L_work and H_work

Open discussion for MAGMA library (Matrix Algebra on GPU and Multicore Architectures)
Post Reply
clattenb
Posts: 8
Joined: Mon Apr 29, 2013 11:46 pm

L_work and H_work

Post by clattenb » Mon May 06, 2013 11:32 am

Hello,

I'm hoping someone can give me a quick explanation of what the common arguments of L_work and H_work are for. I see these arguments in most of the functions (I am only needing to use gels and gesvd) and I just want to have a better understanding of what they're for beyond what is in the testing .cpp files.

Any help is appreciated.

Thanks,
-Nick

mgates3
Posts: 918
Joined: Fri Jan 06, 2012 2:13 pm

Re: L_work and H_work

Post by mgates3 » Tue May 07, 2013 1:50 pm

Most algorithms need some temporary work space. In GELS, that is the array hwork, which has size (dimension) lwork. The exact need varies with the algorithm. For instance, it might be used to store a panel (block column) of A on the CPU, while most of A is stored on the GPU.
-mark

clattenb
Posts: 8
Joined: Mon Apr 29, 2013 11:46 pm

Re: L_work and H_work

Post by clattenb » Tue May 07, 2013 8:16 pm

Hi Mark,

Thanks for that explanation, I'm new to these linear algebra libraries (I have been using my own basic library up till now). For the two functions that I'm using (gels and gesvd) would you recommend that I just use the same formula that's in the code of the respective test application (such as testing_ dgeqrs3_gpu.cpp and testing_dgesvd.cpp)?

Thanks again,
-Nick

Post Reply