srsieve/sr2sieve enhancements

[rogue]

Originally Posted by storm5510

I use only srsieve and sr1sieve. Their parameters are similar. Sr2sieve does a good job, but do not use it because I never work more than once sequence at a time. If sr1sieve could read an ABC2 (PFGW) format, like the example below, then I would use it exclusively. I use srsieve just long enough to create an input file for sr1sieve.

ABC2 98475*2^$b-1
b: from 770000 to 1300000

sr1sieve (by design) can only sieve for p > max k and possibly p > max n. I don't recall off the top of my head. This is one of drivers for its speed.

In some future state, srsieve2 will "do it all", but logic in sr1sieve and sr2sieve is incredibly convoluted so bringing it into srsieve2 is a massive undertaking. Each time I start I only get so far before I give up. This is mainly due to how I want to do the optimization, which is very different than the current logic being used.

[storm5510]

Originally Posted by rogue

sr1sieve (by design) can only sieve for p > max k and possibly p > max n. I don't recall off the top of my head. This is one of drivers for its speed.

In some future state, srsieve2 will "do it all", but logic in sr1sieve and sr2sieve is incredibly convoluted so bringing it into srsieve2 is a massive undertaking. Each time I start I only get so far before I give up. This is mainly due to how I want to do the optimization, which is very different than the current logic being used.

I am quite satisfied to use srsieve and sr1sieve as they are.

As I wrote before, srsieve2 needs to write a new line below the existing line, not replace it. Having a programming background of my own, this would seem a very minor modification.

[rogue]

Originally Posted by storm5510

I am quite satisfied to use srsieve and sr1sieve as they are.

As I wrote before, srsieve2 needs to write a new line below the existing line, not replace it. Having a programming background of my own, this would seem a very minor modification.

I thought that I fixed that. Please check mtsieve framework 1.9.7. If I haven't please let me know.

[storm5510]

Originally Posted by rogue

I thought that I fixed that. Please check mtsieve framework 1.9.7. If I haven't please let me know.

I owe rogue an apology. The behavior I noted was in Windows PowerShell. srsive2 does produce new lines in a Command Prompt console. I have had other programs that did not play well in PowerShell, PFGW for one. If Microsoft wants to push PowerShell as a replacement for Command Prompt, then any console program needs to behave the same as it does in Command Prompt.

srsieve has an option to set a screen update interval, in seconds. As far as I can tell, srsieve2 does not have this. It would be nice if it did.

[rogue]

Originally Posted by storm5510

I owe rogue an apology. The behavior I noted was in Windows PowerShell. srsive2 does produce new lines in a Command Prompt console. I have had other programs that did not play well in PowerShell, PFGW for one. If Microsoft wants to push PowerShell as a replacement for Command Prompt, then any console program needs to behave the same as it does in Command Prompt.

srsieve has an option to set a screen update interval, in seconds. As far as I can tell, srsieve2 does not have this. It would be nice if it did.

What does pfgw do in PowerShell differently than in the command prompt?

srsieve2 is set to 60 seconds between refreshes. I don't have a switch to change it because usage of that switch would be rare (as far as I know). I can't think of any cases where I have done sieving for less than 60 seconds outside of testing code changes. It would also impact the entire mtsieve framework. One of the things about mtsieve is that I want to keep usage of the framework (and its switches) simple and consistent.

[storm5510]

Originally Posted by rogue

What does pfgw do in PowerShell differently than in the command prompt?

It will overwrite an existing output line. This is in Windows 7's PowerShell. I have not tried it with Windows 10.

"The mfsieve framework." This is part of BOINC?


Update: The behavior is the same in Windows 10 Powershell.

[rogue]

Originally Posted by storm5510

It will overwrite an existing output line. This is in Windows 7's PowerShell. I have not tried it with Windows 10.

"The mfsieve framework." This is part of BOINC?


Update: The behavior is the same in Windows 10 Powershell.

