mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Marin's Mersenne-aries (https://www.mersenneforum.org/forumdisplay.php?f=30)
-   -   processed dc and tc posts (https://www.mersenneforum.org/showthread.php?t=24152)

ewmayer 2020-03-17 21:05

[QUOTE=kriesel;539975]It's running now, 4% complete, ETA ~9pm US CDT 2020-03-18 which is UTC -0500.[/QUOTE]

Ken, if your GPU client uses the convention that LL initial seed is iteration 0, you can compare interim every-1M-iter Res64s vs mine:
[code][2020-02-11 10:58:54] M50699483 Iter# = 1000000 Res64: FF270405EA0D7239. shift: 18503814
[2020-02-12 01:52:58] M50699483 Iter# = 2000000 Res64: 10BD8630CDC36E73. shift: 28532790
[2020-02-12 18:32:46] M50699483 Iter# = 3000000 Res64: D4F95C0FB91AD1EE. shift: 1090728
[2020-02-13 12:54:54] M50699483 Iter# = 4000000 Res64: 2F3D4D841DD790ED. shift: 32000164
[2020-02-13 18:00:52] M50699483 Iter# = 5000000 Res64: 6762222E6DAC7A3D. shift: 4094697
[2020-02-13 23:10:28] M50699483 Iter# = 6000000 Res64: 0C3E78B1C77FA688. shift: 33080656
[2020-02-14 04:30:42] M50699483 Iter# = 7000000 Res64: 62AFD692A5EC4EB2. shift: 41528603
[2020-02-14 09:48:36] M50699483 Iter# = 8000000 Res64: 9F2AC42F44FA0D49. shift: 20612489
[2020-02-14 16:10:17] M50699483 Iter# = 9000000 Res64: 2572440B76CF7B14. shift: 8838056
[2020-02-15 07:43:44] M50699483 Iter# = 10000000 Res64: 5D942313DA74C513. shift: 18311478
[2020-02-15 23:11:13] M50699483 Iter# = 11000000 Res64: AF441CB0956493CF. shift: 16158596
[2020-02-16 15:03:47] M50699483 Iter# = 12000000 Res64: FD12342A64D97B8F. shift: 27115678
[2020-02-17 06:06:49] M50699483 Iter# = 13000000 Res64: 2E9753D14381B557. shift: 8538514
[2020-02-17 23:27:22] M50699483 Iter# = 14000000 Res64: DE3C314364459C1B. shift: 17100758
[2020-02-18 13:08:14] M50699483 Iter# = 15000000 Res64: 2EE5C7CC8F97B0B9. shift: 19048758
[2020-02-18 18:25:53] M50699483 Iter# = 16000000 Res64: 42ACBD803C14864F. shift: 14740307
[2020-02-18 23:47:06] M50699483 Iter# = 17000000 Res64: 693F9E609D94F89E. shift: 25311876
[2020-02-19 05:12:41] M50699483 Iter# = 18000000 Res64: 91713ADD0ED97C33. shift: 47268271
[2020-02-19 23:09:48] M50699483 Iter# = 19000000 Res64: 277A592C42FACB53. shift: 1811986
[2020-02-20 04:33:52] M50699483 Iter# = 20000000 Res64: A5682A939EF38D9A. shift: 49290389
[2020-02-20 10:01:20] M50699483 Iter# = 21000000 Res64: 8A8B4492FFC5470B. shift: 36614684
[2020-02-20 15:22:51] M50699483 Iter# = 22000000 Res64: D6D3F91689DDFAF1. shift: 32210977
[2020-02-20 20:44:28] M50699483 Iter# = 23000000 Res64: B48AC590FBA75FE2. shift: 28860043
[2020-02-21 02:08:56] M50699483 Iter# = 24000000 Res64: AD49E3830B9218D2. shift: 50445451
