During computation we must be carful to keep kernels unique: when
we create a new kernel we must check if such kernel was created
earlier and if needed identify it. For this purpose we use
SortedCache. Original version of SortedCache used linear
search. I have now commited a patch which changes it to
use binary search. This gives nice speedup on mapleok:
before change to operator properties on 2.4 GHz Core 2
using sbcl 1.18 mapleok took 89.31s, after change to operator
properties 72.99s, now 28.82s. So, together both changes
make FriCAS 3 times faster on mapleok than it used to be.
I am a bit surprised that we have such a large speedup: I
profiled the same routines more than year ago and it looked
that there is potential for 30-40% speedup. Since I am
now using different machine and different sbcl version
this may affect relative timings. But I think that we
also see cumulative effect of improvements: other parts
got faster (especially due to call speed optimization),
so linear search took larger percentage of total time.
BTW: mapleok stresses "symbolic" manipulations. Effect on
other tests is much smaller.
--
Waldek Hebisch
[EMAIL PROTECTED]
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups
"FriCAS - computer algebra system" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to [EMAIL PROTECTED]
For more options, visit this group at
http://groups.google.com/group/fricas-devel?hl=en
-~----------~----~----~----~------~----~------~--~---
