Hidden affects of setting $MinPrecision/SetPrecision?
- To: mathgroup at smc.vnet.net
- Subject: [mg56428] Hidden affects of setting $MinPrecision/SetPrecision?
- From: terryisnow at yahoo.com
- Date: Mon, 25 Apr 2005 01:30:42 -0400 (EDT)
- Sender: owner-wri-mathgroup at wolfram.com
Hi, I have a Mathematica 3.0 simulation that consists of two parts. The first part loads an 'environment' file (basically a Mathematica DumpSave file with all the relevant variables), runs through 35000 iterations, then makes a determination about whether there were interesting results. If so it signals the second part of the simulation to reload the same environment file and re-run the 35000 iterations but this time saving more details. At the end of this cycle, regardless of whether the second part runs, a new environment file is saved and everything is repeated for the next 35000 iterations. Lately however I've been finding problems with the results of part one of my simulation not matching the results of part two. This coincided with the introduction of $MinPrecision and SetPrecision[] into my calculations (in fact $MinPrecison is now also saved in the environment file, because obviously you need that to recreate a given environment). Basically both parts of my simulation look for special 'events' that indicate that an increased precision is necessary. When such events are found, SetPrecision[] and $MinPrecision are employed to increase and maintain the precision. When the two parts match (as they should), the iterations at which precision is increased are identical. So part one might run and tell me that precision was increased at 35103 and 35355 iterations. Then part two runs after reloading the environment file and tells me the same thing. This is however currently not the case. I've been running these simulations for a long time and the code doesn't change that often, but just to be sure that the two parts are semantically equivalent, I ran each part separately, rebooting my system and reloading the same environment file each time. When I do this the results match. So this isn't the problem. Somehow the first part of my simulation is changing the environment in some way that is not being reset by the reloading of the environment file. As a last resort I carefully took every variable in my program and Clear[]'d and ClearAll[]'d them right before reloading the environment file for part two (including the resetting of $MinPrecision). That too did not help. So I'm at my wits end. If anyone knows why this is happening, I'd be very appreciative to hear from you. Terry