srsieve/sr2sieve enhancements

[rogue]

Originally Posted by pepi37

Any of this three?
Cullen/Woodall Prime Search (Sieve)
The Riesel Problem (Sieve)
Proth Prime Search (Sieve)

There are GPU apps out there already for these. srsieve/sr2sieve are not GPU apps. I suggest you contact one of the ppsieve developers.

[pepi37]

You are wrong; there is no GPU apps fro The Riesel Problem (Sieve)

[rogue]

Originally Posted by pepi37

You are wrong; there is no GPU apps fro The Riesel Problem (Sieve)

As I understand it, ppsieve can be used for sieving Riesels (-R option), although this isn't overly helpful for TRP. Nevertheless, it would make more sense to port sr1sieve/sr2sieve to the GPU than to BOINCify them. I would like to do that someday, but I've been too busy to work on it.

[f1pokerspeed]

Does anyone have a Win32 binary for srsieve that I could obtain please? I have no idea of what I would need to compile the source with.

[VBCurtis]

If you google "srsieve", you'll quickly find the Google page that Geoff Reynolds has set up for this purpose. There are binaries available there.
-Curtis

[henryzz]

I don't know whether you are interested in finding the 64-bit bug in srsieve. I have just found a test case. The stuff required to reproduce is in the attached zip.

edit: I haven't tried to reproduce, but srsieve should be deterministic I think. It's the 64-bit binary you included with 1.0.4.

[rogue]

Originally Posted by henryzz

I don't know whether you are interested in finding the 64-bit bug in srsieve. I have just found a test case. The stuff required to reproduce is in the attached zip.

edit: I haven't tried to reproduce, but srsieve should be deterministic I think. It's the 64-bit binary you included with 1.0.4.

I think it is a miscompile on Windows as I can't reproduce this on a Mac. I'll look at it further tomorrow.

[henryzz]

Originally Posted by rogue

I think it is a miscompile on Windows as I can't reproduce this on a Mac. I'll look at it further tomorrow.

It was alwaya a 64-bit windows bug. There is a warning in warning.txt. If you compile with an old enough version of gcc it disappears. I had assumed the bug fixed although the 64-bit version does check that factors are actually factors.
Sorry for not specifying windows previously.

[rogue]

Originally Posted by henryzz

It was alwaya a 64-bit windows bug. There is a warning in warning.txt. If you compile with an old enough version of gcc it disappears. I had assumed the bug fixed although the 64-bit version does check that factors are actually factors.
Sorry for not specifying windows previously.

I use mingw64. I'll get the latest and try with that.

[rogue]

Try this one. I built it with the latest mingw64, but haven't tested it.

[henryzz]

Originally Posted by rogue

Try this one. I built it with the latest mingw64, but haven't tested it.

I just ran the sieve file in the attachment passed the error and it ran fine. However, a subsequence had been removed so the test wasn't exactly the same. I am running the sieving job again with the new binary.
The faulty binary missed several factors before and after 4302594733 | 20097212*7^23221-1 in addition to having the error. It looks like the 64-bit sieve was completly messed up. Strangely it did manage to find that one factor before finding an error.