Author: Tom Likens
Date: 12:33:35 07/24/03
Go up one level in this thread
On July 24, 2003 at 14:49:05, Uri Blass wrote: >On July 24, 2003 at 14:11:07, Tom Likens wrote: > >>On July 24, 2003 at 03:09:52, Albert Bertilsson wrote: >> >>>After tracing a bug in Sharper for over a week I was starting to loose hope... >>>A ran 600 games with a debug version, no problem. >>>Switched to release version, after 579 games the bastard hanged =(. >>> >>>Would you consider releasing this engine version bad policy? >>> >>>Have you any known bug in your engine that you simply do not care about because >>>it occurs very rarely? >>> >>>I realize that it will not be easier to find the bug later, it's just that I >>>don't have the stamina to try and reproduce it again with the debug version, it >>>can take thousands of games =( >>> >>>As I write Sharper mainly to let it compete against other engines I have no >>>problem with giving away that one point in a thousand games, but I wouldn't like >>>to see Sharper excluded from tournaments because it's buggy. >>> >>>/Regards Albert >> >>If you're running under Windows you might try BoundsChecker from CompuWare >>or PC-Lint from Gimpel. If you do get PC-Lint and you are linting C++ code >>I would recommend trying to get the older 7.5 version, since 8.0 was a >>complete rewrite and from what I understand is still a bit buggy. CompuWare >>will give you a 14-day trial of BoundsChecker, if you ask nicely ;) >> >>A free tool worth trying is Valgrind (not sure if it runs under Windows, >>but I know it definitely runs under Linux and works just fine on C++). >> >>good luck, >>--tom > >I have BoundsChecker from CompuWare but I use it only as another tool to help me >to find more bugs because there are cases when everything seems to be ok but >there is still a bug. > >The problem is that with checkingbounds from compuware movei runs more than >10,000 times slower so most of my tests are without compuware. > >A lot of errors are also not crashes because before I use a new array that I >added I check some conditions about the array and I simply ask the program to >print some information in the screen if the condition does not happen because I >know that there is a good probablility that the array does not have the right >information because of a bug. > >Often the first error happens only after a second of search of the release >version and using compuware means that I need to wait many hours for the same >purpose. > >Uri The new version (7.0.2) seems faster, but you're right, if you enable the Final Check option, then it is *very* slow. A good example is that it takes my program about 15-hrs to finish WAC under Final Check but it finishes it under 5 min. on the same machine when not run under BoundsChecker, for a difference of about 540x (not quite as bad as your 10,000x slowdown, but still painful). There are also levels of checking that can be enabled. If you don't turn on the full debug checking then it will run much faster (of course it may not catch as many bugs, but nothing's perfect). regards, --tom
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.