|
From: Dirk M. <dm...@me...> - 2004-01-14 19:59:59
|
Is there a reason VG_N_SEMAPHORE is set to 50? I end up recompiling on all machines because our software requires orders of magnitude more. Are there some adverse affects that I'm not seeing? -Dirk |
|
From: Nicholas N. <nj...@ca...> - 2004-01-18 18:04:42
|
On Wed, 14 Jan 2004, Dirk Morris wrote: > Is there a reason VG_N_SEMAPHORE is set to 50? > I end up recompiling on all machines because our software requires > orders of magnitude more. > Are there some adverse affects that I'm not seeing? AIUI, it's set to 50 because that's usually enough; the only time someone else complained (that I'm aware of) they had a semaphore leak in their program. Increasing the compile-time constant shouldn't have adverse effects. The only problem might be that Valgrind's libpthread implementation does some things in pretty basic ways, so some semaphore operations might be linear in the number of semaphores, or similar. Is your software doing anything unusual? Needing 10x as many semaphores as everyone else seems extreme... could you have semaphore leaks? N |