mersenneforum.org  

Go Back   mersenneforum.org > Search Forums

Showing results 1 to 25 of 1000
Search took 0.44 seconds.
Search: Posts Made By: Prime95
Forum: PrimeNet 2022-11-22, 22:54
Replies: 5
Views: 250
Posted By Prime95
Make sure your user id is entered properly in the...

Make sure your user id is entered properly in the Test/Primenet dialog box?
Forum: PrimeNet 2022-11-21, 21:15
Replies: 24
Views: 1,015
Posted By Prime95
Be careful. Measure the...

Be careful. Measure the change-in-B2-completed-values divided by time. Using fewer cores may result in a lower target B2 value.
Forum: Software 2022-11-17, 19:55
Replies: 526
Views: 120,402
Posted By Prime95
Depending on the CPU, multithreading isn't even...

Depending on the CPU, multithreading isn't even possible until around 250000 bits and isn't really helpful until around 1.3 million bits or more.

For these small numbers you are better off running...
Forum: Software 2022-11-17, 00:13
Replies: 769
Views: 82,148
Posted By Prime95
No. 30.9 is pre-beta and development is...

No. 30.9 is pre-beta and development is progressing slowly.
Forum: Software 2022-11-15, 03:10
Replies: 769
Views: 82,148
Posted By Prime95
Found it. On Sept. 27 a bug was fixed on...

Found it. On Sept. 27 a bug was fixed on hyperthreaded torture tests with more than 64GB of memory and FFT sizes below 256K.
Forum: PrimeNet 2022-11-14, 21:28
Replies: 10
Views: 876
Posted By Prime95
According to...

According to https://www.mersenne.org/report_exponent/?exp_lo=107349443&exp_hi=&full=1
the proof was uploaded and certified. Digging further indicates that the proof file was only partially...
Forum: Software 2022-11-14, 20:35
Replies: 769
Views: 82,148
Posted By Prime95
That FFT size sounds familiar, but I don't see...

That FFT size sounds familiar, but I don't see anything relevant in the change logs. Have you tried 30.9b3? Is this reproducible (fails the same way every time)?

If this is a new Intel chip,...
Forum: Marin's Mersenne-aries 2022-11-14, 04:18
Replies: 31
Views: 81,874
Posted By Prime95
If I did the query correctly: 34643591 ...

If I did the query correctly:

34643591
34696567
35184673
35381377
35478853
36142801
36211067
36313813
Forum: Marin's Mersenne-aries 2022-11-13, 20:59
Replies: 31
Views: 81,874
Posted By Prime95
IIRC, there are no CudaLucas FFT internals,...

IIRC, there are no CudaLucas FFT internals, simply a call to the CUDA FFT library. That doesn't mean the bug isn't in CudaLucas, there is the weighting and carry propagation code to consider.
...
Forum: Hardware 2022-11-12, 20:04
Replies: 4
Views: 207
Posted By Prime95
Sounds like what you are doing is working. ...

Sounds like what you are doing is working. Errors setting affinity will be harmless in the setup you described.

If the error message bothers you, try setting "NumCPUs=1" in local.txt.
Forum: Hardware 2022-11-12, 19:58
Replies: 20
Views: 3,028
Posted By Prime95
Since this will be used to select default FFTs...

Since this will be used to select default FFTs for many years to come, let's go with the faster settings.
Forum: y-cruncher 2022-11-12, 06:09
Replies: 11
Views: 1,109
Posted By Prime95
Grrr, it seems the v5-2680v4 has an inclusive L3...

Grrr, it seems the v5-2680v4 has an inclusive L3 cache.
Forum: y-cruncher 2022-11-12, 03:45
Replies: 11
Views: 1,109
Posted By Prime95
I can't see how the mesh bus vs ring bus would...

I can't see how the mesh bus vs ring bus would adversely affect my code. My bet is the change from an inclusive L3 cache to an exclusive L3 cache. I haven't come up with a theory as to how this is...
Forum: PrimeNet 2022-11-12, 02:17
Replies: 10
Views: 876
Posted By Prime95
Which exponent? Have you tried submitting your...

Which exponent?
Have you tried submitting your results.json.txt file to https://www.mersenne.org/manual_result/
Forum: y-cruncher 2022-11-10, 02:01
Replies: 11
Views: 1,109
Posted By Prime95
Writes were normal writes. I just tried...

Writes were normal writes. I just tried NT-stores and it helps by about 1% (all cases).



I have two different multi-threading options. The 20% hit takes place in both strategies.

Strategy...
Forum: y-cruncher 2022-11-09, 21:02
Replies: 11
Views: 1,109
Posted By Prime95
The gwnum coefficients are first FFTed using...

The gwnum coefficients are first FFTed using gwnum. Polymult (in AVX) now takes 4 complex values (64 bytes) from every gwnum and does forward FFT, pointwise multiply, inverse FFT, write back to the...
Forum: y-cruncher 2022-11-09, 15:45
Replies: 11
Views: 1,109
Posted By Prime95
Adjacent line prefetch is likely on. However, on...

