Yesterday I commited a patch allows using GMP for bignum
computation in 32-bit sbcl on Intel/AMD processors.
To use it configure using '--with-gmp' option -- you must
have gmp installed (including C header files, that is devel
package), otherwise build will fail.
Actually, there a chance that the code will work on other
32-bit processors, but I have no way of testing on such
processors. Consequently, currently configure will
signal error if you try to use '--with-gmp' on something
other than 32-bit or 64-bit Intel/AMD processors.
Also '--with-gmp' will fail if you try to use different Lisp:
ECL and GCL already use GMP and currently using GMP
with Closure CL and Clisp is unsupported (I have code for
64-bit Closure CL on AMD64, but since number of users of
FriCAS on top of Closure CL seem to be low (0???), I did
not integrate that code in FriCAS build process).
BTW, updated timings for 10^6 digits of e:
native sbcl sbcl + GMP 5.0.1
32-bit 69s 1.35s
64-bit 15s 0.6
As you can see going to 64-bit gives substantial speedup
for arithmetic, which in practice may be more significant
than speedup due to use of GMP: GMP gives large speedup
for large numbers, but for smaller ones effect is smaller.
64-bit arithmetic gives speedups at almost all sizes:
huge speedup between 30 and 60 bits (due to use of fixnums
instead of bignums), small speedup for slightly more than
60 bits growing to about 4 when having about few hunded bits.
--
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.
