https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98426
--- Comment #11 from Jerry DeLisle <jvdelisle at gcc dot gnu.org> --- I am able to run your reproducer and I can see the increasing times as the number of modules goes up. I am curious if you could randomize the subroutine names? These appear fairly repetitive and I wonder if this biases the test. My results: Build time for base.F90 in Modules_100: 0.446443 seconds Number of Modules | Build Time ----------------- | ---------- 10 | 0.0155371 20 | 0.024554 30 | 0.041164 40 | 0.0620602 50 | 0.092014 60 | 0.135193 70 | 0.184979 80 | 0.255272 90 | 0.335244 100 | 0.446443 real 0m27.194s user 2m27.698s sys 0m21.231s The build time presented appears to not be wall time so I wonder what this is.