Page 1 of 1

Performance measurement for BLAS 3 sub-routine

PostPosted: Wed Jan 26, 2011 1:48 pm
by buggedup
Hello,

I want to measure the performance for the cholesky decomposition using dpotrf using C.
Can anyone tell me how can I calculate the flops for dpotrf. Or atleast how many flops are required for one dpotrf call.

BuggedUp BuggedUp BuggedUp

Re: Performance measurement for BLAS 3 sub-routine

PostPosted: Wed Jan 26, 2011 2:06 pm
by admin
Flops can be found on LAWN 41 page 120
Here is the link: http://www.netlib.org/lapack/lawnspdf/lawn41.pdf

From the document flops are:
1/3 n^3 + 1/2 n^2 + 1/6n

Hope it helps
Julie

Re: Performance measurement for BLAS 3 sub-routine

PostPosted: Wed Jan 26, 2011 4:11 pm
by buggedup
Hey...
Thanks for the reply.
I cud see on the page...it is given for SPOTRF...
But, will it be same for double precision matrix also..I mean for DPOTRF?

uggedUp BuggedUp

Re: Performance measurement for BLAS 3 sub-routine

PostPosted: Wed Jan 26, 2011 4:11 pm
by admin
yes it is the same.

Re: Performance measurement for BLAS 3 sub-routine

PostPosted: Wed Jan 26, 2011 4:13 pm
by buggedup
hey...
thanks for d help....

BuggedUp