Hi folks,

Yves has been busy and we have just pushed out a new release of genefer for testing. What's new? On the CPU, expect ~10% speedups for the x87 transform (current n<=20 leading edge tests), especially on 64-bit machines. The vector transforms (sse2, avx, fma3...) have also been optimised with a focus on the large (n>=21) tests. It's hard to generate code that is faster on *all* CPU types, so let us know how it goes for you. There are no speed improvements for the OCL code.

You can download the binaries from here:

https://app.assembla.com/spaces/genefer/subversion/source/HEAD/trunk/bin

Ignore the comments that say 3.3.2 (my bad) - all of the binaries are the new version 3.3.3.

We will do a full test campaign to verify everything works on a range of platforms, but for now it would be interesting to see benchmark results from anyone that wants to give it a try.

Cheers

- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime!

I took production primegrid_genefer_3_3_2_3.16_windows_x86_64__cpuGFN15.exe and mentioned 3.3.3 version of Genefer CPU.

Run specific x87 benchmark on 3 different CPUs, all Win10x64

CPU i5-3570K, version 3.3.2

CPU i5-3570K, version 3.3.3

CPU i5-6600, version 3.3.2

CPU i5-6600, version 3.3.3

CPU i7-8700K, version 3.3.2

CPU i7-8700K, version 3.3.3

Does anyone have a link to the previous version? I'd love to test, but I really have no idea on how to get the old release outside of requesting a WU from Boinc and retrieving the program manually.


Speaking which, it would be nice if "new version, test please" posts in general would include the links for both the old and the new versions. It would be pretty handy to have both in a single, convenient place.

Does anyone have a link to the previous version? I'd love to test, but I really have no idea on how to get the old release outside of requesting a WU from Boinc and retrieving the program manually.


Speaking which, it would be nice if "new version, test please" posts in general would include the links for both the old and the new versions. It would be pretty handy to have both in a single, convenient place.

Speaking which, it would be nice if "new version, test please" posts in general would include the links for both the old and the new versions. It would be pretty handy to have both in a single, convenient place.

Here are results from my X5675:
x87 3.3.2:

Here are some benchmarks for the i5 6600k, overclocked to 4100mhz and with some high speed RAM. Generally speaking, the new version is good if you're sure to use x87, but bad otherwise.


FMA3: both seem about the same, maybe with some small margin in favor of the old version

Ver. 3.3.2

4019150^1024+1 Time: 2.01 us/mul. Err: 0.1602 6763 digits 3287270^2048+1 Time: 3.97 us/mul. Err: 0.1641 13347 digits 2688666^4096+1 Time: 8.29 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 17.5 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 36.6 us/mul. Err: 0.1562 102481 digits 1471094^32768+1 Time: 78.8 us/mul. Err: 0.1562 202102 digits 1203210^65536+1 Time: 166 us/mul. Err: 0.1504 398482 digits 984108^131072+1 Time: 363 us/mul. Err: 0.1562 785521 digits 804904^262144+1 Time: 753 us/mul. Err: 0.1523 1548156 digits 658332^524288+1 Time: 1.7 ms/mul. Err: 0.1562 3050541 digits 538452^1048576+1 Time: 3.64 ms/mul. Err: 0.1406 6009544 digits 440400^2097152+1 Time: 8.35 ms/mul. Err: 0.1406 11836006 digits 360204^4194304+1 Time: 17.3 ms/mul. Err: 0.1328 23305854 digits 294612^8388608+1 Time: 40.3 ms/mul. Err: 0.1367 45879398 digits Genefer Mark = 24.

Ver. 3.3.3

4019150^1024+1 Time: 1.97 us/mul. Err: 0.1562 6763 digits 3287270^2048+1 Time: 4.12 us/mul. Err: 0.1562 13347 digits 2688666^4096+1 Time: 8.64 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 18.5 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 38.5 us/mul. Err: 0.1406 102481 digits 1471094^32768+1 Time: 81.8 us/mul. Err: 0.1406 202102 digits 1203210^65536+1 Time: 172 us/mul. Err: 0.1289 398482 digits 984108^131072+1 Time: 366 us/mul. Err: 0.1406 785521 digits 804904^262144+1 Time: 779 us/mul. Err: 0.1250 1548156 digits 658332^524288+1 Time: 1.68 ms/mul. Err: 0.1250 3050541 digits 538452^1048576+1 Time: 3.76 ms/mul. Err: 0.1172 6009544 digits 440400^2097152+1 Time: 8.27 ms/mul. Err: 0.1172 11836006 digits 360204^4194304+1 Time: 18.1 ms/mul. Err: 0.1094 23305854 digits 294612^8388608+1 Time: 39.9 ms/mul. Err: 0.1104 45879398 digits 240964^16777216+1 Time: 86.9 ms/mul. Err: 0.1016 90294174 digits Genefer Mark = 24.

Ver. 3.3.2

4019150^1024+1 Time: 2.34 us/mul. Err: 0.1562 6763 digits 3287270^2048+1 Time: 5.32 us/mul. Err: 0.1562 13347 digits 2688666^4096+1 Time: 9.82 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 23.2 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 43.3 us/mul. Err: 0.1562 102481 digits 1471094^32768+1 Time: 102 us/mul. Err: 0.1562 202102 digits 1203210^65536+1 Time: 194 us/mul. Err: 0.1406 398482 digits 984108^131072+1 Time: 462 us/mul. Err: 0.1484 785521 digits 804904^262144+1 Time: 876 us/mul. Err: 0.1406 1548156 digits 658332^524288+1 Time: 2.1 ms/mul. Err: 0.1445 3050541 digits 538452^1048576+1 Time: 4.13 ms/mul. Err: 0.1367 6009544 digits 440400^2097152+1 Time: 9.97 ms/mul. Err: 0.1406 11836006 digits 360204^4194304+1 Time: 19.3 ms/mul. Err: 0.1484 23305854 digits 294612^8388608+1 Time: 46.8 ms/mul. Err: 0.1250 45879398 digits 240964^16777216+1 Time: 90.7 ms/mul. Err: 0.1250 90294174 digits Genefer Mark = 21.

Ver. 3.3.3

4019150^1024+1 Time: 2.2 us/mul. Err: 0.1406 6763 digits 3287270^2048+1 Time: 4.65 us/mul. Err: 0.1562 13347 digits 2688666^4096+1 Time: 9.86 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 20.9 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 43.9 us/mul. Err: 0.1562 102481 digits 1471094^32768+1 Time: 93 us/mul. Err: 0.1562 202102 digits 1203210^65536+1 Time: 197 us/mul. Err: 0.1328 398482 digits 984108^131072+1 Time: 414 us/mul. Err: 0.1289 785521 digits 804904^262144+1 Time: 878 us/mul. Err: 0.1406 1548156 digits 658332^524288+1 Time: 1.88 ms/mul. Err: 0.1250 3050541 digits 538452^1048576+1 Time: 4.17 ms/mul. Err: 0.1250 6009544 digits 440400^2097152+1 Time: 9.1 ms/mul. Err: 0.1250 11836006 digits 360204^4194304+1 Time: 19.7 ms/mul. Err: 0.1094 23305854 digits 294612^8388608+1 Time: 43.1 ms/mul. Err: 0.1250 45879398 digits 240964^16777216+1 Time: 93.5 ms/mul. Err: 0.1094 90294174 digits Genefer Mark = 22.

