Closed fknorr closed 9 months ago
Check-perf-impact results: (5a19ced85f862a00d0114dd241122462)
:question: No new benchmark data submitted. :question:
Please re-run the microbenchmarks and include the results if your commit could potentially affect performance.
Yeah, we specifically discussed the buffer naming lifetime concern when I was building this, which is the reason for the design. I generally prefer that the information in the record is as self-contained as possible once it is recorded, and doesn't have dependencies on other data sources.
Thanks for the feedback - we're not doing any of this after all, instead, all DAG generators will track debug names and copy them into every record. One of the future IDAG PRs will address this.
This PR is a clean-up refactoring for parts of #197, and A back-port from the instruction graph branch which will remove
buffer_manager
and extend the recording / printing infrastructure further.buffer_recorder
for buffer-debug-name tracking (so we don't have to askbuffer_manager
)buffer_recorder
intoprint_task_graph
/print_command_graph
, andtask_recorder
intoprint_command_graph
, thereby removing duplicate information from records