Prime95 version 29.6/29.7/29.8

[harlee]

Originally Posted by Prime95

Build 3 is now available. This should fix the reported bugs. Please re-test (some of the reported bugs on older hardware I cannot test here).

Looks good so far. Below are some test results on my Windows 32-bit system.

[Main thread Feb 27 00:34] Mersenne number primality test program version 29.6
[Main thread Feb 27 00:34] Optimizing for CPU architecture: Pentium 4, L2 cache size: 512 KB, L3 cache size: -1 KB

Since the P4 doesn't have L3 cache it is being reported as -1 KB. No worries here.

[Wed Feb 27 00:21:29 2019]
Compare your results to other computers at http://www.mersenne.org/report_benchmarks
Intel(R) Pentium(R) 4 CPU 2.60GHz
CPU speed: 2593.85 MHz, with hyperthreading
CPU features: Prefetch, SSE, SSE2
L1 cache size: 8 KB, L2 cache size: 512 KB, L3 cache size: -1 KB
L1 cache line size: 64 bytes, L2 cache line size: 128 bytesMachine topology as determined by hwloc library:
 Machine#0 (total=2074292KB, Backend=Windows, hwlocVersion=2.0.3, ProcessName=prime95.exe)
  Package (total=2074292KB, CPUVendor=GenuineIntel, CPUFamilyNumber=15, CPUModelNumber=2, CPUModel="Intel(R) Pentium(R) 4 CPU 2.60GHz", CPUStepping=9)
    Core (cpuset: 0x00000003)
      PU#0 (cpuset: 0x00000001)
      PU#1 (cpuset: 0x00000002)
Prime95 32-bit version 29.6, RdtscTiming=1
FFTlen=288K, Type=3, Arch=3, Pass1=384, Pass2=768, clm=4 (1 core, 1 worker): 15.51 ms.  Throughput: 64.48 iter/sec.
FFTlen=288K, Type=3, Arch=2, Pass1=384, Pass2=768, clm=4 (1 core, 1 worker): 15.37 ms.  Throughput: 65.06 iter/sec.
FFTlen=288K, Type=3, Arch=1, Pass1=384, Pass2=768, clm=2 (1 core, 1 worker): 14.22 ms.  Throughput: 70.32 iter/sec.
FFTlen=288K, Type=3, Arch=2, Pass1=384, Pass2=768, clm=2 (1 core, 1 worker): 15.65 ms.  Throughput: 63.91 iter/sec.
FFTlen=288K, Type=2, Arch=3, Pass1=384, Pass2=768, clm=4 (1 core, 1 worker): 15.70 ms.  Throughput: 63.70 iter/sec.
FFTlen=288K, Type=2, Arch=2, Pass1=384, Pass2=768, clm=4 (1 core, 1 worker): 15.55 ms.  Throughput: 64.31 iter/sec.
FFTlen=288K, Type=2, Arch=1, Pass1=384, Pass2=768, clm=2 (1 core, 1 worker): 14.66 ms.  Throughput: 68.21 iter/sec.
FFTlen=288K, Type=2, Arch=2, Pass1=384, Pass2=768, clm=2 (1 core, 1 worker): 15.77 ms.  Throughput: 63.39 iter/sec.

I can provide the benchmark results for 29.4b7 if you want to compare the results. Likes like the build 3 is reading the gwnum.txt file properly. Question, how do we know that the correct Type / Arch is being selected?

[Feb 27 00:34] Using Pentium4 FFT length 288K, Pass1=384, Pass2=768, clm=2

[Prime95]

Originally Posted by harlee

Question, how do we know that the correct Type / Arch is being selected?

[Feb 27 00:34] Using Pentium4 FFT length 288K, Pass1=384, Pass2=768, clm=2

I don't think you can tell (other than if the timings are way off).

[kriesel]

Originally Posted by Prime95

Build 3 is now available. This should fix the reported bugs. Please re-test (some of the reported bugs on older hardware I cannot test here).

Prime95 v29.6b3 on Vista, Pentium M 750. "[Main thread Feb 26 18:23] Mersenne number primality test program version 29.6
[Main thread Feb 26 18:23] Optimizing for CPU architecture: Pentium M, L2 cache size: 2 MB, L3 cache size: -1 KB" CPU-Z indicates no L3 memory on this system.

The skipping a residue at stop/continuedescribed in post 4 is still occurring in build 3; note iteration #71 interim residue is missing in the following.

