Yes, I'm up-to-date. On 09/17/2015 01:07 PM, Fabio Luporini wrote:
Mmm, that is weird. Something odd is going on, I'm curious. At least O1 should be way faster than O2. Are you up-to-date? Will look into it, and will get back to you soon.
Thanks!
-- Fabio
2015-09-17 18:05 GMT+01:00 Tuomas Karna <tuomas.karna@gmail.com <mailto:tuomas.karna@gmail.com>>:
Thanks Fabio,
"parameters["coffee"] = {}" makes a huge difference, compiles in 1 min. Using O1 is almost as slow as O2 (~13 min).
- Tuomas
On 09/17/2015 04:09 AM, Fabio Luporini wrote:
Hello Tuomas!
For disabling coffee adding "parameters["coffee"] = {}" at the top of your file should do the trick.
You could retain some caffeine while decreasing code generation time by instead setting "parameters["coffee"] = {'O1'=True}" (note: default is O2=True, which is more aggressive - as expected - than O1)
Cheers
-- Fabio
2015-09-16 22:42 GMT+01:00 Tuomas Karna <tuomas.karna@gmail.com <mailto:tuomas.karna@gmail.com>>:
Hello Firedrakers,
I'm using "fd_bendy" branches of FFC and UFL, and it takes ~15 min to compile the attached example (with master branches it takes only ~1 min). Is there any way to speed it up? Are there some optimization flags I can turn off? I'm running small test cases so runtime efficiency is not a concern.
Cheers,
Tuomas
_______________________________________________ firedrake mailing list firedrake@imperial.ac.uk <mailto:firedrake@imperial.ac.uk> https://mailman.ic.ac.uk/mailman/listinfo/firedrake
_______________________________________________ firedrake mailing list firedrake@imperial.ac.uk https://mailman.ic.ac.uk/mailman/listinfo/firedrake