[QUOTE=xilman;268883]I've just started looking into exactly the same problem but to build under Fedora 15.

There is a lot of chat on the CUDA and Linux fora about this issue. The summary seems to be that the dependency of gcc is for some debugging structures generated by old gcc and understood by Nvidia's tools. Somewhere in the code (perhaps scripts or Makefiles) there is a #error statement (or statements) which can be safely removed and everything then builds.

Sorry not to be more precise but the material above should give you enough to start looking for the details. Alternatively, you can wait until I've managed to get CUDA reinstalled on my system after its upgrade.[/QUOTE]Miserable failure here :sad:

I can't even install the device driver because the installer checks for elderly versions of the kernel and various crude hacks failed to work.

I've posted to the Cuda-on-Linux forum at Nvidia and hope someone there will be able to help.

Paul

This is a little bit wierd to me. I had no difficulty with installing the whole nine yards earlier this year under xubuntu 10.10, 64-bit version.

[QUOTE=xilman;268894]Miserable failure here :sad:

I can't even install the device driver because the installer checks for elderly versions of the kernel and various crude hacks failed to work.

I've posted to the Cuda-on-Linux forum at Nvidia and hope someone there will be able to help.

Paul[/QUOTE]
Have you tried to install the 275.21 or the 280.13 drivers? The kernel version check has been modified after the release of 270.41.19.

[QUOTE=Ralf Recker;268901]Have you tried to install the 275.21 or the 280.13 drivers? The kernel version check has been modified after the release of 270.41.19.[/QUOTE]No, but thanks for the tip. I'll give it a try.

The standard download for V4 appears to be the 270.40.xx release and I hadn't come across the 280.13 version.

Paul

[QUOTE=Christenson;268895]This is a little bit wierd to me. I had no difficulty with installing the whole nine yards earlier this year under xubuntu 10.10, 64-bit version.[/QUOTE]

I bet because 10.10 had an older version of gcc. I'm using 11.04.

[QUOTE=xilman;268902]No, but thanks for the tip. I'll give it a try.

The standard download for V4 appears to be the 270.40.xx release and I hadn't come across the 280.13 version.

Paul[/QUOTE]AFAICT, 280 is not yet available for 64-bit Linux. About to give the 275 beta a try.

Paul

[QUOTE=xilman;269059]AFAICT, 280 is not yet available for 64-bit Linux. About to give the 275 beta a try.

Paul[/QUOTE]

[B][SIZE=1]Linux x64 (AMD64/EM64T) Display Driver[/SIZE][/B]
[SIZE=1]
[/SIZE][B][SIZE=1]Version:[/SIZE][/B]

[B][SIZE=1]280.13 Certified[/SIZE][/B]

[B][SIZE=1]Release Date:[/SIZE][/B]

[B][SIZE=1]2011.08.01[/SIZE][/B]

[B][SIZE=1]Operating System:[/SIZE][/B]

[B][SIZE=1]Linux 64-bit[/SIZE][/B]

[B][SIZE=1]Language:[/SIZE][/B]

[B][SIZE=1]English (U.S.)[/SIZE][/B]

[B][SIZE=1]File Size:[/SIZE][/B]

[B][SIZE=1]52.4 MB [/SIZE][/B]

[URL]http://www.nvidia.com/object/linux-display-amd64-280.13-driver.html[/URL]

Deutsche Version

[B][SIZE=1]Linux x64 (AMD64/EM64T) Display Driver[/SIZE][/B]
[B][SIZE=1]
Version:[/SIZE][/B]

[B][SIZE=1]280.13 Certified[/SIZE][/B]

[B][SIZE=1]Freigabedatum:[/SIZE][/B]

[B][SIZE=1]2011.08.01[/SIZE][/B]

[B][SIZE=1]Betriebssystem:[/SIZE][/B]

[B][SIZE=1]Linux 64-bit[/SIZE][/B]

[B][SIZE=1]Sprache:[/SIZE][/B]

[B][SIZE=1]Deutsch[/SIZE][/B]

[B][SIZE=1]Dateigröße:[/SIZE][/B]

[B][SIZE=1]52.4 MB [/SIZE] [/B]

[URL]http://www.nvidia.de/object/linux-display-amd64-280.13-driver-de.html[/URL]

[QUOTE=moebius;269062][B][SIZE=1]Linux x64 (AMD64/EM64T) Display Driver[/SIZE][/B]
[SIZE=1]
[/SIZE][B][SIZE=1]Version:[/SIZE][/B]

[B][SIZE=1]280.13 Certified[/SIZE][/B][/QUOTE]

LHQL to the rescue! :big grin:

[QUOTE=ckdo;269065]LHQL to the rescue! :big grin:[/QUOTE]

Does it work with CUDA 3.X (especially 3.0)?

Having a GTX275 I can't access to 2.0 capabilities delivered by CUDA 4.0.

Luigi

Two questions: is there any work to getting mfaktc CPU independent?

And two, more complicated: I have 4 cores, 8 threads. I have 3 P95 workers running on two threads apiece, occupying logical CPU's 3-8. I had the last worker running only on logical CPU 1, leaving 2 empty for mfaktc. When I started mfaktc, it started taking cycles from all of the other CPU's, affecting their performance and not making use of logical thread 2. Is there anyway to set mfaktc onto one particular core/thread?

[QUOTE=ET_;269068]Does it work with CUDA 3.X (especially 3.0)?
Having a GTX275 I can't access to 2.0 capabilities delivered by CUDA 4.0.[/QUOTE]

You can use CUDA 4.0 on your GTX 275, you just can't take advantage of the fancy new features that require CC 2.x. I'm stuck in the same boat with the Tesla S1070.

But yes, the 4.0 driver will run code compiled with 3.x.