When I optimize my code using DropEquivalentBufferResultsPass, I found that it took a long time, almost an hour and a half. To be honest, this module is really big. The main function forward has 4400 call sites, and there are 2500 sub-functions. I wonder if there’s any way to optimize it.
I can’t spot any obvious implementation issues with this pass. I’d try profiling it to see where we are spending time. 1.5 hours is way too slow.
Thank you so much for your help.