Add basic support for tracing

[edgchen1]

Add some basic support for tracing. Currently, we can record durations in a timeline. This is useful for understanding how long different pieces of code take to run.

Tracing can be enabled with CMake build option ENABLE_TRACING.

[baijumeswani]

Could you maybe post a snippet of what the output trace looks like and how it gets rendered through perfetto?

[baijumeswani]

Looks good.

[edgchen1]

Could you maybe post a snippet of what the output trace looks like and how it gets rendered through perfetto?

sure, here's a screenshot from Perfetto UI:

and some of the trace file:

[{"name": "Generator::AppendTokens", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 9},
{"name": "DecoderOnlyPipelineState::Run", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 33},
{"name": "DecoderOnlyPipelineState::RunPipeline[0]", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 2291},
{"name": "State::Run", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 2299},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 10115},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 10150},
{"name": "DecoderOnlyPipelineState::RunPipeline[1]", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 10154},
{"name": "State::Run", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 10162},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 10720},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 10734},
{"name": "DecoderOnlyPipelineState::RunPipeline[2]", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 10737},
{"name": "State::Run", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 10756},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 104758},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 104800},
{"name": "DecoderOnlyPipelineState::RunPipeline[3]", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 104805},
{"name": "State::Run", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 104829},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 230213},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 230246},
{"name": "DecoderOnlyPipelineState::RunPipeline[4]", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 230250},
{"name": "State::Run", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 230279},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 326432},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 326487},
{"name": "DecoderOnlyPipelineState::RunPipeline[5]", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 326492},
{"name": "State::Run", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 326524},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 443239},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 443285},
{"name": "DecoderOnlyPipelineState::RunPipeline[10]", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 443290},
{"name": "State::Run", "cat": "perf", "ph": "B", "pid": 0, "tid": 14528, "ts": 443300},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 449290},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 449318},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 449368},
{"cat": "perf", "ph": "E", "pid": 0, "tid": 14528, "ts": 449374},
...