mtsieve - Page 121

rogue · 2023-06-30, 12:44

The new issues with srsieve2 (which has been around a while) is fixed in sourceforge.

pepi37 · 2023-06-30, 14:13

Quote:

Originally Posted by rogue

The new issues with srsieve2 (which has been around a while) is fixed in sourceforge.

Thanks now it is working, with l increased to 5e13

rogue · 2023-06-30, 15:43

Quote:

Originally Posted by pepi37

Thanks now it is working, with l increased to 5e13

The bug had nothing to do with the value specified by -l. When it runs it tells you how much memory it actually used. In your case it won't need that much memory and would fail if it tried to allocate more than what is physically available on your system.

Think of it this way. If you don't want srsieve2 to use all available memory on your system, then use -l to limit how much it can use. If you don't care, use a large enough value with -l so that it can hold the tables without running out of system memory.

pepi37 · 2023-06-30, 16:49

Quote:

Originally Posted by rogue

The bug had nothing to do with the value specified by -l. When it runs it tells you how much memory it actually used. In your case it won't need that much memory and would fail if it tried to allocate more than what is physically available on your system.

Think of it this way. If you don't want srsieve2 to use all available memory on your system, then use -l to limit how much it can use. If you don't care, use a large enough value with -l so that it can hold the tables without running out of system memory.

Only thing that is "problem " is less speed. I download latest tree from SVN, compile it and it works but in around half speed of your compiled exe... ( or maybe it is not connected to all)

your exe

Code:

srsieve2.exe -P 10000000000 -n 500000 -N 1000000 -W 8 -l 5e13 -s"4*767^n+1" -s"16*767^n+1" -s"52*767^n+1" -f B
srsieve2 v1.7.3, a program to find factors of k*b^n+c numbers for fixed b and variable k and n
Increasing worksize to 400000 since each chunk is tested in less than a second
Increasing worksize to 2000000 since each chunk is tested in less than a second
  p=4856215739, 3.887M p/sec, 1336231 factors found at 164.2 f/sec (last 1 min), 48.6% done. ETC 2023-06-30 17:37

my exe latest revision

Code:

srsieve2.exe -P 10000000000 -n 500000 -N 1000000 -W 8 -l 5e13 -s"4*767^n+1" -s"16*767^n+1" -s"52*767^n+1" -f B
srsieve2 v1.7.3, a program to find factors of k*b^n+c numbers for fixed b and variable k and n
Increasing worksize to 400000 since each chunk is tested in less than a second
  p=1428220663, 1.198M p/sec, 1333856 factors found at 160.4 f/sec (last 1 min), 14.3% done. ETC 2023-06-30 17:51

I also noticed my exe file is around double size of your ( different compiler)?

rogue · 2023-06-30, 19:52

Quote:

Originally Posted by pepi37

Only thing that is "problem " is less speed. I download latest tree from SVN, compile it and it works but in around half speed of your compiled exe... ( or maybe it is not connected to all)

your exe

Code:

srsieve2.exe -P 10000000000 -n 500000 -N 1000000 -W 8 -l 5e13 -s"4*767^n+1" -s"16*767^n+1" -s"52*767^n+1" -f B
srsieve2 v1.7.3, a program to find factors of k*b^n+c numbers for fixed b and variable k and n
Increasing worksize to 400000 since each chunk is tested in less than a second
Increasing worksize to 2000000 since each chunk is tested in less than a second
  p=4856215739, 3.887M p/sec, 1336231 factors found at 164.2 f/sec (last 1 min), 48.6% done. ETC 2023-06-30 17:37

my exe latest revision

Code:

srsieve2.exe -P 10000000000 -n 500000 -N 1000000 -W 8 -l 5e13 -s"4*767^n+1" -s"16*767^n+1" -s"52*767^n+1" -f B
srsieve2 v1.7.3, a program to find factors of k*b^n+c numbers for fixed b and variable k and n
Increasing worksize to 400000 since each chunk is tested in less than a second
  p=1428220663, 1.198M p/sec, 1333856 factors found at 160.4 f/sec (last 1 min), 14.3% done. ETC 2023-06-30 17:51

I also noticed my exe file is around double size of your ( different compiler)?

I use llvm-mingw-20220323-ucrt-x86_64

pepi37 · 2023-06-30, 23:24

Quote:

Originally Posted by rogue

I use llvm-mingw-20220323-ucrt-x86_64

Changed to your compiler, reduce file size, but speed, remain slow.
I will wait your release :)

rogue · 2023-06-30, 23:34

Quote:

Originally Posted by pepi37

Changed to your compiler, reduce file size, but speed, remain slow.
I will wait your release :)

I have no explanation. Are the file sizes the same? Is the speed as slow as with your build using the other compiler?

pepi37 · 2023-06-30, 23:41

Quote:

Originally Posted by rogue

I have no explanation. Are the file sizes the same? Is the speed as slow as with your build using the other compiler?

After installing GMP got same speed as your build
Thanks

rogue · 2023-07-01, 13:10

Quote:

Originally Posted by pepi37

After installing GMP got same speed as your build
Thanks

That's weird. Not certain why. GMP isn't used by srsieve2.

pepi37 · 2023-07-01, 13:44

Quote:

Originally Posted by rogue

That's weird. Not certain why. GMP isn't used by srsieve2.

First I have to say sorry if I was boring. I try to learn to compile so I made many errors, since I was tired, etc etc...
I have many errors in compile process, but as always be ,morning is smarter then evening :)
Today I try many llmv compilers, and as you say, some one is working someone is not.

I found that llvm-mingw-20220323-ucrt-x86_64-3.543.zip , llvm-mingw-20220906-ucrt-x86_64-386.zip and llvm-mingw-20230320-ucrt-x86_64-326.zip working ok, create srsieve2 ok and have same speed as your build.
Maybe fastest build was from llvm-mingw-20220906-ucrt-x86_64-386 (386 at the end show 3.86 MP/sec)
Also I confirm that MSYS64 create double size exe file but doesnot have speed as build from llvm
Also MSYS64 show this warning ( not error)

Code:

sierpinski_riesel/GenericSequenceHelper.cpp:114:13: warning: variable 'A' set but not used [-Wunused-but-set-variable]
  114 |    uint32_t A[9], P[9], M[9], i, k, t;

And last my error: why I dont have speed from your build, I forget to add -l switch to command to start srsieve2 :)
P.S and as you say I dont need GMP at all, I make test build without GMP dir and speed is same.

Citrix · 2023-07-01, 17:34

Code:

gcwsieve -W16   -P10e14 -Ogcwfactors.txt -n299830 -N299840 -b145 -s-
gcwsieve v1.5.2, a program to find factors numbers of the form n*b^n+1 and n*b^n-1
5 terms removed because the term is even
Sieve started: 3 <= p <= 1e15 with 6 terms (299830 <= n <= 299840, n*145^n-1) (expecting 6 factors)
Increasing worksize to 1600000 since each chunk is tested in less than a second
Increasing worksize to 160000000 since each chunk is tested in less than a second
Decreasing worksize to 10000000 since each chunk needs more than 5 seconds to test
  p=59305138787, 42.31M p/sec, 4 factors found at 216 sec per factor (last 1 min), 0.0% done. ETC 2023-07-13 10:13

One of the remaining terms (299832) has a small factor 7 which the program does not detect.

2023-06-30, 12:44	#1321
rogue "Mark" Apr 2003 Between here and the 2·3·1,223 Posts	The new issues with srsieve2 (which has been around a while) is fixed in sourceforge.