These are chat archives for symengine/symengine

26th
Dec 2015
Francesco Biscani
@bluescarni
Dec 26 2015 00:00
pi@raspberrypi:~/repos/piranha/build_perf $ sudo nice -n -19 ./tests/fateman1_perf 4
Running 1 test case...
 14.475813s wall, 56.950000s user + 0.110000s system = 57.060000s CPU (394.2%)

*** No errors detected
Freeing MPFR caches.
Setting shutdown flag.
pi@raspberrypi:~/repos/piranha/build_perf $ sudo nice -n -19 ./tests/fateman1_perf 1
Running 1 test case...
 48.945153s wall, 48.860000s user + 0.080000s system = 48.940000s CPU (100.0%)

*** No errors detected
Freeing MPFR caches.
Setting shutdown flag.
pi@raspberrypi:~/repos/piranha/build_perf $ sudo nice -n -19 ./tests/pearce1_perf 4
Running 1 test case...
 6.993902s wall, 25.040000s user + 1.940000s system = 26.980000s CPU (385.8%)

*** No errors detected
Freeing MPFR caches.
Setting shutdown flag.
pi@raspberrypi:~/repos/piranha/build_perf $ sudo nice -n -19 ./tests/pearce1_perf 1
Running 1 test case...
 25.413737s wall, 24.210000s user + 1.130000s system = 25.340000s CPU (99.7%)

*** No errors detected
Freeing MPFR caches.
Setting shutdown flag.
Francesco Biscani
@bluescarni
Dec 26 2015 00:06
The unit tests and the Python bindings are too voluminous to be compiled directly on the device, but for the perf tests 1GB of memory is enough. I think it uses up to 60% of that at peak during compilation.
Ralf Stephan
@rwst
Dec 26 2015 16:32
Heh I found the code in Pynac that caused super-O(n^2) time behaviour for non-Puiseux series multiplication and fixed it in pynac/pynac@c12fbea. The competition doesn't sleep...