Some CRUS stats

[rogue]

I just added another feature to the unproven page, filtering. Take a look and let me know what you think. I intend to use it to help me choose future conjectures to work on.

[mdettweiler]

Big surprise: Mark's new stats scripts are now online! I had a few minutes today sitting in front of the computer and figured "hmm, I really should do this already..." and now 10 minutes later it's done.

Yeah, gee...this has been sitting for five months and IT'S A STINKING 10 MINUTE JOB! Out of sight, out of mind, you know? Thanks Mathew for the little reminder this morning--this should have been done a long time ago.

To review (and to confirm that I didn't forget anything), the new script generates four pages, crus-stats.htm, crus-top20.htm, crus-proven.htm, and crus-unproven.htm. Each of these are now linked from the main CRUS page:

http://www.noprimeleftbehind.net/crus/

The script runs hourly at :30.

Sorry everyone for the wait on this.

[rogue]

Thanks Max!

I see that the top Riesel in this list: Top 20 Conjectures with 1k Remaining by Lowest Weight, has a negative weight and there are a number on the Unproven Conjectures page with a negative weight. Max, did you delete the temp files that were there before installing the scripts? I wonder if a bad value was picked up from old data. The other possibility is that you are using an older version of srsieve. I am using 1.0.3.

[mdettweiler]

Originally Posted by rogue

Thanks Max!

I see that the top Riesel in this list: Top 20 Conjectures with 1k Remaining by Lowest Weight, has a negative weight and there are a number on the Unproven Conjectures page with a negative weight. Max, did you delete the temp files that were there before installing the scripts? I wonder if a bad value was picked up from old data. The other possibility is that you are using an older version of srsieve. I am using 1.0.3.

Ah, I just extracted the http://home.roadrunner.com/~mrodenkirch/vstats.zip file and ran it like that...seeing as how there's no srsieve in the directory that would probably mess things up a bit.

I downloaded the latest srsieve 1.0.4 package from your website and tried compiling it on jeepford (Ubuntu 9.04 64-bit), but got this error:

gcc -m64 -s  srsieve.o arithmetic32.o arithmetic64.o bitmap.o bsgs.o clock.o events.o factors.o files.o global.o hashtable.o primes.o sequences.o subseq.o util.o priority.o powmod-k8.o  -lm -o srsieve
bsgs.o: In function `bsgs64_noflt':
bsgs.c:(.text+0x294b): undefined reference to `_powmod64_k8'
bsgs.c:(.text+0x296c): undefined reference to `_powmod64_k8'
bsgs.c:(.text+0x29cc): undefined reference to `_powmod64_k8'
bsgs.o: In function `bsgs64_flt':
bsgs.c:(.text+0x2d47): undefined reference to `_powmod64_k8'
bsgs.c:(.text+0x2d68): undefined reference to `_powmod64_k8'
bsgs.o:bsgs.c:(.text+0x2f29): more undefined references to `_powmod64_k8' follow
collect2: ld returned 1 exit status
make: *** [srsieve] Error 1

Are there by chance Linux binaries posted somewhere for this version? I saw Windows .exe's in the file but none for Linux.

Edit: I do have an srsieve 0.6.17 binary on hand; will that do or does the script rely on any of the features you added later?

[rogue]

Originally Posted by mdettweiler

Ah, I just extracted the http://home.roadrunner.com/~mrodenkirch/vstats.zip file and ran it like that...seeing as how there's no srsieve in the directory that would probably mess things up a bit.

I downloaded the latest srsieve 1.0.4 package from your website and tried compiling it on jeepford (Ubuntu 9.04 64-bit), but got this error:

gcc -m64 -s  srsieve.o arithmetic32.o arithmetic64.o bitmap.o bsgs.o clock.o events.o factors.o files.o global.o hashtable.o primes.o sequences.o subseq.o util.o priority.o powmod-k8.o  -lm -o srsieve
bsgs.o: In function `bsgs64_noflt':
bsgs.c:(.text+0x294b): undefined reference to `_powmod64_k8'
bsgs.c:(.text+0x296c): undefined reference to `_powmod64_k8'
bsgs.c:(.text+0x29cc): undefined reference to `_powmod64_k8'
bsgs.o: In function `bsgs64_flt':
bsgs.c:(.text+0x2d47): undefined reference to `_powmod64_k8'
bsgs.c:(.text+0x2d68): undefined reference to `_powmod64_k8'
bsgs.o:bsgs.c:(.text+0x2f29): more undefined references to `_powmod64_k8' follow
collect2: ld returned 1 exit status
make: *** [srsieve] Error 1

Are there by chance Linux binaries posted somewhere for this version? I saw Windows .exe's in the file but none for Linux.

Edit: I do have an srsieve 0.6.17 binary on hand; will that do or does the script rely on any of the features you added later?

That's odd. _powmod64_k8 is in powmod-k8, which is being linked. Do any of you other linux guys have any ideas?

Have you tried a 32-bit build? I wonder if that has the same problem.

You will need 1.0.3 or 1.0.4.

[mdettweiler]

Originally Posted by rogue

That's odd. _powmod64_k8 is in powmod-k8, which is being linked. Do any of you other linux guys have any ideas?

