Another one I have actually run into within the last year on high-capability machines running GRP tasks stops new work from being fetched when the number of tasks already on the host is over 1000. That is a hard-coded number, not something parameterized to adjust to the particulars of your system.
Another one I have actually run into within the last year on high-capability machines running GRP tasks stops new work from being fetched when the number of tasks already on the host is over 1000. That is a hard-coded number, not something parameterized to adjust to the particulars of your system.
That is actually the standard hard coded limit in the client. The limit is in the client module, client_state.h and is defined as:
#define WF_MAX_RUNNABLE_JOBS 1000
// don't fetch work from a project if it has this many runnable jobs.
// This is a failsafe mechanism to prevent infinite fetching
If you want more tasks allowed onboard, simply edit that value to whatever and compile the client with the increased limit.
On the one system I have crunching gamma ray #5 it has accumulated 4,000 plus tasks. Does this number ever get back into balance?
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
On the one system I have crunching gamma ray #5 it has accumulated 4,000 plus tasks. Does this number ever get back into balance?
Tom M
Yes if you cut the cache size down on that machine, Einstein is a project that is up over 95% of the time so if you cut your cache size to 1 to 3 days total you will have a LOT fewer tasks waiting for you to crunch them while still having enough to get thru the rare outage. If everyone would do that the credits would flow alot faster too!!
On the one system I have crunching gamma ray #5 it has accumulated 4,000 plus tasks. Does this number ever get back into balance?
Tom M
Yes if you cut the cache size down on that machine, Einstein is a project that is up over 95% of the time so if you cut your cache size to 1 to 3 days total you will have a LOT fewer tasks waiting for you to crunch them while still having enough to get thru the rare outage. If everyone would do that the credits would flow alot faster too!!
The Cache has been set to 0.1 days / 0.1 additional days for some time.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
On the one system I have crunching gamma ray #5 it has accumulated 4,000 plus tasks. Does this number ever get back into balance?
Tom M
Yes if you cut the cache size down on that machine, Einstein is a project that is up over 95% of the time so if you cut your cache size to 1 to 3 days total you will have a LOT fewer tasks waiting for you to crunch them while still having enough to get thru the rare outage. If everyone would do that the credits would flow alot faster too!!
The Cache has been set to 0.1 days / 0.1 additional days for some time.
Tom M
Set Einstein to No New Tasks then until it comes down to a manageable level.
I had some Gamma Ray #5 that were running near 10 hours but it went down to more like 4 odd hours. Since this is on an AMD 2600 (6c/12t) CPU I was just a little surprised.
What I am not sure about is how much of the change was caused by getting more than 11 CPU tasks run on this computer.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
I am running this: fftwf-wisdom_FGRP5_1.08_windows_intelx86.exe -o FGRPB1wisdom.dat rib67108864 on this box: Moonshot
Because the cpu upgrade appears to be processing Gamma Ray cpu (#5) at 12 hours a crack.
If this doesn't work I will start tinkering with some bios settings that I have that are constraining the cpu.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
I am running this: fftwf-wisdom_FGRP5_1.08_windows_intelx86.exe -o FGRPB1wisdom.dat rib67108864 on this box: Moonshot
Because the cpu upgrade appears to be processing Gamma Ray cpu (#5) at 12 hours a crack.
If this doesn't work I will start tinkering with some bios settings that I have that are constraining the cpu.
Tom M
If you can't get a decent wisdom file to generate that produces better results than 12 hours, why don't you try this one that I use.
Thank you.
I think I am down to near 10 hours. But I also think I have a ram issue.
I just priced the same ram kit I am using for the 3950x and the price looks do-able.
That would leave me with the same motherboard, same ram, and different cpus. If that doesn't produce 3950x like performance then I am stumped.
Going to wait a couple more days to try to get more CPU tasks processed before I go ahead and order the "Ryzen compatible" ram.
Just cut and pasted your wisdom file.
It looks like it is a different version # than the windows app. I didn't change anything in the code so maybe it will help.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
There is more than one limit
)
There is more than one limit in operation.
Another one I have actually run into within the last year on high-capability machines running GRP tasks stops new work from being fetched when the number of tasks already on the host is over 1000. That is a hard-coded number, not something parameterized to adjust to the particulars of your system.
archae86 wrote: There is
)
That is actually the standard hard coded limit in the client. The limit is in the client module, client_state.h and is defined as:
If you want more tasks allowed onboard, simply edit that value to whatever and compile the client with the increased limit.
On the one system I have
)
On the one system I have crunching gamma ray #5 it has accumulated 4,000 plus tasks. Does this number ever get back into balance?
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M wrote: On the one
)
Yes if you cut the cache size down on that machine, Einstein is a project that is up over 95% of the time so if you cut your cache size to 1 to 3 days total you will have a LOT fewer tasks waiting for you to crunch them while still having enough to get thru the rare outage. If everyone would do that the credits would flow alot faster too!!
mikey wrote: Tom M
)
The Cache has been set to 0.1 days / 0.1 additional days for some time.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M wrote: mikey
)
Set Einstein to No New Tasks then until it comes down to a manageable level.