[2020-02-21 07:29:53] M50699483 Iter# = 25000000 Res64: E4382FB2661B845A. shift: 366548
[2020-02-21 12:52:39] M50699483 Iter# = 26000000 Res64: 83C74046877BC1D7. shift: 42829265
[2020-02-21 23:36:31] M50699483 Iter# = 27000000 Res64: EB4330B282026832. shift: 20458504
[2020-02-22 16:51:24] M50699483 Iter# = 28000000 Res64: 54A4F4CAADAAE0F9. shift: 38860600
[2020-02-23 10:15:11] M50699483 Iter# = 29000000 Res64: 9259066017B75695. shift: 11551033
[2020-02-24 03:25:44] M50699483 Iter# = 30000000 Res64: FCEE9793433D6108. shift: 34385914
[2020-02-25 16:59:20] M50699483 Iter# = 31000000 Res64: 0082DF9E0893D9D2. shift: 19736803
[2020-03-02 16:31:16] M50699483 Iter# = 32000000 Res64: 9B2D2709D0339C17. shift: 24549533
[2020-03-03 19:11:20] M50699483 Iter# = 33000000 Res64: 7A2717061965634C. shift: 14550787
[2020-03-04 11:28:49] M50699483 Iter# = 34000000 Res64: 265C4029588A33E5. shift: 38862938
[2020-03-05 15:55:23] M50699483 Iter# = 35000000 Res64: C7EA44FB046C7866. shift: 50647604
[2020-03-06 07:03:11] M50699483 Iter# = 36000000 Res64: FA9720C207C67570. shift: 4745807
[2020-03-06 21:57:14] M50699483 Iter# = 37000000 Res64: 00FBF511991AFED1. shift: 28508816
[2020-03-07 18:07:49] M50699483 Iter# = 38000000 Res64: 9E5851EE2357B20F. shift: 13420991
[2020-03-08 10:31:24] M50699483 Iter# = 39000000 Res64: FBE3CB493A25E922. shift: 9572750
[2020-03-09 01:44:31] M50699483 Iter# = 40000000 Res64: B5BA6759F6360A0C. shift: 18147084
[2020-03-09 17:00:40] M50699483 Iter# = 41000000 Res64: C94D0040BBE2D050. shift: 4939962
[2020-03-10 09:58:22] M50699483 Iter# = 42000000 Res64: 4B934CF3A43F88A2. shift: 41353787
[2020-03-11 03:07:06] M50699483 Iter# = 43000000 Res64: 5E7AEE9752819113. shift: 19740705
[2020-03-12 00:55:07] M50699483 Iter# = 44000000 Res64: A53578C09D91C1BE. shift: 15246709
[2020-03-12 23:38:08] M50699483 Iter# = 45000000 Res64: 561A2C7C7ACDCE8A. shift: 40676835
[2020-03-13 21:23:49] M50699483 Iter# = 46000000 Res64: 4DFDE6A9FF0157A9. shift: 20021361
[2020-03-14 12:26:05] M50699483 Iter# = 47000000 Res64: 7B305450D7F703AA. shift: 20321389
[2020-03-15 03:15:30] M50699483 Iter# = 48000000 Res64: 5D2C15BAA7AD490F. shift: 23413838
[2020-03-15 23:24:49] M50699483 Iter# = 49000000 Res64: 5E37DEB52957C078. shift: 14058798
[2020-03-16 14:40:29] M50699483 Iter# = 50000000 Res64: E1293037BCEA6CE7. shift: 44564712[/code]
FYI, reason for the pokiness is that this is my low-priority "Plan B" Mlucas run, in case the full-priority wavefront run it runs alongside (both using all 4 cores of this non-HT quad) on the same CPU flakes out for any reason - a not-uncommon occurrence on this ever-flaky Haswell CPU. None of my other Mlucas-running devices has ever shown this kind of behavioral tempestuousness.

