mersenneforum.org mtsieve
 Register FAQ Search Today's Posts Mark Forums Read

2023-02-07, 15:14   #991
storm5510
Random Account

Aug 2009
Not U. + S.A.

1010010000102 Posts

Quote:
 Originally Posted by rogue CPU utilization is used in the calculation. Short runtimes will skew the calculations too.
Very well. I noticed srsieve2cl uses the CPU, to a point. Somewhere between 15% to 20% utilization.

2023-02-07, 15:51   #992
rogue

"Mark"
Apr 2003
Between here and the

3×2,357 Posts

Quote:
 Originally Posted by storm5510 Very well. I noticed srsieve2cl uses the CPU, to a point. Somewhere between 15% to 20% utilization.
I should have been more specific. It combines execution time in the GPU with processing time in the CPU. These are not necessarily 1::1. Execution time in the GPU is just a sum of (stopKernelTime - startKernelTime) for each kernel execution. It does not take into account utilization of the GPU. Processing time in the CPU is more precise. The value from the OS takes into account CPU utilization. This makes is easier for the code to compute the percentage of CPU cores that are used with 100% meaning full utilization of a single core.

2023-02-07, 18:48   #993
storm5510
Random Account

Aug 2009
Not U. + S.A.

2×13×101 Posts

Quote:
 Originally Posted by rogue I should have been more specific. It combines execution time in the GPU with processing time in the CPU. These are not necessarily 1::1. Execution time in the GPU is just a sum of (stopKernelTime - startKernelTime) for each kernel execution. It does not take into account utilization of the GPU. Processing time in the CPU is more precise. The value from the OS takes into account CPU utilization. This makes is easier for the code to compute the percentage of CPU cores that are used with 100% meaning full utilization of a single core.
Thank you!

The CPU utilization is being spread across 8 threads at 16%. However, the fine print near the top of the attached image says 100%. Some threads are being used more than others according to the graph. Combining the actual usage of each thread would put it at 100%. Since the removal rate is not real-time, the best thing I can do is use it like a sort of benchmark.

The 60% test I ran with LLR indicated I needed to be consistently above 1,994 seconds per factor with what I am running. This test is real-time so I am not sure I can use this as an indicator of where I need to be with srsieve2cl.
Attached Thumbnails

 2023-02-07, 21:05 #994 Dylan14     "Dylan" Mar 2017 3×199 Posts As a heads up, with new versions of the ld program, the following warning comes up when compiling: Code: /usr/bin/ld: warning: x86_asm/avx_powmod.o: missing .note.GNU-stack section implies executable stack /usr/bin/ld: NOTE: This behaviour is deprecated and will be removed in a future version of the linker This is due to the fact that making the stack executable can result in a security risk (see this for instance). There doesn't seem to be any ill effect with adding -z noexecstack to the LD_FLAGS in the makefile with regards to the program's performance (at least on small files), so it's probably worth adding this.
2023-02-07, 21:07   #995
rogue

"Mark"
Apr 2003
Between here and the

1B9F16 Posts

Quote:
 Originally Posted by Dylan14 As a heads up, with new versions of the ld program, the following warning comes up when compiling: Code: /usr/bin/ld: warning: x86_asm/avx_powmod.o: missing .note.GNU-stack section implies executable stack /usr/bin/ld: NOTE: This behaviour is deprecated and will be removed in a future version of the linker This is due to the fact that making the stack executable can result in a security risk (see this for instance). There doesn't seem to be any ill effect with adding -z noexecstack to the LD_FLAGS in the makefile with regards to the program's performance (at least on small files), so it's probably worth adding this.
Okay. I will look into this when I have some time.

 2023-02-14, 10:09 #996 pepi37     Dec 2011 After milion nines:) 163910 Posts Question for srsieve2 As I can remember I can do start sieving multiple sequence at once. Not it doesnot work. I enter few sequences in command line, but got output only for first one. Srsieve2 is from latest mtsieve package ver 1.6.8
2023-02-14, 10:15   #997
rebirther

Sep 2011
Germany

22·3·172 Posts

Quote:
 Originally Posted by pepi37 Question for srsieve2 As I can remember I can do start sieving multiple sequence at once. Not it doesnot work. I enter few sequences in command line, but got output only for first one. Srsieve2 is from latest mtsieve package ver 1.6.8
I would prefer to put all remaining sequences into a remain.txt and use it as inputfile.

2023-02-14, 10:41   #998
rogue

"Mark"
Apr 2003
Between here and the

3×2,357 Posts

Quote:
 Originally Posted by pepi37 Question for srsieve2 As I can remember I can do start sieving multiple sequence at once. Not it doesnot work. I enter few sequences in command line, but got output only for first one. Srsieve2 is from latest mtsieve package ver 1.6.8
Each sequence requires -s so if you have 5 sequences there should be 5 -s arguments each with a different sequence.

2023-02-14, 16:54   #999
storm5510
Random Account

Aug 2009
Not U. + S.A.

A4216 Posts

Quote:
 Originally Posted by rogue Each sequence requires -s so if you have 5 sequences there should be 5 -s arguments each with a different sequence.
Or put all the sequences in a single text file, each on one line. -s input.txt, for example. It worked for me.

Last fiddled with by storm5510 on 2023-02-14 at 16:56

2023-02-14, 23:36   #1000
pepi37

Dec 2011
After milion nines:)

11·149 Posts

Quote:
 Originally Posted by storm5510 Or put all the sequences in a single text file, each on one line. -s input.txt, for example. It worked for me.
Thanks Storm, first problem is solved

Now second one

Quote:
 srsieve2 -P 50000000000000 -W8 -i b10_n.abcd -O factors.txt -l 0 srsieve2 v1.6.8, a program to find factors of k*b^n+c numbers for fixed b and variable k and n Must use generic sieving logic because no memory is allocated for Legendre tables
If I use any other then 0 with -l option I got invalid factor. And if I read correctly -l build Legendre table so sieving with them should be faster?

And last, this is first time I got that on RTX 2070 Super sieving is 75% faster then on Ryzen 5700x ( 8 threads)

Last fiddled with by pepi37 on 2023-02-14 at 23:46

2023-02-15, 00:06   #1001
rogue

"Mark"
Apr 2003
Between here and the

3×2,357 Posts

Quote:
 Originally Posted by pepi37 Thanks Storm, first problem is solved Now second one If I use any other then 0 with -l option I got invalid factor. And if I read correctly -l build Legendre table so sieving with them should be faster? And last, this is first time I got that on RTX 2070 Super sieving is 75% faster then on Ryzen 5700x ( 8 threads)
Yes, it should be faster with -l > 0. Please PM me with the details for the invalid factor.

All times are UTC. The time now is 23:40.

Thu Mar 30 23:40:38 UTC 2023 up 224 days, 21:09, 1 user, load averages: 0.76, 0.81, 0.82