CoCoA » CoCoA-5

Here is a concrete example... a bit faster, but still sufficient to exhibit the problem.

H := 5000;
foreach C in (0..H)><(-H..H) do
  // something
endforeach;
println "Starting loop 2";
for i := 1 to 30000 do
  if factorial(i) = 120 then println i; endif;
endfor;

Note that, according to top, during 1st loop memory increases a bit beyond 4Gbytes, but during the 2nd loop it drops to about 3.7Gbytes (rather than 0 Gbytes). Why?

This is what valgrind produced:

With CoCoALib and external libraries GMP, Normaliz, Frobby, Gfan, CDD, MathSAT
indent(VersionInfo(), 2); -- for information about this version
# H := 5000;
# foreach C in (0..H)><(-H..H) do
[[foreach]] #   // something
[[foreach]] # endforeach;
==14306== Warning: set address range perms: large range [0x1a627f040, 0x1c627f040) (undefined)
==14306== Warning: set address range perms: large range [0xde27e028, 0xee27e058) (noaccess)
==14306== Warning: set address range perms: large range [0x1a627f028, 0x1c627f058) (noaccess)
# println "Starting loop 2";
Starting loop 2
# for i := 1 to 30000 do
[[For]] #   if factorial(i) = 120 then println i; endif;
[[For]] # endfor;
5
# ==14306== 
==14306== HEAP SUMMARY:
==14306==     in use at exit: 134,751 bytes in 217 blocks
==14306==   total heap usage: 150,887,968 allocs, 150,887,751 frees, 8,190,471,597 bytes allocated
==14306== 
==14306== LEAK SUMMARY:
==14306==    definitely lost: 0 bytes in 0 blocks
==14306==    indirectly lost: 0 bytes in 0 blocks
==14306==      possibly lost: 0 bytes in 0 blocks
==14306==    still reachable: 134,751 bytes in 217 blocks
==14306==         suppressed: 0 bytes in 0 blocks
==14306== Rerun with --leak-check=full to see details of leaked memory
==14306== 
==14306== For lists of detected and suppressed errors, rerun with: -s
==14306== ERROR SUMMARY: 0 errors from 0 contexts (suppressed: 0 from 0)

Does not help to isolate the problem. I wonder what the 135k of still reachable memory is?
Maybe I should try with MEMPOOL_DEBUG? But not today.