i`m new to cuda , but i`m trying to learn some new stuff ,
i found this post and i followed the steps , but my text isn`t highlighted .

i am using ms visual studio 2008 prof . i ve installed the gpu driver (gt 320M) , the cuda toolkit (3.1 i think) , the cuda sdk and ran the bandwidthTest which worked and also worked to build and run the bandwidthTest project, too , but the bandwidthTest.cu text in visual studio is still all black … even after i tried doing what you said right here :

- copied the usertype.dat file to \IDE folder
- added the .cu extension (…and selected the C++ in the drop down menu)
- selected the CUDA rules from the Custom Build Rules…
- and done the thing with the regedit , even though that is for the intellisense

if i missed something , please point it out for me …

thanks in advance .

//adrian

The above post is rather old so if you are using the spreadsheet for occupancy calculations consider switching to the visual profiler.

The number of registers used is decided by the compiler, but you can change the upper bound with the -maxrregcount compiler switch. Be careful with this and check the /cubin file produced as it may make the code more convoluted or start using local memory. “Local” memory incurs the same performance hit as global mem so it will dramatically slow your kernels down. As for shared memory you can specify the amount used in your kernels by specifying a bound in your array.

Hope this helps.
/Barrett

I’m glad the original article helped.

It may be a good idea to install the same version of the toolkit and SDK as I think some of the new examples depend on the new toolkit.

Does your practice.cu file have a custom build rule associated with it in your project? I’ve also found that if you include a .cu file from another one make sure the included one if excluded from the build - nvcc will read and compile it.

I’ve only got the PRJ0019 error when one of my paths was incorrect. Or possibly missing “” around it.

Lastly emulation mode is being deprecated (removed) in future releases. This does make testing on windows a bit different but then I suppose NVidia wants everyone to switch to Nexus ( why win7 only?? ). Another good reason could be that the emulator and non-emulator usually produce different outputs and led many a developer to go on a wild goose chase.

By the way printf does work from cuda code now - so useful for debugging if you don’t have Nexus.

Hope this help.
/Barrett

Have a look at the Gaussian convolution code available at
http://www.bv2.co.uk/?page_id=863

It gives a fairly optimal solution of the first step in the process.

/Barrett

Without actually seeing the results of the individual steps of your Canny Edge I can only guess at possible problems.

Is your Gaussian blur code correct? Ie dividing by the correct divisor for your mask - a rather easy mistake to make if you hand key your masks in.

Have you tried changing your Gausian blur parameters? Ie changing the width and std deviation of your mask? The example I gave in the article works in a wide variety of cases but may not be perfect for you.

It is quite easy to make an error in the non-maximum suppression code. The one mentioned in the wikipedia article should work, but there are other ways of doing it. Basically you are looking for a local maximum in the gradient direction you determined in the Canny step. The direction is the angle obtained by: theta = arctan (Gy/Gx). So if theta is 0 degrees see if the intensity of the pixels above and below are greater than or less than, if both less than we have a local maximum in the correct direction.

Hope this helps you a bit.
/Barrett

I currently only have vs2008 installed. But the installation of CUDA into vs2003 should be similar with the possible exception of the custom build rules.
You may need to set the .cu’s to be excluded from the build and then compile the project with nvcc.

Sorry this is not terribly helpful but I havent used vs2003 in a while.

/Barrett