Ver. 3.3.2

4019150^1024+1 Time: 4.35 us/mul. Err: 0.1562 6763 digits 3287270^2048+1 Time: 8.93 us/mul. Err: 0.1875 13347 digits 2688666^4096+1 Time: 19.1 us/mul. Err: 0.1875 26336 digits 2199064^8192+1 Time: 39.3 us/mul. Err: 0.1875 51956 digits 1798620^16384+1 Time: 85.4 us/mul. Err: 0.1719 102481 digits 1471094^32768+1 Time: 175 us/mul. Err: 0.1797 202102 digits 1203210^65536+1 Time: 378 us/mul. Err: 0.1719 398482 digits 984108^131072+1 Time: 774 us/mul. Err: 0.1562 785521 digits 804904^262144+1 Time: 1.69 ms/mul. Err: 0.1562 1548156 digits 658332^524288+1 Time: 3.47 ms/mul. Err: 0.1562 3050541 digits 538452^1048576+1 Time: 7.61 ms/mul. Err: 0.1484 6009544 digits 440400^2097152+1 Time: 15.6 ms/mul. Err: 0.1562 11836006 digits 360204^4194304+1 Time: 34.3 ms/mul. Err: 0.1406 23305854 digits 294612^8388608+1 Time: 69.6 ms/mul. Err: 0.1406 45879398 digits 240964^16777216+1 Time: 154 ms/mul. Err: 0.1250 90294174 digits Genefer Mark = 13.

Ver. 3.3.3

4019150^1024+1 Time: 5.9 us/mul. Err: 0.1406 6763 digits 3287270^2048+1 Time: 12.4 us/mul. Err: 0.1562 13347 digits 2688666^4096+1 Time: 25.9 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 60.2 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 125 us/mul. Err: 0.1562 102481 digits 1471094^32768+1 Time: 287 us/mul. Err: 0.1562 202102 digits 1203210^65536+1 Time: 598 us/mul. Err: 0.1328 398482 digits 984108^131072+1 Time: 1.34 ms/mul. Err: 0.1289 785521 digits 804904^262144+1 Time: 2.79 ms/mul. Err: 0.1406 1548156 digits 658332^524288+1 Time: 6.18 ms/mul. Err: 0.1250 3050541 digits 538452^1048576+1 Time: 12.8 ms/mul. Err: 0.1250 6009544 digits 440400^2097152+1 Time: 28.1 ms/mul. Err: 0.1250 11836006 digits 360204^4194304+1 Time: 57.7 ms/mul. Err: 0.1094 23305854 digits 294612^8388608+1 Time: 125 ms/mul. Err: 0.1250 45879398 digits 240964^16777216+1 Time: 257 ms/mul. Err: 0.1094 90294174 digits Genefer Mark = 7.

Ver 3.3.2

4019150^1024+1 Time: 5.26 us/mul. Err: 0.1562 6763 digits 3287270^2048+1 Time: 11.4 us/mul. Err: 0.1875 13347 digits 2688666^4096+1 Time: 22.3 us/mul. Err: 0.1875 26336 digits 2199064^8192+1 Time: 48.3 us/mul. Err: 0.1875 51956 digits 1798620^16384+1 Time: 97.3 us/mul. Err: 0.1719 102481 digits 1471094^32768+1 Time: 210 us/mul. Err: 0.1797 202102 digits 1203210^65536+1 Time: 425 us/mul. Err: 0.1719 398482 digits 984108^131072+1 Time: 912 us/mul. Err: 0.1562 785521 digits 804904^262144+1 Time: 1.88 ms/mul. Err: 0.1562 1548156 digits 658332^524288+1 Time: 4.01 ms/mul. Err: 0.1562 3050541 digits 538452^1048576+1 Time: 8.36 ms/mul. Err: 0.1484 6009544 digits 440400^2097152+1 Time: 17.8 ms/mul. Err: 0.1562 11836006 digits 360204^4194304+1 Time: 37.4 ms/mul. Err: 0.1406 23305854 digits 294612^8388608+1 Time: 78.5 ms/mul. Err: 0.1406 45879398 digits Genefer Mark = 11.

Ver. 3.3.3

4019150^1024+1 Time: 8.19 us/mul. Err: 0.1406 6763 digits 3287270^2048+1 Time: 17.1 us/mul. Err: 0.1562 13347 digits 2688666^4096+1 Time: 35.4 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 78.9 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 164 us/mul. Err: 0.1562 102481 digits 1471094^32768+1 Time: 363 us/mul. Err: 0.1562 202102 digits 1203210^65536+1 Time: 752 us/mul. Err: 0.1328 398482 digits 984108^131072+1 Time: 1.64 ms/mul. Err: 0.1289 785521 digits 804904^262144+1 Time: 3.4 ms/mul. Err: 0.1406 1548156 digits 658332^524288+1 Time: 7.38 ms/mul. Err: 0.1250 3050541 digits 538452^1048576+1 Time: 15.2 ms/mul. Err: 0.1250 6009544 digits 440400^2097152+1 Time: 32.8 ms/mul. Err: 0.1250 11836006 digits 360204^4194304+1 Time: 67.4 ms/mul. Err: 0.1094 23305854 digits 294612^8388608+1 Time: 144 ms/mul. Err: 0.1250 45879398 digits 240964^16777216+1 Time: 295 ms/mul. Err: 0.1094 90294174 digits Genefer Mark = 6.

Ver. 3.3.2

10980642^32+1 Time: 434 ns/mul. Err: 0.0001 226 digits 8981090^64+1 Time: 960 ns/mul. Err: 0.0001 446 digits 7345652^128+1 Time: 2.15 us/mul. Err: 0.0001 879 digits 6008024^256+1 Time: 4.86 us/mul. Err: 0.0001 1736 digits 4913974^512+1 Time: 10.8 us/mul. Err: 0.0001 3427 digits 4019150^1024+1 Time: 23.4 us/mul. Err: 0.0001 6763 digits 3287270^2048+1 Time: 51.3 us/mul. Err: 0.0001 13347 digits 2688666^4096+1 Time: 112 us/mul. Err: 0.0001 26336 digits 2199064^8192+1 Time: 241 us/mul. Err: 0.0001 51956 digits 1798620^16384+1 Time: 519 us/mul. Err: 0.0001 102481 digits 1471094^32768+1 Time: 1.11 ms/mul. Err: 0.0001 202102 digits 1203210^65536+1 Time: 2.37 ms/mul. Err: 0.0001 398482 digits 984108^131072+1 Time: 5.03 ms/mul. Err: 0.0001 785521 digits 804904^262144+1 Time: 10.7 ms/mul. Err: 0.0001 1548156 digits 658332^524288+1 Time: 22.5 ms/mul. Err: 0.0001 3050541 digits 538452^1048576+1 Time: 47.6 ms/mul. Err: 0.0001 6009544 digits 440400^2097152+1 Time: 99.5 ms/mul. Err: 0.0001 11836006 digits 360204^4194304+1 Time: 209 ms/mul. Err: 0.0001 23305854 digits 294612^8388608+1 Time: 436 ms/mul. Err: 0.0001 45879398 digits 240964^16777216+1 Time: 912 ms/mul. Err: 0.0001 90294174 digits Genefer Mark = 2.

Ver. 3.3.3

10980642^32+1 Time: 371 ns/mul. Err: 0.0001 226 digits 8981090^64+1 Time: 835 ns/mul. Err: 0.0001 446 digits 7345652^128+1 Time: 1.92 us/mul. Err: 0.0001 879 digits 6008024^256+1 Time: 4.33 us/mul. Err: 0.0001 1736 digits 4913974^512+1 Time: 9.73 us/mul. Err: 0.0001 3427 digits 4019150^1024+1 Time: 21.6 us/mul. Err: 0.0001 6763 digits 3287270^2048+1 Time: 47.3 us/mul. Err: 0.0001 13347 digits 2688666^4096+1 Time: 103 us/mul. Err: 0.0001 26336 digits 2199064^8192+1 Time: 223 us/mul. Err: 0.0001 51956 digits 1798620^16384+1 Time: 480 us/mul. Err: 0.0001 102481 digits 1471094^32768+1 Time: 1.03 ms/mul. Err: 0.0001 202102 digits 1203210^65536+1 Time: 2.19 ms/mul. Err: 0.0001 398482 digits 984108^131072+1 Time: 4.67 ms/mul. Err: 0.0001 785521 digits 804904^262144+1 Time: 9.94 ms/mul. Err: 0.0001 1548156 digits 658332^524288+1 Time: 21 ms/mul. Err: 0.0001 3050541 digits 538452^1048576+1 Time: 44.4 ms/mul. Err: 0.0001 6009544 digits 440400^2097152+1 Time: 93.2 ms/mul. Err: 0.0001 11836006 digits 360204^4194304+1 Time: 196 ms/mul. Err: 0.0001 23305854 digits 294612^8388608+1 Time: 409 ms/mul. Err: 0.0001 45879398 digits 240964^16777216+1 Time: 854 ms/mul. Err: 0.0001 90294174 digits Genefer Mark = 2.

i7-8700K, a bit of slowdown for FMA3 in 3.3.3.

CPU i7-8700K, version 3.3.2 FMA3

CPU i7-8700K, version 3.3.3 FMA3

CPU i7-8700K, version 3.3.2 AVX

CPU i7-8700K, version 3.3.3 FMA3

"Generalized Fermat Prime Search benchmarks" are the most significant.

GFN15-GFN19 => x87.
x87 was written in c++ in ver 3.3.2 and is optimized in asm in 3.3.3: then it should be faster on every CPU.

GFN20 => x87 with 3.3.2. The error of 3.3.3 (fma, avx, ...) is lower then this version may still to be able to test some n=20 with fast transforms... but we are very close to the limit at b ~ 950000.

GFN21 is then only n that uses fma, avx, sse4, or sse2 transform. On a fixed processor, only the faster transform is relevant. fma (i3/i5/i7 4th-8th gen) is expected to run at about the same speed, avx (2th-3th gen) to be faster. I didn't check 1th gen (sse4), it seems to be a major slowdown.

I purchased a 7th gen i5 with a particular focus on fma transform on x-Lake processors, but it will take time before this version is available.

In production, ver 3.3.3 is an improvement for n <= 20.
One may wonder whether 3.3.2 or 3.3.3 is the most appropriate for n=21.

"Generalized Fermat Prime Search benchmarks" are the most significant.

On my Haswell (i7-4750HQ):

New round of benches, using -b2 and with 2 more CPUs. TLDR: FMA3 seems inconsistent (I re-ran the tests manually and sometimes there was improvement, at others slowdown, and even about the same performance), x87 is always better, SSE2 is MUCH slower.

- i5 6600k -
Ver. 3.3.2

I'm going to test the new Genefer on 5 different CPUs in a moment, but first this:

On second thought, I just noticed that the -b2 benchmarks were modified to test different values in 3.3.3 -- so comparing these benchmarks in the old and new versions isn't a good comparison.
____________
My lucky number is 75898⁵²⁴²⁸⁸+1

The new Genefer is missing the old f64 transform. Does this mean the Genefer will no longer run on old CPUs like my recently defunct Sempron?

SSE2 is MUCH slower.

The new Genefer is missing the old f64 transform. Does this mean the Genefer will no longer run on old CPUs like my recently defunct Sempron?

Genefer 3.3.3 64-bit => fma3 avx sse4 sse2 x87
Genefer 3.3.3 32-bit => fma3 avx sse4 sse2 x87 f64

K7 Sempron is a 32-bit processor without SSE2 but K8 Sempron is a 64-bit processor with SSE2.

f64 was removed from the 64-bit version because the SSE2 is included in AMD64 specification.

i7-5820K (Haswell-E) @ 3.4 GHz (slight o/c, nominal is 3.3 GHz)

Updated binaries (version 3.3.3-1) are now available from https://app.assembla.com/spaces/genefer/subversion/source/HEAD/trunk/bin, which address the performance issues of the sse2/sse4 transforms.

- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime!

Testing the new 3.3.3-1 SSE4 and SSE2 (TL:DR better than previous 3.3.3 release, but still not worth it over 3.3.2).

First, the i5 6600k Skylake. I had to go with the -b option, given that -b2 doesn't use the SSE transforms. Both SSE4 and SSE2 saw major improvements, but are still slightly slower than 3.3.2.

For the i5 4590 Haswell, comparing 3.3.3-1 vs 3.3.2, performance is just about the same for SSE4, with a marginal tip towards the slower side. SSE2 is slower; not by that much, but you can definitely notice it. Regardless, there's no reason to use the new version, other than the fact that lower Err. values could allow a faster transform to be used during transition periods.

I won't be able to test my LGA775 pentium next year, but to make up for it, I'm adding benchmarks for an i3 530, which is SSE4. A full suit of benchmarks is available below, with both -b2 and -b options tested. Across the board, 3.3.3-1 is noticeably slower than 3.3.2



- i5 6600k -

SSE4: slight slowdowns (12 vs 13)

Ver. 3.3.2

4019150^1024+1 Time: 4.35 us/mul. Err: 0.1562 6763 digits 3287270^2048+1 Time: 8.93 us/mul. Err: 0.1875 13347 digits 2688666^4096+1 Time: 19.1 us/mul. Err: 0.1875 26336 digits 2199064^8192+1 Time: 39.3 us/mul. Err: 0.1875 51956 digits 1798620^16384+1 Time: 85.4 us/mul. Err: 0.1719 102481 digits 1471094^32768+1 Time: 175 us/mul. Err: 0.1797 202102 digits 1203210^65536+1 Time: 378 us/mul. Err: 0.1719 398482 digits 984108^131072+1 Time: 774 us/mul. Err: 0.1562 785521 digits 804904^262144+1 Time: 1.69 ms/mul. Err: 0.1562 1548156 digits 658332^524288+1 Time: 3.47 ms/mul. Err: 0.1562 3050541 digits 538452^1048576+1 Time: 7.61 ms/mul. Err: 0.1484 6009544 digits 440400^2097152+1 Time: 15.6 ms/mul. Err: 0.1562 11836006 digits 360204^4194304+1 Time: 34.3 ms/mul. Err: 0.1406 23305854 digits 294612^8388608+1 Time: 69.6 ms/mul. Err: 0.1406 45879398 digits 240964^16777216+1 Time: 154 ms/mul. Err: 0.1250 90294174 digits Genefer Mark = 13.

Ver. 3.3.3-1

4019150^1024+1 Time: 4.84 us/mul. Err: 0.1406 6763 digits 3287270^2048+1 Time: 9.61 us/mul. Err: 0.1562 13347 digits 2688666^4096+1 Time: 20.3 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 42.4 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 89.8 us/mul. Err: 0.1562 102481 digits 1471094^32768+1 Time: 189 us/mul. Err: 0.1562 202102 digits 1203210^65536+1 Time: 400 us/mul. Err: 0.1328 398482 digits 984108^131072+1 Time: 837 us/mul. Err: 0.1289 785521 digits 804904^262144+1 Time: 1.77 ms/mul. Err: 0.1406 1548156 digits 658332^524288+1 Time: 3.7 ms/mul. Err: 0.1250 3050541 digits 538452^1048576+1 Time: 7.86 ms/mul. Err: 0.1250 6009544 digits 440400^2097152+1 Time: 16.5 ms/mul. Err: 0.1250 11836006 digits 360204^4194304+1 Time: 34.9 ms/mul. Err: 0.1094 23305854 digits 294612^8388608+1 Time: 73.4 ms/mul. Err: 0.1250 45879398 digits 240964^16777216+1 Time: 155 ms/mul. Err: 0.1094 90294174 digits Genefer Mark = 12.

SSE2: slight slowdowns (10 vs 11)

Ver 3.3.2

4019150^1024+1 Time: 5.26 us/mul. Err: 0.1562 6763 digits 3287270^2048+1 Time: 11.4 us/mul. Err: 0.1875 13347 digits 2688666^4096+1 Time: 22.3 us/mul. Err: 0.1875 26336 digits 2199064^8192+1 Time: 48.3 us/mul. Err: 0.1875 51956 digits 1798620^16384+1 Time: 97.3 us/mul. Err: 0.1719 102481 digits 1471094^32768+1 Time: 210 us/mul. Err: 0.1797 202102 digits 1203210^65536+1 Time: 425 us/mul. Err: 0.1719 398482 digits 984108^131072+1 Time: 912 us/mul. Err: 0.1562 785521 digits 804904^262144+1 Time: 1.88 ms/mul. Err: 0.1562 1548156 digits 658332^524288+1 Time: 4.01 ms/mul. Err: 0.1562 3050541 digits 538452^1048576+1 Time: 8.36 ms/mul. Err: 0.1484 6009544 digits 440400^2097152+1 Time: 17.8 ms/mul. Err: 0.1562 11836006 digits 360204^4194304+1 Time: 37.4 ms/mul. Err: 0.1406 23305854 digits 294612^8388608+1 Time: 78.5 ms/mul. Err: 0.1406 45879398 digits Genefer Mark = 11.

Ver. 3.3.3

4019150^1024+1 Time: 6.55 us/mul. Err: 0.1406 6763 digits 3287270^2048+1 Time: 12.6 us/mul. Err: 0.1562 13347 digits 2688666^4096+1 Time: 26.3 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 54 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 113 us/mul. Err: 0.1562 102481 digits 1471094^32768+1 Time: 234 us/mul. Err: 0.1562 202102 digits 1203210^65536+1 Time: 493 us/mul. Err: 0.1328 398482 digits 984108^131072+1 Time: 1.02 ms/mul. Err: 0.1289 785521 digits 804904^262144+1 Time: 2.16 ms/mul. Err: 0.1406 1548156 digits 658332^524288+1 Time: 4.44 ms/mul. Err: 0.1250 3050541 digits 538452^1048576+1 Time: 9.42 ms/mul. Err: 0.1250 6009544 digits 440400^2097152+1 Time: 19.6 ms/mul. Err: 0.1250 11836006 digits 360204^4194304+1 Time: 41.1 ms/mul. Err: 0.1094 23305854 digits 294612^8388608+1 Time: 85.1 ms/mul. Err: 0.1250 45879398 digits 240964^16777216+1 Time: 180 ms/mul. Err: 0.1094 90294174 digits Genefer Mark = 10.

SSE4: just about the same thing

Ver 3.3.2

4019150^1024+1 Time: 5.66 us/mul. Err: 0.1562 6763 digits 3287270^2048+1 Time: 11.9 us/mul. Err: 0.1875 13347 digits 2688666^4096+1 Time: 25.4 us/mul. Err: 0.1875 26336 digits 2199064^8192+1 Time: 53.5 us/mul. Err: 0.1875 51956 digits 1798620^16384+1 Time: 114 us/mul. Err: 0.1719 102481 digits 1471094^32768+1 Time: 240 us/mul. Err: 0.1797 202102 digits 1203210^65536+1 Time: 511 us/mul. Err: 0.1719 398482 digits 984108^131072+1 Time: 1.06 ms/mul. Err: 0.1562 785521 digits 804904^262144+1 Time: 2.27 ms/mul. Err: 0.1562 1548156 digits 658332^524288+1 Time: 4.86 ms/mul. Err: 0.1562 3050541 digits 538452^1048576+1 Time: 10.4 ms/mul. Err: 0.1484 6009544 digits 440400^2097152+1 Time: 21.4 ms/mul. Err: 0.1562 11836006 digits 360204^4194304+1 Time: 47.7 ms/mul. Err: 0.1406 23305854 digits 294612^8388608+1 Time: 103 ms/mul. Err: 0.1406 45879398 digits 240964^16777216+1 Time: 219 ms/mul. Err: 0.1250 90294174 digits Genefer Mark = 9.

Ver. 3.3.3-1

4019150^1024+1 Time: 6.58 us/mul. Err: 0.1406 6763 digits 3287270^2048+1 Time: 12.2 us/mul. Err: 0.1562 13347 digits 2688666^4096+1 Time: 26.4 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 55.3 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 119 us/mul. Err: 0.1562 102481 digits 1471094^32768+1 Time: 250 us/mul. Err: 0.1562 202102 digits 1203210^65536+1 Time: 535 us/mul. Err: 0.1328 398482 digits 984108^131072+1 Time: 1.11 ms/mul. Err: 0.1289 785521 digits 804904^262144+1 Time: 2.37 ms/mul. Err: 0.1406 1548156 digits 658332^524288+1 Time: 4.94 ms/mul. Err: 0.1250 3050541 digits 538452^1048576+1 Time: 10.7 ms/mul. Err: 0.1250 6009544 digits 440400^2097152+1 Time: 22.3 ms/mul. Err: 0.1250 11836006 digits 360204^4194304+1 Time: 48.4 ms/mul. Err: 0.1094 23305854 digits 294612^8388608+1 Time: 102 ms/mul. Err: 0.1250 45879398 digits 240964^16777216+1 Time: 220 ms/mul. Err: 0.1094 90294174 digits Genefer Mark = 9.

SSE2: slight slowdowns

Ver 3.3.2

4019150^1024+1 Time: 7.16 us/mul. Err: 0.1562 6763 digits 3287270^2048+1 Time: 14.3 us/mul. Err: 0.1875 13347 digits 2688666^4096+1 Time: 31.4 us/mul. Err: 0.1875 26336 digits 2199064^8192+1 Time: 62.5 us/mul. Err: 0.1875 51956 digits 1798620^16384+1 Time: 138 us/mul. Err: 0.1719 102481 digits 1471094^32768+1 Time: 275 us/mul. Err: 0.1797 202102 digits 1203210^65536+1 Time: 604 us/mul. Err: 0.1719 398482 digits 984108^131072+1 Time: 1.21 ms/mul. Err: 0.1562 785521 digits 804904^262144+1 Time: 2.64 ms/mul. Err: 0.1562 1548156 digits 658332^524288+1 Time: 5.35 ms/mul. Err: 0.1562 3050541 digits 538452^1048576+1 Time: 11.9 ms/mul. Err: 0.1484 6009544 digits 440400^2097152+1 Time: 23.8 ms/mul. Err: 0.1562 11836006 digits 360204^4194304+1 Time: 53.7 ms/mul. Err: 0.1406 23305854 digits 294612^8388608+1 Time: 107 ms/mul. Err: 0.1406 45879398 digits 240964^16777216+1 Time: 243 ms/mul. Err: 0.1250 90294174 digits Genefer Mark = 8.

Ver 3.3.3-1

4019150^1024+1 Time: 8.18 us/mul. Err: 0.1406 6763 digits 3287270^2048+1 Time: 16.5 us/mul. Err: 0.1562 13347 digits 2688666^4096+1 Time: 33.9 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 70.3 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 149 us/mul. Err: 0.1562 102481 digits 1471094^32768+1 Time: 309 us/mul. Err: 0.1562 202102 digits 1203210^65536+1 Time: 652 us/mul. Err: 0.1328 398482 digits 984108^131072+1 Time: 1.35 ms/mul. Err: 0.1289 785521 digits 804904^262144+1 Time: 2.84 ms/mul. Err: 0.1406 1548156 digits 658332^524288+1 Time: 5.88 ms/mul. Err: 0.1250 3050541 digits 538452^1048576+1 Time: 12.5 ms/mul. Err: 0.1250 6009544 digits 440400^2097152+1 Time: 26.2 ms/mul. Err: 0.1250 11836006 digits 360204^4194304+1 Time: 56 ms/mul. Err: 0.1094 23305854 digits 294612^8388608+1 Time: 118 ms/mul. Err: 0.1250 45879398 digits 240964^16777216+1 Time: 251 ms/mul. Err: 0.1094 90294174 digits Genefer Mark = 8.

-b2 benchmarks

Ver. 3.3.2

50000000^32768+1 252280 digits x87 (80-bit) Estimated time: 0:25:40 25000000^65536+1 484832 digits x87 (80-bit) Estimated time: 1:47:00 8000000^131072+1 904802 digits x87 (80-bit) Estimated time: 6:37:00 46000000^131072+1 1004373 digits x87 (80-bit) Estimated time: 7:18:00 2800000^262144+1 1690084 digits x87 (80-bit) Estimated time: 26:10:00 1500000^524288+1 3238051 digits x87 (80-bit) Estimated time: 109:00:00 900000^1048576+1 6243476 digits SSE4 Estimated time: 85:00:00 130000^2097152+1 10724717 digits SSE4 Estimated time: 298:00:00 100000^4194304+1 20971521 digits SSE4 Estimated time: 1332:00:00

Ver. 3.3.3-1

75000000^32768+1 258051 digits x87 (80-bit) Estimated time: 0:24:10 27000000^65536+1 487022 digits x87 (80-bit) Estimated time: 1:38:00 10000000^131072+1 917505 digits x87 (80-bit) Estimated time: 6:19:00 48000000^131072+1 1006796 digits x87 (80-bit) Estimated time: 6:52:00 3600000^262144+1 1718696 digits x87 (80-bit) Estimated time: 25:00:00 1700000^524288+1 3266550 digits x87 (80-bit) Estimated time: 100:00:00 950000^1048576+1 6268098 digits SSE4 Estimated time: 94:00:00 180000^2097152+1 11021106 digits SSE4 Estimated time: 352:00:00 110000^4194304+1 21145134 digits SSE4 Estimated time: 1461:00:00

SSE4: noticeable slowdowns

Ver. 3.3.2

4019150^1024+1 Time: 7.52 us/mul. Err: 0.1562 6763 digits 3287270^2048+1 Time: 15.7 us/mul. Err: 0.1875 13347 digits 2688666^4096+1 Time: 34.9 us/mul. Err: 0.1875 26336 digits 2199064^8192+1 Time: 71.1 us/mul. Err: 0.1875 51956 digits 1798620^16384+1 Time: 155 us/mul. Err: 0.1719 102481 digits 1471094^32768+1 Time: 324 us/mul. Err: 0.1797 202102 digits 1203210^65536+1 Time: 700 us/mul. Err: 0.1719 398482 digits 984108^131072+1 Time: 1.44 ms/mul. Err: 0.1562 785521 digits 804904^262144+1 Time: 3.14 ms/mul. Err: 0.1562 1548156 digits 658332^524288+1 Time: 6.61 ms/mul. Err: 0.1562 3050541 digits 538452^1048576+1 Time: 14.7 ms/mul. Err: 0.1484 6009544 digits 440400^2097152+1 Time: 30.1 ms/mul. Err: 0.1562 11836006 digits 360204^4194304+1 Time: 69 ms/mul. Err: 0.1406 23305854 digits 294612^8388608+1 Time: 140 ms/mul. Err: 0.1406 45879398 digits 240964^16777216+1 Time: 318 ms/mul. Err: 0.1250 90294174 digits Genefer Mark = 6.

Ver. 3.3.3-1

4019150^1024+1 Time: 8.14 us/mul. Err: 0.1406 6763 digits 3287270^2048+1 Time: 17.3 us/mul. Err: 0.1562 13347 digits 2688666^4096+1 Time: 37.6 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 80.9 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 175 us/mul. Err: 0.1562 102481 digits 1471094^32768+1 Time: 371 us/mul. Err: 0.1562 202102 digits 1203210^65536+1 Time: 796 us/mul. Err: 0.1328 398482 digits 984108^131072+1 Time: 1.67 ms/mul. Err: 0.1289 785521 digits 804904^262144+1 Time: 3.56 ms/mul. Err: 0.1406 1548156 digits 658332^524288+1 Time: 7.65 ms/mul. Err: 0.1250 3050541 digits 538452^1048576+1 Time: 16.3 ms/mul. Err: 0.1250 6009544 digits 440400^2097152+1 Time: 34.7 ms/mul. Err: 0.1250 11836006 digits 360204^4194304+1 Time: 75 ms/mul. Err: 0.1094 23305854 digits 294612^8388608+1 Time: 160 ms/mul. Err: 0.1250 45879398 digits 240964^16777216+1 Time: 343 ms/mul. Err: 0.1094 90294174 digits Genefer Mark = 6.

SSE2: noticeable slowdowns

Ver. 3.3.2

4019150^1024+1 Time: 11.8 us/mul. Err: 0.1562 6763 digits 3287270^2048+1 Time: 22.8 us/mul. Err: 0.1875 13347 digits 2688666^4096+1 Time: 51.3 us/mul. Err: 0.1875 26336 digits 2199064^8192+1 Time: 99.2 us/mul. Err: 0.1875 51956 digits 1798620^16384+1 Time: 224 us/mul. Err: 0.1719 102481 digits 1471094^32768+1 Time: 435 us/mul. Err: 0.1797 202102 digits 1203210^65536+1 Time: 969 us/mul. Err: 0.1719 398482 digits 984108^131072+1 Time: 1.89 ms/mul. Err: 0.1562 785521 digits 804904^262144+1 Time: 4.22 ms/mul. Err: 0.1562 1548156 digits 658332^524288+1 Time: 8.41 ms/mul. Err: 0.1562 3050541 digits 538452^1048576+1 Time: 19.1 ms/mul. Err: 0.1484 6009544 digits 440400^2097152+1 Time: 37.3 ms/mul. Err: 0.1562 11836006 digits 360204^4194304+1 Time: 86.5 ms/mul. Err: 0.1406 23305854 digits 294612^8388608+1 Time: 168 ms/mul. Err: 0.1406 45879398 digits 240964^16777216+1 Time: 386 ms/mul. Err: 0.1250 90294174 digits Genefer Mark = 5.

Ver. 3.3.3-1

4019150^1024+1 Time: 13 us/mul. Err: 0.1406 6763 digits 3287270^2048+1 Time: 26.3 us/mul. Err: 0.1562 13347 digits 2688666^4096+1 Time: 55.1 us/mul. Err: 0.1562 26336 digits 2199064^8192+1 Time: 116 us/mul. Err: 0.1562 51956 digits 1798620^16384+1 Time: 243 us/mul. Err: 0.1562 102481 digits 1471094^32768+1 Time: 510 us/mul. Err: 0.1562 202102 digits 1203210^65536+1 Time: 1.07 ms/mul. Err: 0.1328 398482 digits 984108^131072+1 Time: 2.22 ms/mul. Err: 0.1289 785521 digits 804904^262144+1 Time: 4.67 ms/mul. Err: 0.1406 1548156 digits 658332^524288+1 Time: 9.75 ms/mul. Err: 0.1250 3050541 digits 538452^1048576+1 Time: 20.6 ms/mul. Err: 0.1250 6009544 digits 440400^2097152+1 Time: 43.4 ms/mul. Err: 0.1250 11836006 digits 360204^4194304+1 Time: 92.4 ms/mul. Err: 0.1094 23305854 digits 294612^8388608+1 Time: 196 ms/mul. Err: 0.1250 45879398 digits 240964^16777216+1 Time: 412 ms/mul. Err: 0.1094 90294174 digits Genefer Mark = 5.

First, the i5 6600k Skylake. I had to go with the -b option, given that -b2 doesn't use the SSE transforms. Both SSE4 and SSE2 saw major improvements, but are still slightly slower than 3.3.2.

Ran new 3.3.3-1 benches on my X5675:

3.3.2/SSE2:

Hi all,

Genefer 3.3.3 is now ready for testing. The test programme is quite comprehensive as we need to check changes (quite substantial) to the CPU transforms, and some minor changes to the OCL transforms.

Binaries are available from SVN:

Windows: https://www.assembla.com/spaces/genefer/subversion/source/HEAD/trunk/bin/windows

Linux: https://www.assembla.com/spaces/genefer/subversion/source/HEAD/trunk/bin/linux

Mac: https://www.assembla.com/spaces/genefer/subversion/source/HEAD/trunk/bin/mac

These binaries are version "3.3.3-2", due to some performance tweaks which had to be made.

There is a full list of tests, with some instructions for how and what to run on a google sheet:

https://docs.google.com/spreadsheets/d/13K5NKoywrlE4UtDZTM0ws8mMoz6y11b7kR4hFbehnBo/edit?usp=sharing

There are a large set of manual tests, for n=16,17 and n=22 (which will take a long time, unfortunately!), and also some BOINC tests to run under app_info.xml. Please check carefully as some tests require specifying the transform to use (via the -x option), and some leave genefer free to select the appropriate transform.

Please post into this thread if you want to reserve tests, and if you have results and/or questions. I will try to keep the thread clean by hiding posts when I update the google sheet. Manual credit will be awarded for the standalone tests.

Thanks in advance for the help - the more people contribute, the sooner we can start using the new apps.

Cheers

- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime!

I want to point out that one significant change to GeneferOCL is that we're reversing a change we made back in 2014.

Originally, GeneferCUDA ran at normal priority. Most BOINC programs run at idle priority so as not to interfere with whatever else the computer is doing. But a GPU program should run at normal priority so that the CPU can keep the GPU running full speed.

When Genefer OCL first came out, because of a driver "feature" the CPU portion of the program could consume a full CPU core, so we had to lower the priority of GeneferOCL.

Eventually that driver 'feature' was fixed, or we found a way around it, and as a result GeneferOCL no longer uses a whole CPU core.

Now, with 3.3.3, we're able to once again raise the priority of GeneferOCL. As a result, the GPU doesn't get starved even if you're running LLR on all cores. You no longer need to keep a core free for the GPU. The GPU program takes precedence over the other BOINC tasks.

Note that the GPU versions of AP27 and PPS-Sieve already work this way, so you don't need to keep a core free for them either.
____________
My lucky number is 75898⁵²⁴²⁸⁸+1

It's been a long time since the last update to 3.3.3, so I definitely want to ask: has the performance regression of SSE2/4 been reversed?

I would like to reserve/do the Windows non-boinc testing for 64 bit CPU and NVIDIA OCL through the n17 tests, after that I'll see about the n22 tests.
____________
Eating more cheese on Thursdays.

geneferocl_linux64 starts 8 threads, but only the main one gets essentially all the CPU time (another one amassed a total of 0.10 seconds by the time the main thread had 10.5 minutes, and a third got 0.01 seconds). Is this normal?

D48, D50, D52, D54, D56; note the failure and successful completion from checkpoint on D56

geneferocl 3.3.3-2 (Linux/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_linux64 -q 52186^131072+1 -x OCL

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL transform
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 760', vendor 'NVIDIA Corporation', version 'OpenCL 1.2 CUDA' and driver '384.111'.
6 computeUnits @ 1137MHz, memSize=1996MB, cacheSize=96kB, cacheLineSize=128B, localMemSize=48kB, maxWorkGroupSize=1024.
Starting initialization...
Initialization complete (0.113 seconds).
Testing 52186^131072+1... 2054077 steps to go
Estimated time for 52186^131072+1 is 0:09:16
52186^131072+1 is composite. (RES=1b196d6c0e4d778f) (618340 digits) (err = 0.0004) (time = 0:09:20) 21:01:23
geneferocl 3.3.3-2 (Linux/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_linux64 -q 52186^131072+1 -x OCL2

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL2 transform
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 760', vendor 'NVIDIA Corporation', version 'OpenCL 1.2 CUDA' and driver '384.111'.
6 computeUnits @ 1137MHz, memSize=1996MB, cacheSize=96kB, cacheLineSize=128B, localMemSize=48kB, maxWorkGroupSize=1024.
Starting initialization...
Initialization complete (0.113 seconds).
Testing 52186^131072+1... 2054077 steps to go
Estimated time for 52186^131072+1 is 0:13:50
52186^131072+1 is composite. (RES=1b196d6c0e4d778f) (618340 digits) (err = 0.0000) (time = 0:13:43) 21:15:06
geneferocl 3.3.3-2 (Linux/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_linux64 -q 52186^131072+1 -x OCL3

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL3 transform
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 760', vendor 'NVIDIA Corporation', version 'OpenCL 1.2 CUDA' and driver '384.111'.
6 computeUnits @ 1137MHz, memSize=1996MB, cacheSize=96kB, cacheLineSize=128B, localMemSize=48kB, maxWorkGroupSize=1024.
Starting initialization...
Initialization complete (0.113 seconds).
Testing 52186^131072+1... 2054077 steps to go
Estimated time for 52186^131072+1 is 0:11:40
52186^131072+1 is composite. (RES=1b196d6c0e4d778f) (618340 digits) (err = 0.0000) (time = 0:11:49) 21:26:55
geneferocl 3.3.3-2 (Linux/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_linux64 -q 52186^131072+1 -x OCL4

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL4 transform
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 760', vendor 'NVIDIA Corporation', version 'OpenCL 1.2 CUDA' and driver '384.111'.
6 computeUnits @ 1137MHz, memSize=1996MB, cacheSize=96kB, cacheLineSize=128B, localMemSize=48kB, maxWorkGroupSize=1024.
Starting initialization...
Initialization complete (0.113 seconds).
Testing 52186^131072+1... 2054077 steps to go
Estimated time for 52186^131072+1 is 0:08:03
52186^131072+1 is composite. (RES=1b196d6c0e4d778f) (618340 digits) (err = 0.0000) (time = 0:08:06) 21:35:01
geneferocl 3.3.3-2 (Linux/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_linux64 -q 52186^131072+1 -x OCL5

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL5 transform
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 760', vendor 'NVIDIA Corporation', version 'OpenCL 1.2 CUDA' and driver '384.111'.
6 computeUnits @ 1137MHz, memSize=1996MB, cacheSize=96kB, cacheLineSize=128B, localMemSize=48kB, maxWorkGroupSize=1024.
Starting initialization...
Initialization complete (0.113 seconds).
Testing 52186^131072+1... 2054077 steps to go
Estimated time for 52186^131072+1 is 0:09:34
maxErr exceeded for 52186^131072+1, 1.0000 > 0.4500
Errors occurred for all available transform implementations

reran OCL5, it picked up from the last checkpoint
geneferocl 3.3.3-2 (Linux/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_linux64 -q 52186^131072+1 -x OCL5

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL5 transform
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 760', vendor 'NVIDIA Corporation', version 'OpenCL 1.2 CUDA' and driver '384.111'.
6 computeUnits @ 1137MHz, memSize=1996MB, cacheSize=96kB, cacheLineSize=128B, localMemSize=48kB, maxWorkGroupSize=1024.
Resuming 52186^131072+1 from a checkpoint (2054077 iterations left)
Estimated time for 52186^131072+1 is 0:09:30
52186^131072+1 is composite. (RES=1b196d6c0e4d778f) (618340 digits) (err = 0.0000) (time = 0:09:28) 21:47:16

It's been a long time since the last update to 3.3.3, so I definitely want to ask: has the performance regression of SSE2/4 been reversed?

C10; did not switch to X87 transform

B10 reserved
...
EDIT: This was completed before the spreadsheet changed. I will run with the other transforms now

geneferocl 3.3.3-2 (Linux/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_linux64 -q 52186^131072+1 -x OCL5

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL5 transform
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 760', vendor 'NVIDIA Corporation', version 'OpenCL 1.2 CUDA' and driver '384.111'.
6 computeUnits @ 1137MHz, memSize=1996MB, cacheSize=96kB, cacheLineSize=128B, localMemSize=48kB, maxWorkGroupSize=1024.
Starting initialization...
Initialization complete (0.113 seconds).
Testing 52186^131072+1... 2054077 steps to go
Estimated time for 52186^131072+1 is 0:09:34
maxErr exceeded for 52186^131072+1, 1.0000 > 0.4500
Errors occurred for all available transform implementations

reran OCL5, it picked up from the last checkpoint
geneferocl 3.3.3-2 (Linux/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_linux64 -q 52186^131072+1 -x OCL5

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL5 transform
Running on platform 'NVIDIA CUDA', device 'GeForce GTX 760', vendor 'NVIDIA Corporation', version 'OpenCL 1.2 CUDA' and driver '384.111'.
6 computeUnits @ 1137MHz, memSize=1996MB, cacheSize=96kB, cacheLineSize=128B, localMemSize=48kB, maxWorkGroupSize=1024.
Resuming 52186^131072+1 from a checkpoint (2054077 iterations left)
Estimated time for 52186^131072+1 is 0:09:30
52186^131072+1 is composite. (RES=1b196d6c0e4d778f) (618340 digits) (err = 0.0000) (time = 0:09:28) 21:47:16

geneferocl_linux64 starts 8 threads, but only the main one gets essentially all the CPU time (another one amassed a total of 0.10 seconds by the time the main thread had 10.5 minutes, and a third got 0.01 seconds). Is this normal?

This is strange. b=52186 is well below the b limit, and for OCL5 the calculation should be exact (err=0.0000). Does this recreate on your hardware? It seems to run fine for Grebuloner and Van Zimmerman, and also I tested on my own linux system with a Tesla K40m. If not I can only think it is either a hardware glitch, or a (very rare) software bug. In either case the error correction worked and it restarted from the last known good checkpoint...

There's something slightly fishy given that the error value was exacly 1.000.

Is this the high limit of the range of error values?

Might this indicate some very early failure before the calculation has propagated and averaged down?

There's something slightly fishy given that the error value was exacly 1.000.

Is this the high limit of the range of error values?

Might this indicate some very early failure before the calculation has propagated and averaged down?

F74 & F76 are in progress, with G74 and G76 starting shortly.

Please note there are new Windows CPU builds (3.3.3-3), which fix a minor issue with priority when running standalone (i.e. BOINC is not affected). Anyone who is running the CPU code on Windows - please go back to https://app.assembla.com/spaces/genefer/subversion/source/HEAD/trunk/bin/windows and grab the updated binary!

Thanks for all the testing so far!

Cheers

- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime!

Bit of a toaster of a datapoint compared to most things, but on my i7-4710MQ (2.5GHz) laptop (w7x64),

G48-G56:

OCL5 Returned an error, and it looks like it isn't the first time it has occurred. I attempted to re-run it, and got the same result.

dora:test vzimmerman$ ./test.sh
geneferocl 3.3.3-2 (Apple-x86/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_macintel -q 52186^131072+1 -x OCL

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL transform
Running on platform 'Apple', device 'AMD Radeon RX 480 Compute Engine', vendor 'AMD', version 'OpenCL 1.2 ' and driver '1.2 (Nov 9 2017 18:48:40)'.
36 computeUnits @ 1266MHz, memSize=8192MB, cacheSize=0kB, cacheLineSize=0B, localMemSize=32kB, maxWorkGroupSize=256.
Starting initialization...
Initialization complete (0.117 seconds).
Estimated time for 52186^131072+1 is 0:07:18
52186^131072+1 is composite. (RES=1b196d6c0e4d778f) (618340 digits) (err = 0.0004) (time = 0:07:20) 16:54:56
geneferocl 3.3.3-2 (Apple-x86/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_macintel -q 52186^131072+1 -x OCL2

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL2 transform
Running on platform 'Apple', device 'AMD Radeon RX 480 Compute Engine', vendor 'AMD', version 'OpenCL 1.2 ' and driver '1.2 (Nov 9 2017 18:48:40)'.
36 computeUnits @ 1266MHz, memSize=8192MB, cacheSize=0kB, cacheLineSize=0B, localMemSize=32kB, maxWorkGroupSize=256.
Starting initialization...
Initialization complete (0.117 seconds).
Estimated time for 52186^131072+1 is 0:13:00
52186^131072+1 is composite. (RES=1b196d6c0e4d778f) (618340 digits) (err = 0.0000) (time = 0:13:01) 17:07:57
geneferocl 3.3.3-2 (Apple-x86/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_macintel -q 52186^131072+1 -x OCL3

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL3 transform
Running on platform 'Apple', device 'AMD Radeon RX 480 Compute Engine', vendor 'AMD', version 'OpenCL 1.2 ' and driver '1.2 (Nov 9 2017 18:48:40)'.
36 computeUnits @ 1266MHz, memSize=8192MB, cacheSize=0kB, cacheLineSize=0B, localMemSize=32kB, maxWorkGroupSize=256.
Starting initialization...
Initialization complete (0.117 seconds).
Estimated time for 52186^131072+1 is 0:10:50
52186^131072+1 is composite. (RES=1b196d6c0e4d778f) (618340 digits) (err = 0.0000) (time = 0:11:00) 17:18:57
geneferocl 3.3.3-2 (Apple-x86/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_macintel -q 52186^131072+1 -x OCL4

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL4 transform
Running on platform 'Apple', device 'AMD Radeon RX 480 Compute Engine', vendor 'AMD', version 'OpenCL 1.2 ' and driver '1.2 (Nov 9 2017 18:48:40)'.
36 computeUnits @ 1266MHz, memSize=8192MB, cacheSize=0kB, cacheLineSize=0B, localMemSize=32kB, maxWorkGroupSize=256.
Starting initialization...
Initialization complete (0.117 seconds).
Estimated time for 52186^131072+1 is 0:09:19
52186^131072+1 is composite. (RES=1b196d6c0e4d778f) (618340 digits) (err = 0.0000) (time = 0:09:21) 17:28:18
geneferocl 3.3.3-2 (Apple-x86/OpenCL/64-bit)

Copyright 2001-2018, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2018, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./geneferocl_macintel -q 52186^131072+1 -x OCL5

Normal priority change succeeded.

Testing 52186^131072+1...
Using OCL5 transform
Running on platform 'Apple', device 'AMD Radeon RX 480 Compute Engine', vendor 'AMD', version 'OpenCL 1.2 ' and driver '1.2 (Nov 9 2017 18:48:40)'.
36 computeUnits @ 1266MHz, memSize=8192MB, cacheSize=0kB, cacheLineSize=0B, localMemSize=32kB, maxWorkGroupSize=256.
Starting initialization...
Initialization complete (0.117 seconds).
Estimated time for 52186^131072+1 is 0:10:10
Testing 52186^131072+1... 1855860 steps to go (0:09:18 remaining)
maxErr exceeded for 52186^131072+1, 1.0000 > 0.4500
Errors occurred for all available transform implementations

G48-G56:

OCL5 Returned an error, and it looks like it isn't the first time it has occurred. I attempted to re-run it, and got the same result.

G48-G56:

OCL5 Returned an error, and it looks like it isn't the first time it has occurred. I attempted to re-run it, and got the same result.

Would be good if you could try running this test several times - and also with the previous genefer release (current stock PrimeGrid app will do). These does seem to be an issue here that we need to understand!

Retesting the Win64 CPU tests on 3.3.3-3.

I am attempting to run the 32 bit Win CPU tests. It's not a pure environment (x64 Core 2 Duo on x86 Win 7), and I'm turning this laptop on for the first time in about 3 years. If it fails, would you accept a virtualized 32 bit Windows test?

EDIT: It's dead, Jim. We'll see if I can reinstall...
____________
Eating more cheese on Thursdays.

Not enough memory in an IBM T40 Thinkpad to emulate a disk big enough to run GFN when running Debian Live i386 on it. I guess it would be easiest to plug in a USB hard disk and tell it to swap there.

Repeat Mac/AMD OCL5 tests resulted in errors. Same error occurred on production version.

dora:test1 vzimmerman$ ./primegrid_genefer_3_3_2_3.18_x86_64-apple-darwin__openclatiGFN18 -q 52186^131072+1 -x OCL5
geneferocl 3.3.2-7 (Apple-x86/OpenCL/64-bit)

Copyright 2001-2017, Yves Gallot
Copyright 2009, Mark Rodenkirch, David Underbakke
Copyright 2010-2012, Shoichiro Yamada, Ken Brazier
Copyright 2011-2014, Michael Goetz, Ronald Schneider
Copyright 2011-2017, Iain Bethune
Genefer is free source code, under the MIT license.

Command line: ./primegrid_genefer_3_3_2_3.18_x86_64-apple-darwin__openclatiGFN18 -q 52186^131072+1 -x OCL5

Low priority change succeeded.

Testing 52186^131072+1...
Using OCL5 transform
Running on platform 'Apple', device 'ATI Radeon RX 480 Compute Engine', vendor 'AMD', version 'OpenCL 1.2 ' and driver '1.2 (Nov 9 2017 18:48:40)'.
36 computeUnits @ 1266MHz, memSize=8192MB, cacheSize=0kB, cacheLineSize=0B, localMemSize=32kB, maxWorkGroupSize=256.
Starting initialization...
Initialization complete (0.120 seconds).
Estimated time for 52186^131072+1 is 0:11:10
Testing 52186^131072+1... 381732 steps to go (0:02:03 remaining)
maxErr exceeded for 52186^131072+1, 1.0000 > 0.4500
Errors occurred for all available transform implementations

Repeat Mac/AMD OCL5 tests resulted in errors. Same error occurred on production version.

I got my little C2D back up again, reserving/doing 32 bit Win CPU tests.

EDIT: Nevermind. narrowly beaten to the punch.
____________
Eating more cheese on Thursdays.

I might even be able to run tests on Pentium 4. Which would be an even "purer" in that it doesn't have 64-bit instructions at all. Sorry, Grebuloner :[

Ok, that second computer turned out to be Pentium 4 511 (prescott, 90nm), one of the first chips to get Intel's x86-64-bit instruction set. So there's no real "pureness" difference testing on it or on that pentium dual-core. Meh.
So I ran a bunch of tests to see the application works there, but I'm only going to commit to fully running the listed tests on the first computer above which I posted first.

Grebuloner, by the way, I'm not going to run the BOINC win32 test, you can still grab that one, if you want!

I have an old pc running 32-bit windows xp. It has an Intel Pentium E2180 (65nm technology, no SSE4 yet, but dual core). Should be fine for a "pure" windows32 test?

Update on OCL5 "bug" seen on 52186^131072+1:

As far as we know at present this is likely to be a hardware-related issue rather than a code bug. Round-off errors have been seen on Van's Mac with Radeon RX 480 and composite's Linux box with GeForce GTX 760. On Van's machine the error is reproducible on most runs, but not always at the same point in the calculation. On composite's it has only been seen once. Van also sees the same behaviour with the current stock app (3.3.2).

We have over 40 successful tests in both Linux/Nvidia and Mac/AMD setups, and have additional successful tests on Windows/Nvidia and Mac/Nvidia.

composite - would you be able to re-run the test several times on your host where you first saw the bug and see if appears to be recreatable?

In any case, I think we can continue with the test programme - if anyone has Linux and/or Windows systems with AMD GPUs, that could run some tests that would be very helpful - PM me if you're unsure what to do!

- Iain


____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime!

I still get the same error with OCL5 on Linux testing 52186^131072+1. The first time I tried it today, I got the error twice during the same test (with resumption). Sometimes I don't get the error. It's random.

This GPU works reliably with OCL2. No invalid tasks returned in the last 1256 GFN16 tasks completed during the challenge.

EDIT: When I remove the checkpoint file after failure, the error occurs at different iterations.

G76 errored out, but I suspect that is my now-untrustworthe RX480.

Genefer 3.3.3 has now been released and I've applied manual credit for all of the tests that were run. Thanks to everyone who contributed to the testing - and congrats to dukebg, composite and Kiska on your new "Volunteer tester" status!

- Iain
____________
Twitter: IainBethune
Proud member of team "Aggie The Pew". Go Aggie!
3073428256125*2^1290000-1 is Prime!