Standard Loops ???

[*hondo*]

Although I've been F@H for many years I can't say that I've noticed this previously altho TBH I don't really look at the log file all that often, having looked over the last few days I've noticed this rather too often it always happens if I've rebooted the PC.
The file reads as follows
[18:14:15] - Looking at optimizations...
[18:14:15] - Working with standard loops on this execution.
[18:14:15] - Previous termination of core was improper.
[18:14:15] - Going to use standard loops.

The PC is stable so I guess its me but I don't know why.

Can someone tell me please how I'm terminating the core improperly?
What are standard loops?
Does this matter?
Finally how can I stop it doing this, if it is a of concern?

Thanks *hondo*

Mod Edit: Fixed Underline Tag - PantherX

[JimboPalmer]

Normally, F@H uses SSE, to speed up CPU calculations. If I am doing repeated upgrades to my computer and kill power badly more than once, it may do 'standard loops' which is much slower. (8 times slower?)

The next WU goes back to SSE

http://en.wikipedia.org/wiki/SIMD

[*hondo*]

Normally, F@H uses SSE, to speed up CPU calculations. If I am doing repeated upgrades to my computer and kill power badly more than once, it may do 'standard loops' which is much slower. (8 times slower?)

The next WU goes back to SSE

http://en.wikipedia.org/wiki/SIMD

STRUTH!! 8 times slower Thanks for the info JimboPalmer, is there any way other than waiting for the next WU to make it go back to working without these standard loops? I've still got 10.75 Hrs to go on the current SMP WU. Whats even worse this is only the 4th or 5th SMP therefore no extra credit for finishing the WU on time.

[ChasR]

THe standard loops message is a cosmetic error in the v6 smp and gpu client. Neither will run at all without optimizations. If it bothers you a lot seeing the message in the log, the addition of the -forceasm flag will keep the "standard loops" message from being displayed, but that is all it does for SMP and GPU clients. It will force optimizations on uniprocessor work (I'm not sure about the hybrid a4 core)

[codysluder]

Normally, F@H uses SSE, to speed up CPU calculations. If I am doing repeated upgrades to my computer and kill power badly more than once, it may do 'standard loops' which is much slower. (8 times slower?)

The next WU goes back to SSE

http://en.wikipedia.org/wiki/SIMD

8x doesn't apply here. Gromacs for the x86 hardware uses single precision SSE which means four FP calculations can be completed in a single instruction instead of four separate operations. Only perfectly optimized code can achieve a 4x speedup. Compiler optimizations often achieve a speedup of as much as 2x or if they're lucky, 3x. Fah's inner loops are hand optimized assembly code and achieve a speedup of 3.8x.

If you run the uniprocessor client (fahcore_78) which has both optimized an unoptimized code, the message is important by the factor of 3.8. You can bypass this problem by adding the -forceasm flag. For any other client, the message should be ignored. It's meaningless for either the SMP or GPU cores since neither one actually contains unoptimized code. The gpu core does the work in the gpu, itself, not the cpu, so it doesn't use SSE. Any x86 cpu which has more than one core will always have SSE so the code has been configured to always use SSE, no matter what the messaged suggests.

[*hondo*]

Thanks guys for the replies TBH I'm a very much Older Client & don't understand much of the above, I'm running both an SMP on the CPU & a GPU I'm now getting the vibes that this Standard loops stuff in my case dosent matter. Right?

Cheers *hondo*

[bruce]

The message about standard loops only matters if you run the uniprocessor client and you're getting assigned projects that use FahCore_78.

[*hondo*]

Great, I can see an A3 & an A4 no signs of FahCore_78.
Thanks Guys

Bruce can you tell Standford to send me smaller WUs until I get more than 10 of these SMPS done please, these
P10133 are playing lights out with me.

[bruce]

Sorry. but Stanford applies the same assignment logic to everybody; no exceptions.

[*hondo*]

Sorry. but Stanford applies the same assignment logic to everybody; no exceptions.

Thanks but no problems now Bruce 10 are completed on SMP2 Plus almost 5000 on GPU & many others dating back to Tinker core days, albeit under different teams & user names.


[13:24:14] Completed 750000 out of 750000 steps (100%)
[13:24:14] DynamicWrapper: Finished Work Unit: sleep=10000
[13:24:24]
[13:24:24] Finished Work Unit:
[13:24:24] - Reading up to 1152288 from "work/wudata_00.trr": Read 1152288
[13:24:24] trr file hash check passed.
[13:24:24] - Reading up to 1108788 from "work/wudata_00.xtc": Read 1108788
[13:24:24] xtc file hash check passed.
[13:24:24] edr file hash check passed.
[13:24:24] logfile size: 131365
[13:24:24] Leaving Run
[13:24:28] - Writing 2500945 bytes of core data to disk...
[13:24:28] Done: 2500433 -> 2311474 (compressed to 92.4 percent)
[13:24:28] ... Done.
[13:24:28] - Shutting down core
[13:24:28]
[13:24:28] Folding@home Core Shutdown: FINISHED_UNIT
[13:24:31] CoreStatus = 64 (100)
[13:24:31] Sending work to server
[13:24:31] Project: 7200 (Run 23, Clone 9, Gen 80)


[13:24:31] + Attempting to send results [January 30 13:24:31 UTC]
[13:24:49] + Results successfully sent
[13:24:49] Thank you for your contribution to Folding@Home.
[13:24:49] + Number of Units Completed: 10