[Feb 26 18:38] Iteration: 68 / 82589933 [0.00%], ms/iter: 443.897, ETA: 424d 07:43
[Feb 26 18:38] M82589933 interim LL residue CCCA452E1C116AB5 at iteration 68
[Feb 26 18:38] Iteration: 69 / 82589933 [0.00%], ms/iter: 416.372, ETA: 398d 00:14
[Feb 26 18:38] M82589933 interim LL residue 43F6A14DA66096A3 at iteration 69
[Feb 26 18:38] Iteration: 70 / 82589933 [0.00%], ms/iter: 406.014, ETA: 388d 02:37
[Feb 26 18:38] M82589933 interim LL residue 7FC9987DAFA68A81 at iteration 70
[Feb 26 18:38] Iteration: 71 / 82589933 [0.00%], ms/iter: 409.558, ETA: 391d 11:56
[Feb 26 18:38] Stopping primality test of M82589933 at iteration 71 [0.00%]
[Feb 26 18:38] Worker stopped.
[Feb 26 18:39] Worker starting
[Feb 26 18:39] Running Jacobi error check.  Passed.  Time: 259.729 sec.
[Feb 26 18:43] Resuming primality test of M82589933 using Pentium4 FFT length 4480K, Pass1=896, Pass2=5K, clm=4
[Feb 26 18:43] Iteration: 72 / 82589933 [0.00%].
[Feb 26 18:43] M82589933 interim LL residue 9E38150B427A99E1 at iteration 72
[Feb 26 18:43] Iteration: 73 / 82589933 [0.00%], ms/iter: 208.759, ETA: 199d 13:15
[Feb 26 18:43] M82589933 interim LL residue 3260B712A16C7199 at iteration 73
[Feb 26 18:43] Iteration: 74 / 82589933 [0.00%], ms/iter: 409.600, ETA: 391d 12:52
[Feb 26 18:43] M82589933 interim LL residue BE64D87AE6907EE7 at iteration 74

The offset appearance of PRP interim residues, and skipped interim residue at stop start of PRP also remain. Here, output of interim residue for iteration 36 is skipped.

[Feb 26 18:58] Iteration: 34 / 1257787 [0.00%], ms/iter:  4.300, ETA: 01:30:08
[Feb 26 18:58] M1257787 interim PRP residue 199770DE73DBE343 at iteration 33
[Feb 26 18:58] Iteration: 35 / 1257787 [0.00%], ms/iter:  3.852, ETA: 01:20:45
[Feb 26 18:58] M1257787 interim PRP residue BFCE96FBA995F5E1 at iteration 34
[Feb 26 18:58] Iteration: 36 / 1257787 [0.00%], ms/iter:  4.170, ETA: 01:27:24
[Feb 26 18:58] M1257787 interim PRP residue 59EFF1E40063D039 at iteration 35
[Feb 26 18:58] Iteration: 37 / 1257787 [0.00%], ms/iter:  3.848, ETA: 01:20:39
[Feb 26 18:58] Stopping PRP test of M1257787 at iteration 37 [0.00%]
[Feb 26 18:58] Worker stopped.
[Feb 26 18:58] Worker starting
[Feb 26 18:58] Resuming Gerbicz error-checking PRP test of M1257787 using Pentium4 FFT length 64K, Pass1=256, Pass2=256, clm=4
[Feb 26 18:58] Iteration: 38 / 1257787 [0.00%].
[Feb 26 18:58] M1257787 interim PRP residue FE5C44613154E578 at iteration 37
[Feb 26 18:58] Iteration: 39 / 1257787 [0.00%], ms/iter:  2.156, ETA: 00:45:11
[Feb 26 18:58] M1257787 interim PRP residue 794A82BDC4619CF4 at iteration 38
[Feb 26 18:58] Iteration: 40 / 1257787 [0.00%], ms/iter:  3.850, ETA: 01:20:42
[Feb 26 18:58] M1257787 interim PRP residue 61DA20FFA80D2AE1 at iteration 39

The application crash on launch documented in post 9 is reproduced in build 3 also.

[Prime95]

Originally Posted by harlee

L3 cache it is being reported as -1 KB.

Fixed in build 4 (if there is one).

[M0CZY]

32-bit Linux 29.6b3 detects my CPU cache a bit better than 29.6b1.

CPU Information:
Intel(R) Pentium(R) M processor 2.26GHz
CPU speed: 1063.79 MHz
CPU features: Prefetch, SSE, SSE2
L1 cache size: 32 KB, L2 cache size: 2 MB, L3 cache size: -1 KB

My CPU doesn't actually have an L3 cache, and it seems to report the CPU speed incorrectly!

[Mark Rose]

Originally Posted by M0CZY

32-bit Linux 29.6b3 detects my CPU cache a bit better than 29.6b1.

CPU Information:
Intel(R) Pentium(R) M processor 2.26GHz
CPU speed: 1063.79 MHz
CPU features: Prefetch, SSE, SSE2
L1 cache size: 32 KB, L2 cache size: 2 MB, L3 cache size: -1 KB

My CPU doesn't actually have an L3 cache, and it seems to report the CPU speed incorrectly!

It's not a 2.26 GHz processor? The speed is probably the actual clock it's running at, at the time.

[Flammo]

In build 3, I can't load the torture test at all. A dialog box says 'An unsupported operation was attempted'.

This is the same PC I previously reported torture test issues on in this thread with build 2.

If I load a new installation of Prime95, I click to say 'just stress testing', and then this same dialog box appears.

