CADO and WinBlows
 

We're currently in the process of improving portability of the [URL="http://cado-nfs.gforge.inria.fr/"]CADO NFS[/URL] tools, in particular we're trying to get them to compile under Windows. There are a bunch of annoying catches, however, and we're taking stock of the options. One option that was discussed was simply preparing a Linux environment for a Virtual Machine under which CADO could run, with minimal porting work on our part.

So we were wondering how much demand there actually is for CADO on Windows, and whether a VM solution would be considered acceptable.

If you are interested in running CADO on a Windows system, can you please tell which Windows version (incl. 32 or 64 bit) you are running, and whether running a Linux CADO in a VM would work for you?

I would use CADO for Windows 7 64-bit.

How good is CADO's siever in comparison to ggnfs's?

[QUOTE=henryzz;325664]How good is CADO's siever in comparison to ggnfs's?[/QUOTE]

It's under active development, which is a rather big plus. :razz: (In all seriousness, I'd love to see some numbers, like everyone else :razz:)

[QUOTE=Dubslow;325770]It's under active development, which is a rather big plus. :razz: (In all seriousness, I'd love to see some numbers, like everyone else :razz:)[/QUOTE]

I doubt it would beat the 64-bit asm but it might stand a chance against the 32-bit asm(which is all we have on windows).
Anyone got numbers?

I can run a comparison on Linux, maybe with polynomials for one of the current team sieve projects. The Windows build does not work yet (tho it appears we're getting closer, now it seems to be mostly a matter of parsing CRLF in the scripts correctly); once it works, I can post a comparison for 32-bit builds on Windows.

This [URL="http://www.mersenneforum.org/showthread.php?t=17695"]team sieve[/URL] could handle some extra relations. :smile: 1M for each siever would be a reasonable large-scale test. That's around two days on one core of an i5-2500, so perhaps splitting a 1M range among the various sievers would be better.

[QUOTE]So we were wondering how much demand there actually is for CADO on Windows…[/QUOTE]

[YOUTUBE]1mXeG5xyV2E[/YOUTUBE]

Hi
It would be nice to have a version for Windows 7 64 bit, I mean a native version, not a virtualized one.

TBH I would recommend making a full windows version now. Eventually I would guess that bits of the CADO tools will become a regular part of people's toolset. It would be better to convert it earlier while the project is simpler than later when it is more complex.
We don't want a situation like we have with the 64-bit ggnfs siever.

Is the Windows version going to use the GMP libraries or the MPIR ones?

[QUOTE=skan;330862]Is the Windows version going to use the GMP libraries or the MPIR ones?[/QUOTE]

It could presumably be linked with either. MPIR is designed to be a "drop in" replacement for GMP, meaning the function calls and data types are all the same.

Which they would use for a pre-compiled binary, I would hazard a guess at MPIR, since that's what Gilchrist/Gladman primarily use. (This is all hypothetical of course.)

[QUOTE=Dubslow;330896]It could presumably be linked with either. MPIR is designed to be a "drop in" replacement for GMP, meaning the function calls and data types are all the same.

Which they would use for a pre-compiled binary, I would hazard a guess at MPIR, since that's what Gilchrist/Gladman primarily use. (This is all hypothetical of course.)[/QUOTE]

At the moment I believe that mingw and mingw64 are being used so linking could be with either GMP or MPIR, which are (mostly) compatible.

GMP has recently caught up with MPIR in speed on Windows x64 by getting its x64 assembler to work with the Windows ABI. So the speed difference is now much reduced (and may even be reversed).

I am looking at the possibility of a CADO-NFS build with Visual Studio but this will probably follow if the mingw activity succeeds. If I do make a Visual Studio build, this will only be with MPIR as GMP doesn't build natively with Visual Studio (I believe that it may be possible to link with mingw/mingw64 built libraries).

Currently the test builds link against the GMP library provided by MinGW, although a separately compiled GMP should work just as well. We tested only MinGW32 so far, not MinGW64, so right now that performance of the siever, as a native Windows binary, is pretty bad.

Hi

Could you post any already compiled Windows binary, please?

regards

I would much rather prefer a windows version and I have always regarded virtual machines as much less efficient, but recently I've heard that VMware Workstation can run with relatively small overhead. Since I can't get CADO-NFS to compile currently, I'm left with either a VM or dual booting.

[QUOTE=jux;420416]I would much rather prefer a windows version and I have always regarded virtual machines as much less efficient, but recently I've heard that VMware Workstation can run with relatively small overhead. Since I can't get CADO-NFS to compile currently, I'm left with either a VM or dual booting.[/QUOTE]

I have used virtualbox in the past and it ran at pretty much the same speed as linux.
My problem is that with only 4GB of memory I don't have any spare for a virtual machine.

If you can, I would highly recommend buying and installing more RAM (or even [URL="http://downloadmoreram.com/index.php"]download it[/URL]). Memory these days is very cheap and your computer will benefit overall.

[QUOTE=jux;420453]If you can, I would highly recommend buying and installing more RAM (or even [URL="http://downloadmoreram.com/index.php"]download it[/URL]). Memory these days is very cheap and your computer will benefit overall.[/QUOTE]

Please forgive me for this, but this seems like a scam to me.

One cannot "download RAM". RAM stands for "Random Access Memory", and RAM needs to be ***very*** close to the processors for effectiveness.

Could you please explain yourself?

[QUOTE=chalsall;420455]Please forgive me for this, but this seems like a scam to me.

One cannot "download RAM". RAM stands for "Random Access Memory", and RAM needs to be ***very*** close to the processors for effectiveness.

Could you please explain yourself?[/QUOTE]

It's a well known joke. The site he linked is a parody of such scams.

[QUOTE=Dubslow;420466]It's a well known joke. The site he linked is a parody of such scams.[/QUOTE]

Colour me embarrassed.... :blush:

[QUOTE=chalsall;420469]Colour me embarrassed.... :blush:[/QUOTE]

I would call it a case of cultural ignorance :smile: I certainly suffered from an extreme lack of exposure to things basically all my peers had heard of ("I was sheltered"), so I can sympathize.

In this case: there are surely as many computer scams as there are computers, and it just so happens that the intersection of computer/computer literate culture and internet culture settled on this particular example as the archetype of such things, and built it into the "in joke" that it is now. It is perhaps somewhat surprising that you hadn't heard of this cultural phenomenon, though as I said, I can very definitely sympathize :smile: no big deal, at the end of the day.

I would if I wasn't stuck with a ddr2 PC for the time being. I plan on upgrading the whole thing next year.