David Sanchez
David Sanchez
WIP ## Checklist - [ ] Change(s) are motivated and described in the PR description - [ ] Testing strategy is described if automated tests are not included in the...
This is an interface purely for diagnostics and testing. ## Checklist - [ ] Change(s) are motivated and described in the PR description - [ ] Testing strategy is described...
Profiling-by-endpoint doesn't work when using the libdatadog uploader because it uses a completely wrong mechanism for traversing the span object. ## Checklist - [x] Change(s) are motivated and described in...
PR does - Patches `torch.profiler.profile` class by adding our own `on_trace_ready` handler - Adds GPU time/flops/memory samples via libdatadog interface in `on_trace_ready` event handler - Ensures that libdd exporter is...
It appears that using stack v2 may extend the lifetime of gunicorn workers in such a way that the gunicorn arbiter has to SIGKILL them in some configurations. We should...
Whenever a contextvar is reassociated, it causes the underlying HAMT data structure to clone a node. This clone operation requires de-referencing stored Python objects, which can cause segmentation faults if...
## Checklist - [x] PR author has checked that all the criteria below are met - The PR description includes an overview of the change - The PR description articulates...
Since profiling contains quite a bit of native code, it has a number of native checks (running with sanitizers, static analysis, etc) that's done outside of our normal build operation....
Customers report that some ways of using a non-default tracer configuration can result in the profiler failing to acquire the correct configuration. This patch makes the profiler fully dynamic, like...