product launch readiness.