I’m new in GPGPU programming and besides i’m almost new with VS 2010 - so, newbie i’m!
Ok, what’s the problem:
Pair of 470s in SLI with latest drivers installed - ok
Installed VS2010 - ok
Installed CUDA Toolkit 3.2.7 - ok
Installed Computing SDK 3.2 - ok (ran samples from browser, looks fantastic)
CUDA VS Wizard didn’t help (only for vs 2008)
Hmm… I can’t see any CUDA projects in Project menu, but i’ve checked $(CUDA_INC_PATH) and other variables - they’re ok
I’m new in GPGPU programming and besides i’m almost new with VS 2010 - so, newbie i’m!
Ok, what’s the problem:
Pair of 470s in SLI with latest drivers installed - ok
Installed VS2010 - ok
Installed CUDA Toolkit 3.2.7 - ok
Installed Computing SDK 3.2 - ok (ran samples from browser, looks fantastic)
CUDA VS Wizard didn’t help (only for vs 2008)
Hmm… I can’t see any CUDA projects in Project menu, but i’ve checked $(CUDA_INC_PATH) and other variables - they’re ok
I am also trying to get CUDA working in Visual Studio 2010. I am working on Windows Server 2008. I tried looking for some articles/steps on how to do this,
but could not find it.
I have installed CUDA Toolkit v3.2, CUDA SDK 3.2 and Visual Studio 2010 Premium. It seems v3.2 should support VS 2010 (mentioned here:
I want to get CUDA working with this combination. I have Tesla M1060 card. The binaries distributed with the SDK 3.2 work fine, but compiling the code in
VS 2010 is a problem. If I try to open the *.sln file from src, then VS 2010 tries to convert it and fails to do so.
If anyone has got CUDA working with VS 2010 before, please let us know how to do it.
I am also trying to get CUDA working in Visual Studio 2010. I am working on Windows Server 2008. I tried looking for some articles/steps on how to do this,
but could not find it.
I have installed CUDA Toolkit v3.2, CUDA SDK 3.2 and Visual Studio 2010 Premium. It seems v3.2 should support VS 2010 (mentioned here:
I want to get CUDA working with this combination. I have Tesla M1060 card. The binaries distributed with the SDK 3.2 work fine, but compiling the code in
VS 2010 is a problem. If I try to open the *.sln file from src, then VS 2010 tries to convert it and fails to do so.
If anyone has got CUDA working with VS 2010 before, please let us know how to do it.
I am also trying to get CUDA working in Visual Studio 2010. I am working on Windows Server 2008. I tried looking for some articles/steps on how to do this,
but could not find it.
I have installed CUDA Toolkit v3.2, CUDA SDK 3.2 and Visual Studio 2010 Premium. It seems v3.2 should support VS 2010 (mentioned here:
I want to get CUDA working with this combination. I have Tesla M1060 card. The binaries distributed with the SDK 3.2 work fine, but compiling the code in
VS 2010 is a problem. If I try to open the *.sln file from src, then VS 2010 tries to convert it and fails to do so.
If anyone has got CUDA working with VS 2010 before, please let us know how to do it.
I am also trying to get CUDA working in Visual Studio 2010. I am working on Windows Server 2008. I tried looking for some articles/steps on how to do this,
but could not find it.
I have installed CUDA Toolkit v3.2, CUDA SDK 3.2 and Visual Studio 2010 Premium. It seems v3.2 should support VS 2010 (mentioned here:
I want to get CUDA working with this combination. I have Tesla M1060 card. The binaries distributed with the SDK 3.2 work fine, but compiling the code in
VS 2010 is a problem. If I try to open the *.sln file from src, then VS 2010 tries to convert it and fails to do so.
If anyone has got CUDA working with VS 2010 before, please let us know how to do it.
i’ve had to struggle through this process. i believe i’ve got it.
firstly, make sure you have the developer driver. (though i develop on my laptop, too, which doesn’t have an nvidia card actually, so i just dont’ fully compile / run it there.)
the main problem is that the sdk doesn’t include a solution for 2010, and when you try to convert for a 2008 / 2005 file, it errors out on NvRuntimeApi or something. the file really isn’t there. so i decided to try editing it out of the .vcproj or whatever file. it turns out it’s just xml and you can just remove it. so i did. and then it was able to convert successfully.
i think there was still stuff i had to mess around with. most important step was right-clicking on the project in solution explorer and selecting build customizations…, and checking cuda 3.2.targets.
also you might have to the compile cutil and shrutil projects first.
