Prime95 version 29.6/29.7/29.8

[harlee]

This morning 29.6b1 ran another benchmark. Notice the benchmark was for FFTlen=288K while the P1 was using a FFTlen=320K.

Stopped and restarted Prime95 29.6b1 and now it is reporting a FFTlen=288K.

[Feb 24 14:30] Worker starting
[Feb 24 14:30] Optimal P-1 factoring of M5339707 using up to 2048MB of memory.
[Feb 24 14:30] Assuming no factors below 2^62 and 10 primality tests saved if a factor is found.
[Feb 24 14:30] Optimal bounds are B1=320000, B2=9520000
[Feb 24 14:30] Chance of finding a factor is an estimated 7.47%
[Feb 24 14:30] Using Pentium4 FFT length 288K, Pass1=384, Pass2=768, clm=4

[harlee]

8)  Default memory available for prime95 changed from 8MB to 1/16th of RAM.

The 32-bit Windows 29.6b1 version is still defaulting to 8MB of memory. As you can see from my early posts that I've allocated 2GB of memory for Prime95 P-1 use.

[Feb 24 14:40] Worker starting
[Feb 24 14:40] Optimal P-1 factoring of M5339707 using up to 8MB of memory.
[Feb 24 14:40] Assuming no factors below 2^62 and 10 primality tests saved if a factor is found.
[Feb 24 14:40] Optimal bounds are B1=525000, B2=525000
[Feb 24 14:40] Chance of finding a factor is an estimated 4.1%
[Feb 24 14:40] Using Pentium4 type-0 FFT length 320K, Pass1=320, Pass2=1K, clm=4

Same thing on my MacBook Air with 8GB of memory:

[Feb 24 09:51] Optimal P-1 factoring of M18524839 using up to 8MB of memory.
[Feb 24 09:51] Assuming no factors below 2^65 and 3 primality tests saved if a factor is found.
[Feb 24 09:51] Optimal bounds are B1=480000, B2=480000
[Feb 24 09:51] Chance of finding a factor is an estimated 3.4%
[Feb 24 09:51] Using FMA3 FFT length 960K, Pass1=384, Pass2=2560, clm=2
[Feb 24 09:51] Worker stopped.
[Feb 24 09:52] Worker starting
[Feb 24 09:52] Optimal P-1 factoring of M18524839 using up to 4096MB of memory.
[Feb 24 09:52] Assuming no factors below 2^65 and 3 primality tests saved if a factor is found.
[Feb 24 09:52] Optimal bounds are B1=330000, B2=8745000
[Feb 24 09:52] Chance of finding a factor is an estimated 6.4%
[Feb 24 09:52] Using FMA3 FFT length 960K, Pass1=384, Pass2=2560, clm=2
[Feb 24 09:52] M18524839 stage 1 is 0.32% complete.

[Prime95]

Originally Posted by harlee

8)  Default memory available for prime95 changed from 8MB to 1/16th of RAM.

The 32-bit Windows 29.6b1 version is still defaulting to 8MB of memory.

That new default is for new installs on machines with more than 4GB of memory. If you are upgrading prime95 then the maximum amount of RAM is determined from the existing prime.txt.

Edit: My bad, the new default is only for the blend torture test. I'm changing whatsnew.txt to reflect this. The options/cpu memory-to-use-for-P-1/ECM remains 8MB by default.

[Prime95]

Originally Posted by harlee

Found an issue with the Windows 32-bit version. Noticed that 29.6b1 is running much slower then 29.4b7.

I think I'll need gwnum.txt, prime.txt, local.txt and your CPU info to investigate.

[harlee]

sent via e-mail

[Madpoo]

Originally Posted by harlee

Madpoo, Please look at M18521411 as 29.6b1 found and reported a a factor but the P1 bounds are not being shown.

[Sat Feb 23 14:42:47 2019]
{"status":"F", "exponent":18521411, "worktype":"P-1", "factors":"165718901219828172167153", "b1":330000, "b2":8745000, "brent-suyama":12, "fft-length":983040, "security-code":"DFB5B559", "program":{"name":"Prime95", "version":"29.6", "build":1, "port":10}, "timestamp":"2019-02-23 19:42:47", "user":"harlee", "computer":"i5-5250U_1600", "aid":"<deleted>"}

I don't remember ever showing the bounds for a "factor found" result in that history section. I could be wrong, but it's not ringing a bell. I'll double-check and if the old style non-JSON was doing it, I'll match that.

[Madpoo]

Originally Posted by Madpoo

I don't remember ever showing the bounds for a "factor found" result in that history section. I could be wrong, but it's not ringing a bell. I'll double-check and if the old style non-JSON was doing it, I'll match that.

I was wrong... I just hadn't been including that info since the switch to the JSON result format.

Fixed now.

[M0CZY]

I am testing 32-bit Linux 29.6b1, and I have noticed that the program is unable to detect my CPU cache sizes.

CPU Information:
Intel(R) Pentium(R) M processor 2.26GHz
CPU speed: 2260.72 MHz
CPU features: Prefetch, SSE, SSE2
L1 cache size: unknown, L2 cache size: unknown

I don't know if this is very important in the correct running of the program.

[Prime95]

Originally Posted by M0CZY

I am testing 32-bit Linux 29.6b1, and I have noticed that the program is unable to detect my CPU cache sizes.
I don't know if this is very important in the correct running of the program.

In 29.6 I switched to using hwloc to determine cache sizes. Apparently, that library has trouble with older architectures. I've fixed this in build 3 (prime95 falls back to old methods if hwloc cannot detect cache sizes).

[kriesel]

Originally Posted by M0CZY

I am testing 32-bit Linux 29.6b1, and I have noticed that the program is unable to detect my CPU cache sizes.

CPU Information:
Intel(R) Pentium(R) M processor 2.26GHz
CPU speed: 2260.72 MHz
CPU features: Prefetch, SSE, SSE2
L1 cache size: unknown, L2 cache size: unknown

I don't know if this is very important in the correct running of the program.

Similar occurs in Windows 29.6b1 x32 on a Pentium M. Not an issue in 29.4b7 on the same system as it occurs in 29.6b1. Win XP affected, as is Vista; possibly later OS versions also. Not an issue in 29.6b1 x64 on Win 10 i7-8750H, or 29.6b1 x64 on Win 7 i3-370M, or Vista X64 prime95 v29.6b1 X64 on Core 2 Duo.

[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).