Uncwilly 2020-03-17 21:09

[QUOTE=kriesel;539975]Thanks. I tried to reserve it on the manual page both before and immediately after claiming it in this thread, and the error message I got in both cases was it was not available, even though pulling up the exponent detail showed no one else had it at the time.[/QUOTE]Because you can't do that via the manual page (for exponents in the lowest categories. I popped it into a Prime95 worktodo and did a manual communication with sending new dates.

kriesel 2020-03-17 21:34

[QUOTE=ewmayer;539977]Ken, if your GPU client uses the convention that LL initial seed is iteration 0, you can compare interim every-1M-iter Res64s vs mine:[/QUOTE]
Looking good through 4M, remainder TBD. This is CUDALucas v2.06 on a GTX1080Ti. It does not output the offset value until the result record. It's logging interim residues every 50K iterations, so if we diverge, we might be able to refine when.

[CODE]Starting M50699483 fft length = 2688K
| Date Time | Test Num Iter Residue | FFT Error ms/It Time | ETA Done |
| Mar 17 14:40:04 | M50699483 1000000 0xff270405ea0d7239 | 2688K 0.26563 2.0281 101.40s | 1:03:59:10 1.97% |
| Mar 17 15:13:53 | M50699483 2000000 0x10bd8630cdc36e73 | 2688K 0.28125 2.0286 101.43s | 1:03:26:06 3.94% |
| Mar 17 15:47:43 | M50699483 3000000 0xd4f95c0fb91ad1ee | 2688K 0.25000 2.0291 101.45s | 1:02:52:47 5.91% |
| Mar 17 16:21:34 | M50699483 4000000 0x2f3d4d841dd790ed | 2688K 0.28125 2.0296 101.48s | 1:02:19:17 7.88% |
[/CODE][QUOTE=Uncwilly;539978]Because you can't do that via the manual page (for exponents in the lowest categories. I popped it into a Prime95 worktodo and did a manual communication with sending new dates.[/QUOTE]Uggh, I think I used to know that.

kriesel 2020-03-18 14:49

Ernst: Do you have interim save files you could resume from? My run is keeping save files spaced every 1M iterations.[QUOTE=ewmayer;539977]Ken, if your GPU client uses the convention that LL initial seed is iteration 0, you can compare interim every-1M-iter Res64s vs mine:
[code][2020-02-11 10:58:54] ...
[2020-03-03 19:11:20] M50699483 Iter# = 33000000 Res64: 7A2717061965634C. shift: 14550787
[2020-03-04 11:28:49] M50699483 Iter# = 34000000 Res64:[B] 265C4029588A33E5[/B]. shift: 38862938
[2020-03-05 15:55:23] M50699483 Iter# = 35000000 Res64: C7EA44FB046C7866. shift: 50647604
[2020-03-06 07:03:11] M50699483 Iter# = 36000000 Res64: FA9720C207C67570. shift: 4745807
[2020-03-06 21:57:14] M50699483 Iter# = 37000000 Res64: 00FBF511991AFED1. shift: 28508816
[2020-03-07 18:07:49] M50699483 Iter# = 38000000 Res64: 9E5851EE2357B20F. shift: 13420991
[2020-03-08 10:31:24] M50699483 Iter# = 39000000 Res64: FBE3CB493A25E922. shift: 9572750
[2020-03-09 01:44:31] M50699483 Iter# = 40000000 Res64: B5BA6759F6360A0C. shift: 18147084
[2020-03-09 17:00:40] M50699483 Iter# = 41000000 Res64: C94D0040BBE2D050. shift: 4939962
[2020-03-10 09:58:22] M50699483 Iter# = 42000000 Res64: 4B934CF3A43F88A2. shift: 41353787
[2020-03-11 03:07:06] M50699483 Iter# = 43000000 Res64: 5E7AEE9752819113. shift: 19740705
[2020-03-12 00:55:07] M50699483 Iter# = 44000000 Res64: A53578C09D91C1BE. shift: 15246709
[2020-03-12 23:38:08] M50699483 Iter# = 45000000 Res64: 561A2C7C7ACDCE8A. shift: 40676835
[2020-03-13 21:23:49] M50699483 Iter# = 46000000 Res64: 4DFDE6A9FF0157A9. shift: 20021361
[2020-03-14 12:26:05] M50699483 Iter# = 47000000 Res64: 7B305450D7F703AA. shift: 20321389
[2020-03-15 03:15:30] M50699483 Iter# = 48000000 Res64: 5D2C15BAA7AD490F. shift: 23413838
[2020-03-15 23:24:49] M50699483 Iter# = 49000000 Res64: 5E37DEB52957C078. shift: 14058798
[2020-03-16 14:40:29] M50699483 Iter# = 50000000 Res64: E1293037BCEA6CE7. shift: 44564712[/code][/QUOTE]
My CUDALucas v2.06 on GTX1080Ti run diverges from yours between 33M and 34M:

