Dear Florian, I am profoundly grateful for your precise answer, as I can now focus on my thesis peacefully. For my new 7.2 MB 328963-lines-long C file: clang -c -O0: 23s, 34 MB gcc -c -O0: 1m8s, 31 MB clang -c -Oz: 1m59s, 23 MB gcc -c -Oz: 4m23s, 23 MB https://github.com/bottle2/swf2c/blob/085de13eb535a90b3c2a8b2f200bda00d8671102/there_she_is.c Best regards, Bento Em qua., 12 de fev. de 2025 às 05:25, Florian Weimer <fweimer@xxxxxxxxxx> escreveu: > > * Bento Borges Schirmer: > > > [3] https://github.com/bottle2/swf2c/blob/88f9ccb7912d55002e87f1efb11f21720d97e4ec/tests/thousands-of-functions.c > > You should turn L and B into proper functions instead of macros, then > compilation time will decrease significantly. If compilation time is > still too high, consider adopting a table-based approach. > > > [4] https://github.com/bottle2/swf2c/blob/c4e4faaa50fda0876fa877e68f50de2c6291a7c2/tests/gigaenormous-function.c > > Similar changes as above. The OpenMP sections should probably be > separate functions. Maybe store their addresses in an array and invoke > them from an OpenMP parallel loop. > > Thanks, > Florian >