Adjacent line prefetch is likely on. However, on Skylake-X where I'm getting the 20% performance drop I actually read and write 128-byte chunks because its using AVX-512 instructions.



I tried...
Forum: y-cruncher 2022-11-09, 02:47
Replies: 11
Views: 1,109
Posted By Prime95
Strided memory access

I have a mystery that perhaps Mystical can provide some insight.

Some background info, the details of which aren't extremely important. I'm optimizing the polynomial multiplication code where...
Forum: PrimeNet 2022-11-08, 02:01
Replies: 2,640
Views: 357,105
Posted By Prime95
True, but the extra reported curves won't affect...

True, but the extra reported curves won't affect future ECM assignments very much.
Forum: GPU Computing 2022-11-07, 03:24
Replies: 10
Views: 1,250
Posted By Prime95
Thanks for the insight. Yes, gpuowl on Radeon...

Thanks for the insight. Yes, gpuowl on Radeon VII is near or at max memory bandwidth which is why Radeon VII Pro benchmarks are not much faster than a Radeon VII.



This probably belongs in a...
Forum: GPU Computing 2022-11-06, 19:04
Replies: 10
Views: 1,250
Posted By Prime95
Something weird with the chart. The 6950 has...

Something weird with the chart. The 6950 has half the memory bandwidth and half the FP64 throughput but comes out faster than the VII?

BTW, I just tuned some of my Radeon VIIs for maximum energy...
Forum: PrimeNet 2022-11-06, 18:46
Replies: 10
Views: 876
Posted By Prime95
If you have any save files from late in the PRP...

If you have any save files from late in the PRP run, save them somewhere in case we need to re-run the end of the PRP test.

I've not seen this behavior before, please email results.txt,...
Forum: Hardware 2022-11-06, 18:37
Replies: 20
Views: 3,028
Posted By Prime95
A big ask. Can you run an all FFT...

A big ask. Can you run an all FFT implementations benchmark on 16 cores, 2 workers, no hyperthreading for all FFT sizes from 1024K to the maximum 65536K? Email the results to me and I can make your...
Forum: PrimeNet 2022-11-04, 22:14
Replies: 1
Views: 289
Posted By Prime95
P-1 results where a factor is not found are not...

P-1 results where a factor is not found are not verifiable at this time.
Forum: Software 2022-11-02, 22:56
Replies: 6
Views: 441
Posted By Prime95
The Test/Primenet dialog box should have the "Use...

The Test/Primenet dialog box should have the "Use PrimeNet..." checkbox marked.
Showing results 1 to 25 of 1000

 
All times are UTC. The time now is 04:34.


Tue Nov 29 04:34:23 UTC 2022 up 103 days, 2:02, 0 users, load averages: 1.31, 1.10, 1.07

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

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.

≠ ± ∓ ÷ × · − √ ‰ ⊗ ⊕ ⊖ ⊘ ⊙ ≤ ≥ ≦ ≧ ≨ ≩ ≺ ≻ ≼ ≽ ⊏ ⊐ ⊑ ⊒ ² ³ °
∠ ∟ ° ≅ ~ ‖ ⟂ ⫛
≡ ≜ ≈ ∝ ∞ ≪ ≫ ⌊⌋ ⌈⌉ ∘ ∏ ∐ ∑ ∧ ∨ ∩ ∪ ⨀ ⊕ ⊗ 𝖕 𝖖 𝖗 ⊲ ⊳
∅ ∖ ∁ ↦ ↣ ∩ ∪ ⊆ ⊂ ⊄ ⊊ ⊇ ⊃ ⊅ ⊋ ⊖ ∈ ∉ ∋ ∌ ℕ ℤ ℚ ℝ ℂ ℵ ℶ ℷ ℸ 𝓟
¬ ∨ ∧ ⊕ → ← ⇒ ⇐ ⇔ ∀ ∃ ∄ ∴ ∵ ⊤ ⊥ ⊢ ⊨ ⫤ ⊣ … ⋯ ⋮ ⋰ ⋱
∫ ∬ ∭ ∮ ∯ ∰ ∇ ∆ δ ∂ ℱ ℒ ℓ
𝛢𝛼 𝛣𝛽 𝛤𝛾 𝛥𝛿 𝛦𝜀𝜖 𝛧𝜁 𝛨𝜂 𝛩𝜃𝜗 𝛪𝜄 𝛫𝜅 𝛬𝜆 𝛭𝜇 𝛮𝜈 𝛯𝜉 𝛰𝜊 𝛱𝜋 𝛲𝜌 𝛴𝜎𝜍 𝛵𝜏 𝛶𝜐 𝛷𝜙𝜑 𝛸𝜒 𝛹𝜓 𝛺𝜔