the benchmark of t1

[zhao-denghui]

Hello, I'm a beginner of t1. I wang to figuer it out that does t1 have a benchmark? if not , how can we compare the performance between t1 and other architectures?

[sequencer]

Thanks for your interest! For now, we have only a bunch of tests, including some MLIR based workload under tests dir, feel free to add your own benchmark!

[shinezyy]

@sequencer Is there an example integration? For example, with Rocket? If RV64GCBV is enabled, it is possible to test with auto-vectorized benchmarks, like some of SPECCPU benchmarks.

[sequencer]

Not currently. Because T1 uses a lot of new chisel features, we just forked the rocket core for integrating it as a demo. But please note we currently don’t support this flow. For compiler engineers, the emulator fully emulate scalar core with spike as well as memory subsystem. It also expose uarch events to users, you can directly use that.

[sequencer]

Additionally, we don’t support MMU because we don’t want to suffer with pf and af or any other memory coherence overhead. We only support machine mode to use this framework as pure accelerator.