Have you tried a 32-bit build? I wonder if that has the same problem.

You will need 1.0.3 or 1.0.4.

I tried a 32-bit build on a VM running Linux Mint Debian Edition (it's a rolling distribution, so in lieu of a version number all I can give you is that it's up to date as of today ), but it gave me a different error:

gcc -O2 -fomit-frame-pointer -ffast-math -march=k8 -m64  -Wall -DHAVE_CMOV  -DUSE_ASM -DNDEBUG  -c -o srsieve.o srsieve.c
In file included from /usr/include/assert.h:37:0,
                 from srsieve.c:11:
/usr/include/features.h:323:26: fatal error: bits/predefs.h: No such file or directory
compilation terminated.
make: *** [srsieve.o] Error 1

[mdettweiler]

Mathew just sent me his 64-bit Linux build of srsieve; however, it's not working either:

gary@jeepford:/home/nplb/html/crus/vstats_new$ ./srsieve -v
./srsieve: /lib/libm.so.6: version `GLIBC_2.15' not found (required by ./srsieve)

I suspect this is because jeepford is running an old version of Ubuntu (9.04) which hasn't been supported for a year, and thus probably has all sorts of out-of-date libraries. Dave and I are working on getting it upgraded to the latest version, but in the meantime, we'll need either a statically-linked binary or one compiled on something at least as old as 9.04. Ideally, compiling on jeepford itself would be sure to give us a working binary...if we can get it to compile.

[gd_barnes]

Max,

I've sieved a fair amount on Jeepford. There should be several older versions of sr(x)sieve that work on it. Just look in the "Prime" folder on the desktop. Then look in any of the "primex" folders within that. You should see folders for srsieve, sr1sieve, and sr2sieve.

The only problem is that they are old enough that they probably wouldn't remove the n's with algebraic factors like the more recent versions of srsieve would.


Gary

[henryzz]

Originally Posted by mdettweiler

Ah, I just extracted the http://home.roadrunner.com/~mrodenkirch/vstats.zip file and ran it like that...seeing as how there's no srsieve in the directory that would probably mess things up a bit.

I downloaded the latest srsieve 1.0.4 package from your website and tried compiling it on jeepford (Ubuntu 9.04 64-bit), but got this error:

gcc -m64 -s  srsieve.o arithmetic32.o arithmetic64.o bitmap.o bsgs.o clock.o events.o factors.o files.o global.o hashtable.o primes.o sequences.o subseq.o util.o priority.o powmod-k8.o  -lm -o srsieve
bsgs.o: In function `bsgs64_noflt':
bsgs.c:(.text+0x294b): undefined reference to `_powmod64_k8'
bsgs.c:(.text+0x296c): undefined reference to `_powmod64_k8'
bsgs.c:(.text+0x29cc): undefined reference to `_powmod64_k8'
bsgs.o: In function `bsgs64_flt':
bsgs.c:(.text+0x2d47): undefined reference to `_powmod64_k8'
bsgs.c:(.text+0x2d68): undefined reference to `_powmod64_k8'
bsgs.o:bsgs.c:(.text+0x2f29): more undefined references to `_powmod64_k8' follow
collect2: ld returned 1 exit status
make: *** [srsieve] Error 1

Are there by chance Linux binaries posted somewhere for this version? I saw Windows .exe's in the file but none for Linux.

Edit: I do have an srsieve 0.6.17 binary on hand; will that do or does the script rely on any of the features you added later?

Relatively easy to fix. I just tried on an almost identical system(9.04 64-bit Q6600).
In powmod-k8.S, #define powmod64_k8 _powmod64_k8 is not being called when it is needed since NEED_UNDERSCORE is not defined. I removed that #if and its #endif and it compiled for me. Obviously there might be a less hackish fix but it seems to work for now.

[rogue]

Originally Posted by henryzz

Relatively easy to fix. I just tried on an almost identical system(9.04 64-bit Q6600).
In powmod-k8.S, #define powmod64_k8 _powmod64_k8 is not being called when it is needed since NEED_UNDERSCORE is not defined. I removed that #if and its #endif and it compiled for me. Obviously there might be a less hackish fix but it seems to work for now.

Thanks. I'll fix the sources. In the meantime, can you provide Max with a static-linked build?

[mdettweiler]

Originally Posted by henryzz

Relatively easy to fix. I just tried on an almost identical system(9.04 64-bit Q6600).
In powmod-k8.S, #define powmod64_k8 _powmod64_k8 is not being called when it is needed since NEED_UNDERSCORE is not defined. I removed that #if and its #endif and it compiled for me. Obviously there might be a less hackish fix but it seems to work for now.

Thanks, that worked! I got it to compile and work on jeepford with that change. (The dynamically linked binary will do fine since it's compiled on a suitably old system.)

The negative weights seem to be fixed now--a quick Ctrl-F search through the Top 20 page showed no "-" characters except those in the "even-n" and "odd-n" wordings.

(BTW, if anyone needs an srsieve 1.0.4 binary that runs on old(ish) 64-bit Linuxes, feel free to snarf the one I just built out of the script folder. )