You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For some reason the create-ethernet-map call hangs, and I can't figure out why.
This is called from Hal::Hal() > tt_metal/llrt/get_platform_architecture.hpp, and this seems like a test intialization. What's more interesting is that this works at other places. For instance, in the referenced PR I also changed detect_arch to tt_ClusterDescriptor::create() in tests/tt_metal/test_utils/env_vars.hpp and that call passes when later it is called from the tests itself.
I compared ENV variables and ls /dev/tenstorrent and they're exactly the same for the two calls.
Here's the stack that I managed to extract with GDB
This blocks removing detect_arch() from our code
The text was updated successfully, but these errors were encountered:
I saw this while working on #345 and tenstorrent/tt-metal#15564
To repro, add tt_ClusterDescriptor::create() to tt_metal/llrt/get_platform_architecture.hpp, then do this:
For some reason the create-ethernet-map call hangs, and I can't figure out why.
This is called from Hal::Hal() > tt_metal/llrt/get_platform_architecture.hpp, and this seems like a test intialization. What's more interesting is that this works at other places. For instance, in the referenced PR I also changed detect_arch to tt_ClusterDescriptor::create() in tests/tt_metal/test_utils/env_vars.hpp and that call passes when later it is called from the tests itself.
I compared ENV variables and ls /dev/tenstorrent and they're exactly the same for the two calls.

Here's the stack that I managed to extract with GDB

This blocks removing detect_arch() from our code
The text was updated successfully, but these errors were encountered: