search

tensorflow / tflite-micro

Infrastructure to enable deployment of ML models to low-power resource-constrained embedded targets (including microcontrollers and digital signal processors).
Apache License 2.0
1.85k stars 805 forks source link

Usage of NaN and Infinity not supported by default in armclang #2518

Closed AdrianLundell closed 3 months ago

AdrianLundell commented 6 months ago

Using these is not supported by the std/fast ffp-mode in armclang, forcing useage of the --ffp-mode=full flag. This is fixed for all armclang builds with #2517, but it could be worth the effort to look into supporting other ffp-modes, in particular since they may be faster.

See https://developer.arm.com/documentation/101754/0622/armclang-Reference/armclang-Command-line-Options/-ffp-mode

rascani commented 6 months ago

@AdrianLundell - is this fixed now? If so, please close.

AdrianLundell commented 6 months ago

It is fixed to the extent that we do not build anything with undefined behaviour anymore since #2517, but the issue has a broader scope to add support for all fpp-modes in the future, so it would be nice to keep around for that if that is ok.

github-actions[bot] commented 4 months ago

"This issue is being marked as stale due to inactivity. Remove label or comment to prevent closure in 5 days."

github-actions[bot] commented 3 months ago

"This issue is being marked as stale due to inactivity. Remove label or comment to prevent closure in 5 days."

github-actions[bot] commented 3 months ago

"This issue is being closed because it has been marked as stale for 5 days with no further activity."