[CODE]| Mar 18 07:35:50 | M50699483 31000000 0x0082df9e0893d9d2 | 2688K 0.25000 2.0313 101.56s | 11:06:56 61.14% |
| Mar 18 08:09:39 | M50699483 32000000 0x9b2d2709d0339c17 | 2688K 0.28125 2.0253 101.26s | 10:33:03 63.11% |
| Mar 18 08:43:25 | M50699483 33000000 0x7a2717061965634c | 2688K 0.25000 2.0253 101.26s | 9:59:09 65.08% |
| Mar 18 09:17:11 | M50699483 34000000 [B]0xa22077a84ffa1c25[/B] | 2688K 0.26563 2.0254 101.27s | 9:25:15 67.06% |
[/CODE]At finer granularity, with notes on progress of repeat from 33M:[CODE]| Mar 18 08:45:06 | M50699483 33050000 0x37c64cf8f8be0f90 | 2688K 0.26563 2.0362 101.81s | 9:57:28 65.18% | reproduced
| Mar 18 08:46:47 | M50699483 33100000 0xbb849ff55738c8d5 | 2688K 0.26563 2.0253 101.26s | 9:55:46 65.28% | reproduced
| Mar 18 08:48:29 | M50699483 33150000 0x1cd84913571585e1 | 2688K 0.26563 2.0253 101.26s | 9:54:04 65.38% | reproduced
| Mar 18 08:50:10 | M50699483 33200000 0xef8c9ea2bb982cd2 | 2688K 0.26563 2.0254 101.27s | 9:52:23 65.48% | reproduced
| Mar 18 08:51:51 | M50699483 33250000 0xe80447b573ef030a | 2688K 0.25000 2.0253 101.26s | 9:50:41 65.58% | reproduced
| Date Time | Test Num Iter Residue | FFT Error ms/It Time | ETA Done |
| Mar 18 08:53:32 | M50699483 33300000 0x0ec1cc2621d92648 | 2688K 0.26563 2.0254 101.26s | 9:48:59 65.68% | reproduced
| Mar 18 08:55:14 | M50699483 33350000 0xac334a394a51f8df | 2688K 0.28125 2.0253 101.26s | 9:47:17 65.77% | reproduced
| Mar 18 08:56:55 | M50699483 33400000 0x5647ecc4a6bfe11b | 2688K 0.26563 2.0252 101.26s | 9:45:36 65.87% | reproduced
| Mar 18 08:58:36 | M50699483 33450000 0xab3476d0c81b41cc | 2688K 0.25000 2.0252 101.26s | 9:43:54 65.97% | reproduced
| Mar 18 09:00:17 | M50699483 33500000 0x64d625bfc4429707 | 2688K 0.28125 2.0254 101.27s | 9:42:12 66.07% | reproduced
| Mar 18 09:01:59 | M50699483 33550000 0xdf554459fc451fe2 | 2688K 0.28125 2.0253 101.26s | 9:40:31 66.17% | reproduced
| Mar 18 09:03:40 | M50699483 33600000 0x219a46abd5061aa9 | 2688K 0.25000 2.0253 101.26s | 9:38:49 66.27% | reproduced
| Mar 18 09:05:21 | M50699483 33650000 0xccc27947bc7707c8 | 2688K 0.28125 2.0252 101.26s | 9:37:07 66.37% | reproduced
| Mar 18 09:07:03 | M50699483 33700000 0xf77df0cc388e7949 | 2688K 0.26563 2.0254 101.27s | 9:35:26 66.47% | reproduced
| Mar 18 09:08:44 | M50699483 33750000 0xa64fba39022d141f | 2688K 0.28125 2.0254 101.27s | 9:33:44 66.56% | reproduced
| Mar 18 09:10:25 | M50699483 33800000 0x56e89f72185d56f7 | 2688K 0.25000 2.0253 101.26s | 9:32:02 66.66% | reproduced
| Mar 18 09:12:06 | M50699483 33850000 0x0cbcd708bbcfae8b | 2688K 0.26563 2.0253 101.26s | 9:30:20 66.76% | reproduced
| Mar 18 09:13:48 | M50699483 33900000 0x820a85c3d8c9752c | 2688K 0.26563 2.0253 101.26s | 9:28:39 66.86% | reproduced
| Mar 18 09:15:29 | M50699483 33950000 0x54799a3e9105fa36 | 2688K 0.25000 2.0253 101.26s | 9:26:57 66.96% | reproduced
| Mar 18 09:17:11 | M50699483 34000000 0xa22077a84ffa1c25 | 2688K 0.26563 2.0254 101.27s | 9:25:15 67.06% | reproduced
[/CODE]Halting repeat run, resuming from 35M+ save file.

