View Single Post
  #13 (permalink)  
Old 24th July, 2004, 02:33 AM
BuiesCreek847's Avatar
BuiesCreek847 BuiesCreek847 is offline
Join Date: July 2004
Posts: 32

1. -forceSSE as I have an AMD processor, and this will use the optimized code for AMD's. What I think it does is download the FAH_78 core if you don't use it and the FAH_65 core if you do. Anyone can correct me if this is wrong.

2. -forceasm "Instruct the core to use SSE/3DNow assembly instructions if possible", this is from the users guide. I just know MY sysetems seems to process WU's faster with this option.
3. -advmethods This allows you to be assigned new cores &/or WUs, look in the forums for discussion on this feature.
As found in the F@H FAQ about switches:
Instruct the core to use SSE/3DNow assembly instructions if possible, even if it has previously made the determination that the machine may not be handling this well.
On machines with an AMD processor, Core_78 gives priority to 3DNow over SSE -- this overrides that.
Request to be assigned work units from new projects, before they are broadly released to the rest of Folding@home. As among the first to try out new advanced simulations, you may wish to post any observations (positive or negative), questions, or issues at the user Forum.

-forceasm: This forces your rig to run tha advanced assembly methods that are available with the newer CPUs. The F@H client, by default will turn those off should it see any problems during crunching. I'll stay off till you restart the client after making adjustments. This switch simply forces them to stay on. If there are continued problems, the client will most likely do the early unit end thing.

-forceSSE: When it comes to crunching Gromacs WUs, SSE instructions are way faster than 3DNow instructions. This switch forces the SSE over the 3DNow. The client, by default starts with 3DNow. This switch will be phased out for -forceasm with the next client release.

-Advmethods: As you may know, most all of us do the regular WUs. Well, to get to us, there are some folks who do initial testing of Clients, cores, and WUs. Those folks are in the BETA stages of testing. When most of the kinks are worked out, they take the testing one step further, and open them to those who run the advmethods switch. With that, those still experimental clients, cores, and WUs have the chance to work with some of the wild equipment that you and I run, and see if things still hold up. If more bugs are found, they are corrected and then finally released to the general public. If one is to run the -advmethods switch, they should understand that they are working with clients, cores, and/or WUs who may crash, or act stupid otherwise. If there is anything weird about them, it should be reported to so that said info can be forwarded to Stanford.

There was a time that, one could get a lot more points, very fast by running with the -advmethods switch. That was once true in that Gromacs were just coming out. They were still mostly experimental, still though, many were jumping on that switch just for the points. The problem was, and still is, that with running the -advmethods switch, you much contact the scientists about any problem found. You must watch things much closer. If you ain't into all that, don't bother running the switch.

Further, the _65core is exclusive to Tinker WUs. The _78core is exclusive to Gromacs. I can't remember what the Genome core was.


Last edited by BuiesCreek847; 24th July, 2004 at 10:20 PM.
Reply With Quote