[Libre-soc-dev] daily kan-ban update 28jun2021

Luke Kenneth Casson Leighton lkcl at lkcl.net
Tue Jun 29 19:16:17 BST 2021


just saw this, tobias: i don't know if you tried it already, i found
that you can cut down on time and also often find out patterns by
cutting out some of the test data (one at a time is the slow way) and
do a binary search on what actually causes the failure.

then when *only* the data left causes a failure you can sometimes see
that the addresses are inter-related, or some timing etc.

but with the extra non-related data you can't see that.

also, by identifying which piece of data had to be removed to get the
test to *not* fail, this is also valuable info that is otherwise
missing from either staring at code or VCD traces.


More information about the Libre-soc-dev mailing list