🐛 Describe the bug

One of the pattern-matching tests fails on A2/A16 and likely A10 (untested). Not particularly urgent but I would like to learn more about how to expose which part of Inductor is failing to pattern match on certain cases. Since I suspect this is using pre-compiled FA kernels, my guess is there is a hidden constraint that isn't met somewhere (not in the pattern definition itself) or there is a runtime failure that is caught and suppressed.

For working configs it seems to match to pattern 11 (not 12):

Is there a way to get the Inductor logs that would be relevant here?

CC @drisspg

Versions

~week old main branch

cc @ptrblck @msaroufim @chauhang @penguinwu @voznesenskym @EikanWang @jgong5 @Guobing-Chen @XiaobingSuper @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @aakhundov @ezyang @drisspg @mikaylagawarecki