#lisp - Mon 30 Apr 2007 between 18:16 and 18:36

NY Lost Funds



jsnellassuming your platform is using the brk for malloc, it's possible that the sbcl memory spaces are preventing the brk heap from expanding
alastair37the call (foreign-alloc 'contact-geom :count 1) will fail very infrequently, but at the most inopportune times
so what do i do about that?
jsnellif that were the problem, moving the dynamic space higher into memory would help
alastair37meaning, give --dynamic-space-size <bigger-number> would help?
jsnellno, that would just move the end of the dynamic space. you'd need to move the start
alastair37how do i move the start?
jsnellmodify dynamic-space-start in src/compiler/foo/parms.lisp
alastair37i see. and then rebuild sbcl, i take it?
jsnellright
alastair37all right
jsnellit should be easy to test whether that helps, just try allocating memory in the loop before the change, and see how much you get before running out
then see whether moving the dynamic space higher allows you to allocate more
alastair37so, it doesn't seem like it's running out, because it's not like i do foreign-alloc n times and then afterwards i can't do it anymore-- it's like, i do foreign-alloc n times, the n+1th fails, and then i try to foreign-alloc again from the repl and it works.
would that be inconsistent with the dynamic space start problem?
jsnellthat was a hypothesis based on one way that malloc could be implemented. I don't know whether it actually is implemented that way in glibc
alastair37i'm also allocating tiny amounts of space, here-- 38 /bytes/ at a time
foomif you're using linux, the location of the dynamic heap shouldn't have any effect on how much memory you can malloc.
glibc uses mmap if it can't brk any more memory
unless maybe you're using some incredibly ancient version
alastair37no, i'm on an up-to-date ubuntu box
rahulI doubt sbcl would work on that old of a libc
jsnellwhere's the memory fault happening?
alastair37what do you mean, where
jsnellwhat's the backtrace like?
alastair37i lost the backtrace. is there something in particular you wanted to know about it, or should i try to keep poking it until it happens again?
jsnellnothing in particular, just thought that it might contain some hint
alastair37i'll poke it, then, while you think about it
Krystofwhat if "storing the wrapper in the same place" in handle-single-step-around-trap is doing the wrong thing on sparc?
then something bad might happen the /next/ time we went in
jsnellit must be working correctly, since the wrapper did get called
that's what's printing the "Evaluating call: (PRINT 1) ..." and "(PRINT 1) => 1" messages
Krystofdarn
alastair37in what kinds of situations is it possible to get sbcl not to return anything, to act as if it's still evaluating an expression, but also not use any CPU or ram or disk
(and to be recoverable, usually, but not always, via C-c C-c interrupt)
froydnjconnected to slime?
alastair37yes, connected to slme.
HoA46complete noob here, but in lisp, what would the following command evaluate to in common lisp?
alastair37*clime
*slime

Page: 2 9 16 23 

IrcArchive

NY Lost Funds