ewmayer 2020-03-18 18:29

Ken, Mlucas saves every-10M savefiles by default. Residue-reporting granularity on runs of <= 4 threads is every 10Kiter, I have narrowed the divergence to the 33.10-33.15Miter interval, the 2 Res64s which bookend that are bolded in my logfile excerpt below:
[code][2020-03-03 20:39:45] M50699483 Iter# = 33100000 Res64: [b]BB849FF55738C8D5[/b]. AvgMaxErr = 0.071159536. MaxErr = 0.109375000. Residue shift count = 9258343.
[2020-03-03 20:48:40] M50699483 Iter# = 33110000 Res64: 5B7BD8EA7D60E545. AvgMaxErr = 0.071040688. MaxErr = 0.109375000. Residue shift count = 24888588.
[2020-03-03 20:57:28] M50699483 Iter# = 33120000 Res64: 98FBFA574398CDC8. AvgMaxErr = 0.071026953. MaxErr = 0.109375000. Residue shift count = 4513422.
[2020-03-03 21:06:14] M50699483 Iter# = 33130000 Res64: DBEC805AFE89F02C. AvgMaxErr = 0.071080518. MaxErr = 0.109375000. Residue shift count = 26304104.
M50699483 Roundoff warning on iteration 33139667, maxerr = 0.500000000000
Retrying iteration interval to see if roundoff error is reproducible.
Restarting M50699483 at iteration = 33130000. Res64: DBEC805AFE89F02C, residue shift count = 26304104
M50699483: using FFT length 2816K = 2883584 8-byte floats, initial residue shift count = 26304104
this gives an average 17.582107197154652 bits per digit
Retry of iteration interval with fatal roundoff error was successful.
[2020-03-03 21:23:43] M50699483 Iter# = 33140000 Res64: 3FE1873AE0CFDBAD. AvgMaxErr = 0.071077051. MaxErr = 0.160156250. Residue shift count = 10073663.
[2020-03-03 21:32:33] M50699483 Iter# = 33150000 Res64: [b]68D413A52601DAB7[/b]. AvgMaxErr = 0.071095947. MaxErr = 0.101562500. Residue shift count = 12690212.[/code]
Note the kind of data-corruption-detected sudden-fatal-ROE as seen above is not at all unusual on my notoriously flaky Haswell system - I find running in such a context to be a valuable QA exercise because it approximates the worst-case scenario users of my code may face with their own hardware. So you can see from the above the program detected a glitch in the matrix, as a result of which it restarted from the iteration = 33130000 savefile, and retry of the ensuing 10Kiter interval was successful. But for every such detected data-corruption error there is a smaller number of 'silent' ones, which are the reason the PRP+Gerbicz option is so valuable on this kind of hardware - my current production runs are doing PRP-tests, this DC was just last of a long-running low-priority LL-DC batch.

Ken, Uncwilly, can either of you rerun the same 50K interval with the finer 10Kiter reporting granularity? It would be intersting to see if the divergence occurred before the above fatal-ROE/retry incident, or after.

In the meantime I've low-prioritized my 2 production tests (one main, one lower-priority 'backup' run in case the first crashes) on the Haswell and restarted the run from the 30M savefile. But weirdly, doing a 'top' just now I see my 2 production runs are listed at the lowest priority (renice -n 19) but are still grabbing cycles at a higher priority than the retry-DC-from-30M run ... 'sudo renice -n -4' of the latter doesn't help, DC run still just getting 1 core's worth of cycles. Frickin' Ubuntu ... so 'fg' both production jobs and ctrl-z to suspend them. That's better ... with ~3Miter to go, my DC-retry run will hit the divergence point in ~7 hours.

kriesel 2020-03-18 19:32

[QUOTE=ewmayer;540068]Ken, Mlucas saves every-10M savefiles by default. Residue-reporting granularity on runs of <= 4 threads is every 10Kiter, I have narrowed the divergence to the 33.10-33.15Miter interval, the 2 Res64s which bookend that are bolded in my logfile excerpt below
...

Ken, Uncwilly, can either of you rerun the same 50K interval with the finer 10Kiter reporting granularity? It would be interesting to see if the divergence occurred before the above fatal-ROE/retry incident, or after.[/QUOTE]Okay. Forked it to a GTX1080 on a different system that was doing interruptible P-1 in gpuowl, CUDALucas v2.06 at 5K iterations console printout, 100K save file interval, from 33M save file.
[CODE]CUDALucas v2.06beta 64-bit build, compiled May 5 2017 @ 13:02:54

binary compiled for CUDA 8.0
CUDA runtime version 8.0
CUDA driver version 8.0

------- DEVICE 0 -------
name GeForce GTX 1080
UUID GPU-5e2c5531-4684-57ec-6393-8b762f286c70
ECC Support? Disabled
Compatibility 6.1
clockRate (MHz) 1797
memClockRate (MHz) 5005
totalGlobalMem 8589934592
totalConstMem 65536
l2CacheSize 2097152
sharedMemPerBlock 49152
regsPerBlock 65536
warpSize 32
memPitch 2147483647
maxThreadsPerBlock 1024
maxThreadsPerMP 2048
multiProcessorCount 20
maxThreadsDim[3] 1024,1024,64
maxGridSize[3] 2147483647,65535,65535
textureAlignment 512
deviceOverlap 1
pciDeviceID 0
pciBusID 4

You may experience a small delay on 1st startup to due to Just-in-Time Compilation

Using threads: square 32, splice 128.