eventually i discovered something much simpler, though:
the parrallel nsight program actually comes with its own set of samples that actually work pretty much write off the bat in 2010 (they have 2010 solution files). so download and install it, (you’ll need both the monitor and the host), then on your start menu go to nvidia parallel nsight 1.5->browse cuda samples, then open one of those solution files and it should be able to compile.
then i suppose you can copy the settings from those working ones to new ones. that’s what i intend to do. make sure if you’re using win32 vs. x64, you’re using it all across the board. (the xx32.dll’s, etc.)
i’ve had to struggle through this process. i believe i’ve got it.
firstly, make sure you have the developer driver. (though i develop on my laptop, too, which doesn’t have an nvidia card actually, so i just dont’ fully compile / run it there.)
the main problem is that the sdk doesn’t include a solution for 2010, and when you try to convert for a 2008 / 2005 file, it errors out on NvRuntimeApi or something. the file really isn’t there. so i decided to try editing it out of the .vcproj or whatever file. it turns out it’s just xml and you can just remove it. so i did. and then it was able to convert successfully.
i think there was still stuff i had to mess around with. most important step was right-clicking on the project in solution explorer and selecting build customizations…, and checking cuda 3.2.targets.
also you might have to the compile cutil and shrutil projects first.
eventually i discovered something much simpler, though:
the parrallel nsight program actually comes with its own set of samples that actually work pretty much write off the bat in 2010 (they have 2010 solution files). so download and install it, (you’ll need both the monitor and the host), then on your start menu go to nvidia parallel nsight 1.5->browse cuda samples, then open one of those solution files and it should be able to compile.
then i suppose you can copy the settings from those working ones to new ones. that’s what i intend to do. make sure if you’re using win32 vs. x64, you’re using it all across the board. (the xx32.dll’s, etc.)
Thanks a lot !! I got CUDA working with VS 2010. I had to Install VS 2008 and use its cl.exe compiler. Initially, I installed Parallel NSight and opened its sample CUDA program from its SDK in VS 2010 and tried to compile it. But it said that Visual Studio 2008 is reqd. After I installed VS 2008, it compiled successfully. I also tried another program using the setting from Parallel NSight and that worked too. Thank you very much !!
Thanks a lot !! I got CUDA working with VS 2010. I had to Install VS 2008 and use its cl.exe compiler. Initially, I installed Parallel NSight and opened its sample CUDA program from its SDK in VS 2010 and tried to compile it. But it said that Visual Studio 2008 is reqd. After I installed VS 2008, it compiled successfully. I also tried another program using the setting from Parallel NSight and that worked too. Thank you very much !!
the 2010 compiler (v100) is not supported yet. only the 2008 one (v90). that’s actually explicitly checked by the build customization file (CUDA 3.2.targets). there’s a setting in projects settings somewhere on what compiler version to use. you just need to make sure it’s set to v90 instead of v100. that way you needn’t bother manualy specifying a different compiler. and i think 2010 comes w/a v90 compiler so you shouldn’t have to install 2008, either. (though if i’m wrong someone please correct me.)
anyways, glad to hear i was helpful! 'twas quite the struggle for me. i’m happy i could save others the same work.
the 2010 compiler (v100) is not supported yet. only the 2008 one (v90). that’s actually explicitly checked by the build customization file (CUDA 3.2.targets). there’s a setting in projects settings somewhere on what compiler version to use. you just need to make sure it’s set to v90 instead of v100. that way you needn’t bother manualy specifying a different compiler. and i think 2010 comes w/a v90 compiler so you shouldn’t have to install 2008, either. (though if i’m wrong someone please correct me.)
anyways, glad to hear i was helpful! 'twas quite the struggle for me. i’m happy i could save others the same work.
Yeah, samples from nsight work correctly but still i’ve a mess with VS2010 and CUDA (didn’t managed targets file…)
And still, only v90 compiler supported yet. I haven’t been able to enable (able to enable, heh) correct work of CUDA in VS2010, so I decided to jump back to VS2008
Exact path for CUDA + VS2008 looks simple:
VS2008 (default)
SDK (default)
Toolkit (default)
CUDA VS Wizard 2.2
So, I’ll try in VS2008 for now, maybe VS2010 will get in hands later.
Yeah, samples from nsight work correctly but still i’ve a mess with VS2010 and CUDA (didn’t managed targets file…)
And still, only v90 compiler supported yet. I haven’t been able to enable (able to enable, heh) correct work of CUDA in VS2010, so I decided to jump back to VS2008
Exact path for CUDA + VS2008 looks simple:
VS2008 (default)
SDK (default)
Toolkit (default)
CUDA VS Wizard 2.2
So, I’ll try in VS2008 for now, maybe VS2010 will get in hands later.