SVD algorithm

Support for issues specific to the Windows operating systems.

SVD algorithm

Postby linus » Thu Feb 25, 2010 5:41 pm

Is there a "minimum version" of a CUDA enabled device that we need to use for getting correct results for the SVD algorithm?

I am running the gesvd.exe example as is with the cullPercentage set to 0.0 to check if the exact original image is reproduced. But this is not the case.

Can anyone throw some light on this? Am I missing something?

Posts: 3
Joined: Thu Feb 25, 2010 4:48 pm

Re:SVD algorithm

Postby linus » Thu Feb 25, 2010 5:43 pm

FYI, This is my graphics card...I know its not the best :)

Device 0: "Quadro NVS 295"
CUDA Driver Version: 2.30
CUDA Runtime Version: 2.30
CUDA Capability Major revision number: 1
CUDA Capability Minor revision number: 1
Total amount of global memory: 268435456 bytes
Number of multiprocessors: 1
Number of cores: 8
Total amount of constant memory: 65536 bytes
Total amount of shared memory per block: 16384 bytes
Total number of registers available per block: 8192
Warp size: 32
Maximum number of threads per block: 512
Maximum sizes of each dimension of a block: 512 x 512 x 64
Maximum sizes of each dimension of a grid: 65535 x 65535 x 1
Maximum memory pitch: 262144 bytes
Texture alignment: 256 bytes
Clock rate: 1.30 GHz
Concurrent copy and execution: No
Run time limit on kernels: No
Integrated: No
Support host page-locked memory mapping: No
Compute mode: Default (multiple host threads
can use this device simultaneously)
Posts: 3
Joined: Thu Feb 25, 2010 4:48 pm

Re:SVD algorithm

Postby kyle » Fri Feb 26, 2010 10:38 am

Hi Linus,

There is no "minimum version" needed to get a correct answer. If there was a GPU related issue (out of memory, double precision unsupported, etc), the demo will catch and report the error.

The non-perfect results are simply the effects of using a simple image compression algorithm on the image. This demo was created to show non-trivial usage of a CULA routine; not to demonstrate a robust image compression method.

If you'd like to truly test the accuracy of the SVD, I'd suggest trying to test the orthogonality of U and VT (I=U*U', I=VT'*VT) as well as the reconstruction of A (A=U*S*VT). For these tests you should see very good results!

I hope this helps,

Posts: 301
Joined: Fri Jun 12, 2009 7:47 pm

Re:SVD algorithm

Postby linus » Fri Feb 26, 2010 11:18 am

Hi Kyle,

That's great!!! Makes perfect sense.

I wasn't thinking of the image compression algorithm. I was only thinking in terms of "I should get back what I started from if I don't remove anything".

I might try out a simple least squares fit. Will let you know how it goes.

Thanks again,
Posts: 3
Joined: Thu Feb 25, 2010 4:48 pm

Return to Windows Support

Who is online

Users browsing this forum: No registered users and 1 guest