In limit-context (defined in limit.lisp), comment out the lines
(assume '((mlessp) lim-epsilon 1e-8))
(assume '((mgreaterp) prin-inf 1e+8))
Also, comment out the lines (in the function restore-assumptions)
(forget '((mlessp) lim-epsilon 1.0e-8))
(forget '((mgreaterp) prin-inf 1.0e+8))
And in make-global-assumptions (defined in defint.lisp) comment out
(setq *global-defint-assumptions*
(cons (assume '((mgreaterp) prin-inf 1.0e+8))
*global-defint-assumptions*))
Then:
(1) The testsuite and the share testsuite report no unexpected failures or unexpected fixes.
(2) The number of calls to $sign decreases from 604,896 to 562,547 for the test suite;
for the share test suite, the number of calls to $sign decreases from 1,350,048 to
1,303,314. The time for garbage collection changes from 15.13 seconds to 13.946 seconds.
(3) The running time for the testsuite decreases from 215.749 seconds to 199.228 seconds;
for the share test suite, the running time decreases from 330.767 seconds to
316.649 seconds. The time for garbage collection didn't change that much (20.37
seconds to 20.39 seconds).
My configuration: Intel i3 380M at 2.53 GHz, 64 bit Clozure CL 1.7, .... My to_poly_solver
has some uncommitted changes, so your results will differ. I tried these tests only once, but I did
try to run them under the same conditions (no camera, no wifi, no other applications running).
It would be great if others would try similar experiments. What do we lose by removing some of
the prin-inf > 1e8 (junk) code.
--Barton