Continuing M50699483 @ iteration 33000001 with fft length 2688K, 65.09% done

| Date Time | Test Num Iter Residue | FFT Error ms/It Time | ETA Done |
| Mar 18 14:17:10 | M50699483 33005000 0xdd429c1ccbd7a62a | 2688K 0.25000 2.8121 14.05s | 9:59:01 65.09% |
| Mar 18 14:17:24 | M50699483 33010000 0xff4bbbe125ee1176 | 2688K 0.23438 2.8099 14.04s | 9:58:53 65.10% |
| Mar 18 14:17:38 | M50699483 33015000 0x7bc220867e3d7921 | 2688K 0.25000 2.8094 14.04s | 9:58:45 65.11% |
| Mar 18 14:17:52 | M50699483 33020000 0x4ab93d7fb1b9410a | 2688K 0.25000 2.8088 14.04s | 9:58:37 65.12% |
| Mar 18 14:18:06 | M50699483 33025000 0x0dd3e63997a7117e | 2688K 0.25000 2.8055 14.02s | 9:58:29 65.13% |
| Mar 18 14:18:20 | M50699483 33030000 0xdb45cff854629006 | 2688K 0.25000 2.8100 14.05s | 9:58:21 65.14% |
| Mar 18 14:18:34 | M50699483 33035000 0xc927fe56adabaf33 | 2688K 0.25000 2.8171 14.08s | 9:58:13 65.15% |
| Mar 18 14:18:48 | M50699483 33040000 0x671966f21715efe7 | 2688K 0.25000 2.8176 14.08s | 9:58:05 65.16% |
| Mar 18 14:19:02 | M50699483 33045000 0xc24775f499646abb | 2688K 0.25000 2.8184 14.09s | 9:57:56 65.17% |
| Mar 18 14:19:16 | M50699483 33050000 0x37c64cf8f8be0f90 | 2688K 0.25000 2.8173 14.08s | 9:57:48 65.18% |
| Mar 18 14:19:31 | M50699483 33055000 0xea0b3d9144c22bab | 2688K 0.23438 2.8408 14.20s | 9:57:40 65.19% |
| Mar 18 14:19:45 | M50699483 33060000 0xf4e172ae2efbf3c0 | 2688K 0.25000 2.8356 14.17s | 9:57:32 65.20% |
| Mar 18 14:19:59 | M50699483 33065000 0x72232fa331f55cf9 | 2688K 0.25000 2.8219 14.10s | 9:57:24 65.21% |
| Mar 18 14:20:13 | M50699483 33070000 0x207baea46fe47f92 | 2688K 0.23438 2.8226 14.11s | 9:57:16 65.22% |
| Mar 18 14:20:27 | M50699483 33075000 0x07d271a17b19c3d5 | 2688K 0.21875 2.8245 14.12s | 9:57:08 65.23% |
| Mar 18 14:20:41 | M50699483 33080000 0x1ee9e9c6554ccf1e | 2688K 0.25000 2.8241 14.12s | 9:57:00 65.24% |
| Mar 18 14:20:55 | M50699483 33085000 0xdc08ddc9d7820922 | 2688K 0.25000 2.8248 14.12s | 9:56:52 65.25% |
| Mar 18 14:21:10 | M50699483 33090000 0xd19878f22036fac1 | 2688K 0.21875 2.8245 14.12s | 9:56:44 65.26% |
| Mar 18 14:21:24 | M50699483 33095000 0x3e5e1634ab32f2e3 | 2688K 0.25000 2.8275 14.13s | 9:56:36 65.27% |
| Date Time | Test Num Iter Residue | FFT Error ms/It Time | ETA Done |
| Mar 18 14:21:38 | M50699483 33100000 0xbb849ff55738c8d5 | 2688K 0.25000 2.8287 14.14s | 9:56:28 65.28% |