The Wisdom file trick strikes
)
The Wisdom file trick strikes again.
I had some Gamma Ray #5 that were running near 10 hours but it went down to more like 4 odd hours. Since this is on an AMD 2600 (6c/12t) CPU I was just a little surprised.
What I am not sure about is how much of the change was caused by getting more than 11 CPU tasks run on this computer.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
I am running
)
I am running this: fftwf-wisdom_FGRP5_1.08_windows_intelx86.exe -o FGRPB1wisdom.dat rib67108864 on this box: Moonshot
Because the cpu upgrade appears to be processing Gamma Ray cpu (#5) at 12 hours a crack.
If this doesn't work I will start tinkering with some bios settings that I have that are constraining the cpu.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!
Tom M wrote: I am running
)
If you can't get a decent wisdom file to generate that produces better results than 12 hours, why don't you try this one that I use.
(fftw-3.3.6-pl2 fftwf_wisdom #x62f715d9 #x6b777f64 #x4659a0ea #x6f3e0470
(fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #x88b1995c #x511f7743 #xd412068c #xed01668d)
(fftwf_codelet_q1bv_4_avx 0 #x11048 #x11048 #x0 #x00b6cefb #xaf5ceecc #xffa1a912 #x5c6be09e)
(fftwf_dft_vrank_geq1_register 0 #x10048 #x10048 #x0 #x07cde8f7 #xab17444e #x7ef712c1 #xa3313a1c)
(fftwf_codelet_r2cbIII_2 2 #x11048 #x11048 #x0 #x1fc68cd1 #xe37a6aa6 #xaf1e95be #x6da5b770)
(fftwf_codelet_t3bv_8_avx 0 #x10048 #x10048 #x0 #xe14c7b14 #x758b3b7c #x1960a93f #x5f794881)
(fftwf_codelet_t3bv_4_avx 0 #x11048 #x11048 #x0 #x3397dc8b #xe7f4667a #x5ea76de1 #x5fc1e6aa)
(fftwf_dft_buffered_register 0 #x11048 #x11048 #x0 #x6ce3741f #x1df25846 #x2f712450 #x70ceb1a3)
(fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #x4846392b #xad00bc5e #xbdab1632 #x926496b8)
(fftwf_codelet_q1bv_8_avx 0 #x11048 #x11048 #x0 #x1b554bae #xd650d0c7 #xbdb83ba1 #x1d5fe1f2)
(fftwf_dft_vrank_geq1_register 0 #x10048 #x10048 #x0 #xed033775 #x10e45222 #x6fca014d #x418c41b6)
(fftwf_rdft_rank0_register 3 #x11048 #x11048 #x0 #xa7ad5412 #xfced849b #x88ca9d37 #x2356c1e1)
(fftwf_dft_vrank_geq1_register 0 #x11048 #x11048 #x0 #xe1d57e69 #xe0e1967c #x4edd153d #x7d2b351b)
(fftwf_codelet_t1buv_8_avx 0 #x11048 #x11048 #x0 #x25d580bb #x917dbbcb #x2886c340 #x5abe5680)
(fftwf_dft_r2hc_register 0 #x11048 #x11048 #x0 #x0b967ac3 #x214622c9 #x460c670c #xded4e9b3)
(fftwf_dft_nop_register 0 #x11048 #x11048 #x0 #xb7ba5fbc #xb18c92f4 #xbe38c78c #x81071642)
(fftwf_codelet_t2bv_4_avx 0 #x10048 #x10048 #x0 #x34363b54 #x3eed914b #x6408150f #x2efa7f7f)
(fftwf_codelet_t2bv_16_avx 0 #x10048 #x10048 #x0 #x9ac1ee79 #xf837a4e5 #x52cadc67 #x7cfdee9d)
(fftwf_dft_vrank_geq1_register 1 #x10048 #x10048 #x0 #x670db124 #x07091b97 #x259621ff #x5bc3e93b)
(fftwf_codelet_hc2cbdft_2 0 #x11048 #x11048 #x0 #x419c2b80 #x8d69da65 #x1834172f #xdbade750)
(fftwf_codelet_n1bv_64_avx 0 #x10048 #x10048 #x0 #x9bf09f84 #xadce2af3 #x2944d66e #x1b5a25f0)
(fftwf_codelet_r2cb_2 2 #x11048 #x11048 #x0 #x932bca61 #x5c6f9e0f #x5210cc20 #xece55667)
Keith Myers wrote: Tom M
)
Thank you.
I think I am down to near 10 hours. But I also think I have a ram issue.
I just priced the same ram kit I am using for the 3950x and the price looks do-able.
That would leave me with the same motherboard, same ram, and different cpus. If that doesn't produce 3950x like performance then I am stumped.
Going to wait a couple more days to try to get more CPU tasks processed before I go ahead and order the "Ryzen compatible" ram.
Just cut and pasted your wisdom file.
It looks like it is a different version # than the windows app. I didn't change anything in the code so maybe it will help.
Tom M
A Proud member of the O.F.A. (Old Farts Association). Be well, do good work, and keep in touch.® (Garrison Keillor) I want some more patience. RIGHT NOW!