[harlee]

Running 29.6b3 on a MacBook Air. Selected Torture Test and then Prime95 closed without updating the save file. After the second attempt, a Problem Report for Prime95 window popped up. I'll forward the info to Prime95 via e-mail.

[harlee]

Originally Posted by harlee

Running 29.6b3 on a MacBook Air. Selected Torture Test and then Prime95 closed without updating the save file. After the second attempt, a Problem Report for Prime95 window popped up. I'll forward the info to Prime95 via e-mail.

Never mind - my bad!!! Thought I had upgraded to 29.6b3 but was still running b1. Testing Torture Test now but it is working properly so far.

[kriesel]

Originally Posted by Prime95

Build 3 is now available. This should fix the reported bugs. Please re-test (some of the reported bugs on older hardware I cannot test here).

Prime95 v29.6b3 x32 on Win XP and PentiumII-400, not primenet connected:

Comm window shows it taking exception to the same exponent I have currently running P-1 on a primenet connected system. Corresponding worktodo entry is

Test=N/A,90094027,75,1
(same form as for the77263057 entry that runs without error) Perhaps fft lengths required for p>77300000 are unsupported on ancient processors because of the huge ETA.

[Main thread Feb 27 17:51] Mersenne number primality test program version 29.6
[Main thread Feb 27 17:51] Error: Worktodo.txt file contained bad LL exponent: 90094027
[Main thread Feb 27 17:51] Illegal line in worktodo.txt file: Test=90094027,75,1
[Main thread Feb 27 17:51] Optimizing for CPU architecture: Pre-SSE2, L2 cache size: 512 KB, L3 cache size: -1 KB

Worker window

[Feb 27 17:53] Worker starting
[Feb 27 17:54] Starting primality test of M77263057 using x87 FFT length 4M, Pass1=1K, Pass2=4K, clm=1
[Feb 27 17:54] Iteration: 3 / 77263057 [0.00%], ms/iter: 4190.434, ETA: 3747d 06:56
[Feb 27 17:54] M77263057 interim LL residue 000000000000000E at iteration 3
[Feb 27 17:54] Iteration: 4 / 77263057 [0.00%], ms/iter: 3689.181, ETA: 3299d 01:02
[Feb 27 17:54] M77263057 interim LL residue 00000000000000C2 at iteration 4
[Feb 27 17:54] Iteration: 5 / 77263057 [0.00%], ms/iter: 3756.399, ETA: 3359d 03:41
[Feb 27 17:55] M77263057 interim LL residue 0000000000009302 at iteration 5
[Feb 27 17:55] Iteration: 6 / 77263057 [0.00%], ms/iter: 3739.550, ETA: 3344d 02:04
[Feb 27 17:55] M77263057 interim LL residue 00000000546B4C02 at iteration 6
[Feb 27 17:55] Iteration: 7 / 77263057 [0.00%], ms/iter: 3770.378, ETA: 3371d 15:41
[Feb 27 17:55] M77263057 interim LL residue 1BD696D9F03D3002 at iteration 7
[Feb 27 17:55] Iteration: 8 / 77263057 [0.00%], ms/iter: 3703.744, ETA: 3312d 01:36
[Feb 27 17:55] M77263057 interim LL residue 8CC88407A9F4C002 at iteration 8
[Feb 27 17:55] Iteration: 9 / 77263057 [0.00%], ms/iter: 3762.832, ETA: 3364d 21:45
[Feb 27 17:55] M77263057 interim LL residue 55599F9D37D30002 at iteration 9
[Feb 27 17:56] Iteration: 10 / 77263057 [0.00%], ms/iter: 3883.107, ETA: 3472d 11:04
[Feb 27 17:56] M77263057 interim LL residue F460D65DDF4C0002 at iteration 10

Taskmgr says prime95 is getting 95% of the cpu during these iterations.

Jacobi timings were interesting; restart at iteration 30 took 26 minutes, a far cry from 13 seconds at iteration 12. I switched from Interimresidues=1 to 10 at iteration 30. Iteration 30's interim residue was skipped, but 31 and 31 were output. Iteration times improved to 3.0 to 3.5 seconds/iter.

A try on a P3-450 running Win98 ran long enough to show some cache info before crashing rather immediately on launch. Since this is a quite old system, it could be a hardware reliability issue. Or maybe it's the old OS. It did not get as far as a P2-400 running XP (no LL iterations from M77263057 in the worktodo.txt). It only crashes if there's a worktodo file.

[Prime95]

Originally Posted by kriesel

Prime95 v29.6b3 x32 on Win XP and PentiumII-400, not primenet connected:

Comm window shows it taking exception to the same exponent I have currently running P-1 on a primenet connected system. Corresponding worktodo entry is

Test=N/A,90094027,75,1
(same form as for the77263057 entry that runs without error) Perhaps fft lengths required for p>77300000 are unsupported on ancient processors because of the huge ETA.

Yes, x87 FFTs only support exponents up to about 79.3 million.