| Mar 18 14:21:52 | M50699483 33105000 0xfe2846d8d8d8e12f | 2688K 0.25000 2.9227 14.61s | 9:56:20 65.29% |
| Mar 18 14:22:07 | M50699483 33110000 0x5b7bd8ea7d60e545 | 2688K 0.25000 2.8426 14.21s | 9:56:12 65.30% |
| Mar 18 14:22:21 | M50699483 33115000 0x3dc9f02704f5e3cc | 2688K 0.25000 2.8421 14.21s | 9:56:04 65.31% |
| Mar 18 14:22:35 | M50699483 33120000 0x98fbfa574398cdc8 | 2688K 0.25000 2.8424 14.21s | 9:55:56 65.32% |
| Mar 18 14:22:49 | M50699483 33125000 0x3c08397825944984 | 2688K 0.25000 2.8407 14.20s | 9:55:48 65.33% |
| Mar 18 14:23:03 | M50699483 33130000 [B][COLOR=SeaGreen]0xdbec805afe89f02c[/COLOR][/B] | 2688K 0.25000 2.8380 14.19s | 9:55:40 65.34% |
| Mar 18 14:23:18 | M50699483 33135000 0xcc37b1c7d24241ee | 2688K 0.25000 2.8364 14.18s | 9:55:32 65.35% |
| Mar 18 14:23:32 | M50699483 33140000 [B][COLOR=Red]0x067c546da1f13507[/COLOR][/B] | 2688K 0.25000 2.8237 14.11s | 9:55:24 65.36% |
| Mar 18 14:23:46 | M50699483 33145000 0x62d3b4b0d086b501 | 2688K 0.26563 2.8250 14.12s | 9:55:16 65.37% |
| Mar 18 14:24:00 | M50699483 33150000 0x1cd84913571585e1 | 2688K 0.23438 2.8247 14.12s | 9:55:08 65.38% |
| Mar 18 14:24:14 | M50699483 33155000 0x5b55b049e6038c81 | 2688K 0.24219 2.8244 14.12s | 9:55:00 65.39% |
| Mar 18 14:24:28 | M50699483 33160000 0x4fa86767dd1f39bf | 2688K 0.25000 2.8245 14.12s | 9:54:52 65.40% |
| Mar 18 14:24:42 | M50699483 33165000 0x2d062794c09182d6 | 2688K 0.23438 2.8246 14.12s | 9:54:43 65.41% |
| Mar 18 14:24:56 | M50699483 33170000 0xf644b10ec263f600 | 2688K 0.25000 2.8246 14.12s | 9:54:35 65.42% |
| Mar 18 14:25:11 | M50699483 33175000 0xf9a9dc418fc724ea | 2688K 0.25000 2.8249 14.12s | 9:54:27 65.43% |
| Mar 18 14:25:25 | M50699483 33180000 0xb461f655a081955b | 2688K 0.25000 2.8244 14.12s | 9:54:19 65.44% |
| Mar 18 14:25:39 | M50699483 33185000 0xf43ce356e8e26ea7 | 2688K 0.22461 2.8258 14.12s | 9:54:11 65.45% |
| Mar 18 14:25:53 | M50699483 33190000 0xba94f36ee57b6775 | 2688K 0.24219 2.8256 14.12s | 9:54:03 65.46% |
| Date Time | Test Num Iter Residue | FFT Error ms/It Time | ETA Done |
| Mar 18 14:26:07 | M50699483 33195000 0x13135c7b4ff819a7 | 2688K 0.23438 2.8256 14.12s | 9:53:55 65.47% |
| Mar 18 14:26:22 | M50699483 33200000 [B]0xef8c9ea2bb982cd2[/B] | 2688K 0.25000 2.8248 14.12s | 9:53:47 65.48% |[/CODE]Matches the earlier [B]GTX1080Ti[/B] residues, reposted below:

[CODE]| Date Time | Test Num Iter Residue | FFT Error ms/It Time | ETA Done |
| Mar 18 09:57:27 | M50699483 33050000 0x37c64cf8f8be0f90 | 2688K 0.25000 2.0097 100.48s | 9:57:27 65.18% |
| Mar 18 09:59:08 | M50699483 33100000 0xbb849ff55738c8d5 | 2688K 0.25000 2.0190 100.95s | 9:55:45 65.28% |
| Mar 18 10:00:49 | M50699483 33150000 0x1cd84913571585e1 | 2688K 0.26563 2.0231 101.15s | 9:54:03 65.38% |
| Mar 18 10:02:30 | M50699483 33200000 [B]0xef8c9ea2bb982cd2[/B] | 2688K 0.25000 2.0237 101.18s | 9:52:22 65.48% |
[/CODE]CUDALucas is doing well, considering it hasn't even the Jacobi check.

ewmayer 2020-03-18 19:46

Ken, is every-10K reporting an option for you?

kriesel 2020-03-18 20:01

[QUOTE=ewmayer;540079]Ken, is every-10K reporting an option for you?[/QUOTE]Yes, and finer or coarser. If Mlucas can't adjust, you may want to edit mlucas source and recompile so you can look at where the divergence occurs at finer resolution when divergences occur.
Fine/frequent output can be edited down. Coarse/infrequent has to be rerun when questions arise, which might be enlightened by more frequent res64 output.
I was not sure what the lower limit is in CUDALucas. It is apparently a single iteration. I'm not sure what the upper limit is, but I've run 10[SUP]5[/SUP].
From CUDALucas.ini:[CODE]# ErrorIterations tells how often the roundoff error is checked. Larger values
# give shorter iteration times, but introduce some uncertainty as to the actual
# maximum roundoff error that occurs during the test. Default is 100.
# ReportIterations is the same as the -x option; it determines how often
# screen output is written. Default is 10000.
# CheckpointIterations is the same as the -c option; it determines how often
# checkpoints are written. Default is 100000.
# Each of these values should be of the form k * 10^n with k = 1, 2, or 5.

ErrorIterations=100
ReportIterations=50000
CheckpointIterations=1000000[/CODE]A quick experiment shows that ReportIterations = 1 works, although the speed penalty is considerable relative to the 2 ms/iter normal case, and the log file would explode in size:[CODE]| Date Time | Test Num Iter Residue | FFT Error ms/It Time | ETA Done |
| Mar 18 14:55:54 | M50699483 42788602 0xc672edd2fec9d543 | 2688K 0.25000 1.#INF 0.00s | 4:27:44 84.39% |
| Mar 18 14:55:54 | M50699483 42788603 0x073e9eb282310661 | 2688K 0.20313 10.4790 0.01s | 4:27:44 84.39% |
| Mar 18 14:55:54 | M50699483 42788604 0xbe3d42c802206844 | 2688K 0.20313 9.8760 0.00s | 4:27:44 84.39% |[/CODE]There's little speed penalty at ReportIterations = 100 or higher.[CODE]Iter time on GTX1080Ti, CUDALucas v2.06, M53M, versus ReportIterations
# ms/iter
1 9.5
10 2.75
100 2.05
1000 2.02
10000 2.02
50000 2.03[/CODE]

kriesel 2020-03-18 21:15

Ironically, Ernst's res64 was:[CODE][2020-03-03 21:23:43] M50699483 Iter# = 33140000 Res64: [B]3FE1873AE0CFD[/B][COLOR=Red][B]BAD[/B][/COLOR]. AvgMaxErr = 0.071077051. MaxErr = 0.160156250. Residue shift count = 10073663.
[/CODE]vs. my CUDALucas output[CODE]| Mar 18 14:23:03 | M50699483 33130000 0xdbec805afe89f02c | 2688K 0.25000 2.8380 14.19s | 9:55:40 65.34% |
| Mar 18 14:23:18 | M50699483 33135000 0xcc37b1c7d24241ee | 2688K 0.25000 2.8364 14.18s | 9:55:32 65.35% |
| Mar 18 14:23:32 | M50699483 33140000 [B]0x067c546da1f13507[/B] | 2688K 0.25000 2.8237 14.11s | 9:55:24 65.36% |
[/CODE]

ewmayer 2020-03-18 21:57

[QUOTE=kriesel;540089]Ironically, Ernst's res64 was:[snip][/QUOTE]

LOL, I hadn't noticed that the Res64 was itself confessing its badness! :) Anyhow, once my re-run-from-30M gets close, I'll save a copy of the restartfile in case I want to re-run a small subinterval with finer-than-10K granularity.

Ken, I assume your DC run has completed or is close? Still don't see your result appearing on the exponent page.

Uncwilly 2020-03-18 22:08

[code][2020-02-13 23:10:28] M50699483 Iter# = 6000000 Res64: 0C3E78B1C77FA688. shift: 33080656[/code]
[CODE][Mar 18 14:19] M50699483 interim LL residue AF0BF1AEDBD6D468 at iteration 6000000[/CODE]
I guess my shift is not zero.


All times are UTC. The time now is 02:48.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2022, Jelsoft Enterprises Ltd.