No. mtsieve has nothing to do with BOINC.

Can you explain what you mean by "overwrite and existing output line"? All of my sieves and pfgw can overwrite the current output line and is the default behavior. There are some cases where it will do a new line. I'm trying to understand your expectations.

[storm5510]

Originally Posted by rogue

...Can you explain what you mean by "overwrite and existing output line"? All of my sieves and pfgw can overwrite the current output line and is the default behavior. There are some cases where it will do a new line. I'm trying to understand your expectations.

Think of overwrite as "replaced by." Example: I have 1000 on the screen. 1000 changes to 1001 in the same location instead of dropping down to the next line below. Prime95 does this. Each new item is placed below the existing one.

"Expectations." I do not have any. There is one item: Suppose I start a sieve with srsieve2 which will take 36 hours to complete. There is a power outage with only 4 hours left. As far as I can tell, everything would be lost because of no intermittent saving of the of the results. Is there an interval written into the code?

[rogue]

Originally Posted by storm5510

Think of overwrite as "replaced by." Example: I have 1000 on the screen. 1000 changes to 1001 in the same location instead of dropping down to the next line below. Prime95 does this. Each new item is placed below the existing one.

"Expectations." I do not have any. There is one item: Suppose I start a sieve with srsieve2 which will take 36 hours to complete. There is a power outage with only 4 hours left. As far as I can tell, everything would be lost because of no intermittent saving of the of the results. Is there an interval written into the code?

For pfgw, try -Cverbose. It will put the results of each PRP test on their own line. By default most users do not want to see intermediate outputs written to their own line as it just clutters the window with mostly useless information and one some systems will fill the command window buffer causing you to lose more important information. srsieve is one of few programs that does not work this way.

For all of the mtsieve programs, it will write to the output file once per hour. It is possible to lose up to just under 60 minutes of sieving time.

[storm5510]

Originally Posted by rogue

For pfgw, try -Cverbose. It will put the results of each PRP test on their own line. By default most users do not want to see intermediate outputs written to their own line as it just clutters the window with mostly useless information and one some systems will fill the command window buffer causing you to lose more important information. srsieve is one of few programs that does not work this way.

For all of the mtsieve programs, it will write to the output file once per hour. It is possible to lose up to just under 60 minutes of sieving time.

Look at the attached image. This is two instances of LLR stacked on the screen. I suppose I am different because I do not consider everything there as useless. It is the same with sr1sieve. I like to see what it has done and what it is doing at the same time. I started using computers in 1988 when there were few options about screen content. I became used to it.

I started out using PFGW for Riesel searches. It was rapidly pointed out that I needed to be using LLR. Probabilistic versus deterministic. I rarely use PFGW now.

I solved my own problem with srsieve2 by not using PowerShell. Writing to an output file every 60 minutes is good.

You refer to mfsieve as a group of programs. Where can I find them?

[kuratkull]

Originally Posted by storm5510

Look at the attached image. This is two instances of LLR stacked on the screen. I suppose I am different because I do not consider everything there as useless. It is the same with sr1sieve. I like to see what it has done and what it is doing at the same time. I started using computers in 1988 when there were few options about screen content. I became used to it.

I started out using PFGW for Riesel searches. It was rapidly pointed out that I needed to be using LLR. Probabilistic versus deterministic. I rarely use PFGW now.

I solved my own problem with srsieve2 by not using PowerShell. Writing to an output file every 60 minutes is good.

You refer to mfsieve as a group of programs. Where can I find them?

Printing Carriage Return instead of the Newline symbol does that to a line.
Also I wouldn't say "Probabilistic vs Deterministic" is the reason we use the LLR test vs the Fermat test. The LLR test is more optimized for that problem - it needs fewer multiplications to reach a result. For a 100k digit number the Fermat test has 1.3*10^-10584 chance of a false positive - it only gets more unrealistic from there [https://primes.utm.edu/notes/prp_prob.html].