Sr2sieve on PPC/Linux

[BlisteringSheep]

With the bugfix, plus changing CONST_EMPTY_SLOT to 0 and HASH_MAX_DENSITY to 0.32, compiling with gcc4, my 2.5 GHz 970MPs are up from ~292000 p/sec to ~308000 p/sec.

[Greenbank]

Originally Posted by rogue

I have isolated and resolved the issue. Apparently r2 must be saved and restored in this environment. Under OS X, r2 can be used without such a precaution. I've send the updated code to geoff.

Rogue, In which function? Will it affect the expmod and mulmod code you wrote for the G5 PPC proth_sieve? I need to make sure the source is ok so I can give Sheep a linux build of proth_sieve for G5 PPC.

I've been away with work and won't have access to the proth_sieve source until Monday. Plus I'm doing a 220km cycle ride on Sunday so I'll probably be a daze all day Monday. Sorry for the long delay on this Sheep.

[BlisteringSheep]

Originally Posted by Greenbank

Rogue, In which function? Will it affect the expmod and mulmod code you wrote for the G5 PPC proth_sieve? I need to make sure the source is ok so I can give Sheep a linux build of proth_sieve for G5 PPC.

I've been away with work and won't have access to the proth_sieve source until Monday. Plus I'm doing a 220km cycle ride on Sunday so I'll probably be a daze all day Monday. Sorry for the long delay on this Sheep.

This was in expmod.

Enyoy your ride

[rogue]

Originally Posted by Greenbank

Rogue, In which function? Will it affect the expmod and mulmod code you wrote for the G5 PPC proth_sieve? I need to make sure the source is ok so I can give Sheep a linux build of proth_sieve for G5 PPC.

I've been away with work and won't have access to the proth_sieve source until Monday. Plus I'm doing a 220km cycle ride on Sunday so I'll probably be a daze all day Monday. Sorry for the long delay on this Sheep.

It only affects expmod and will have a negligable affect on OS X builds. You can grab the source directly from srsieve as it was (originally) the same source I sent to you.

[geoff]

Originally Posted by BlisteringSheep

With the bugfix, plus changing CONST_EMPTY_SLOT to 0 and HASH_MAX_DENSITY to 0.32, compiling with gcc4, my 2.5 GHz 970MPs are up from ~292000 p/sec to ~308000 p/sec.

Thanks, I'll make those the default for ppc64 in future.

[Greenbank]

Thanks, I'll grab the latest source and see if I can get it done today.

Not feeling as knackered as expected as I had to abandon at 80km (not bad by midday) for various reasons.

[BlisteringSheep]

Geoff,
I got the new 1.4.7 tarball. Is it possible to create the cache file once and then link to it? I have a ton of machines all working out off of a NFS tree. I'm currently hard-linking all of the executables and SoB.dat, and would like to link the sr2cache.bin file as well (I prefer hard links to symlinks for things on the same file system).

I should be able to run some benchmarks & report back.

Thanks,

[BlisteringSheep]

All from a 2.2 GHz PPC970FX, with a sr2work.txt and inital checkpoint.txt of

sr2work.txt: 273020, 273021
checkpoint.txt: pmin=273020049854500,factors_count=0

• Default compile with gcc3

  p=273020986686673, 261113 p/sec, 2 factors, 98.67% done, ETA 05 Dec 10:45
  3639.176u 0.164s 1:00:40.15 99.9%

• Compile with gcc4

  p=273020988990671, 263521 p/sec, 2 factors, 98.90% done, ETA 05 Dec 10:44
  3600.868u 0.140s 1:00:01.75 99.9%

• Compile with gcc4 plus changing HASH_MAX_DENSITY to 0.32

  p=273020991038581, 265975 p/sec, 2 factors, 99.10% done, ETA 05 Dec 10:44
  3571.896u 0.155s 59:33.42 99.9%

• Version 1.4.3 with gcc4, HASH_MAX_DENSITY 0.32 and CONST_EMPTY_SLOT 0

  p=273020991166519, 266097 p/sec, 2 factors, 99.12% done, ETA 05 Dec 10:44
  3570.397u 0.160s 59:31.38 99.9%

[BlisteringSheep]

Note that USE_INLINE_MULMOD makes it run really fast (over 461000 p/sec), but it finds no factors at all (duplicates or new).

[axn]

Originally Posted by BlisteringSheep

• Compile with gcc4 plus changing HASH_MAX_DENSITY to 0.32

  p=273020991038581, 265975 p/sec, 2 factors, 99.10% done, ETA 05 Dec 10:44
  3571.896u 0.155s 59:33.42 99.9%

• Version 1.4.3 with gcc4, HASH_MAX_DENSITY 0.32 and CONST_EMPTY_SLOT 0

  p=273020991166519, 266097 p/sec, 2 factors, 99.12% done, ETA 05 Dec 10:44
  3570.397u 0.160s 59:31.38 99.9%

There are no speed improvements between 1.4.3 and 1.4.7 -- just some minor improvements in handling of srwork.txt and usage of a cache file for lookup tables. So you'll find that the speeds are virtually identical.

[BlisteringSheep]

Originally Posted by axn1

There are no speed improvements between 1.4.3 and 1.4.7 -- just some minor improvements in handling of srwork.txt and usage of a cache file for lookup tables. So you'll find that the speeds are virtually identical.

I expected as much looking at the code. I just like to be thorough while also being helpful.

For PPC64, there was the improvement of making CONST_EMPTY_SLOT be 0. I think it also shows that it's worthwhile to have MAX_HASH_DENSITY reduced. I haven't experimented with different values; I've just been using Geoff's recommendation to halve it. I'll try some others.