CLAPACK 3.2.1 is released

[peng.du]

CLAPACK 3.2.1 was released on Jun 25, 2009 at http://www.netlib.org/clapack/

Update from clapack3.2:
(1) a few bugs about the return type of smaxloc and dmaxloc are fixed
(2) A clapack.h file is included as part of the package

For function what's-new info, please refer to http://www.netlib.org/lapack/lapack-3.2.1.html.

Thanks,
Peng

[lucky0002]

Hello ,
I am experimenting something with the lapack.
I called the 'ssyev'( routine for calculating eigen values of symmetric matrix) , and linked with the lapack.lib, blas.lib, libf2c.lib ( the prebuilt binaries I downloaded).
When I run the program , for 128X128, and 256X256 matrix it takes very less time, less that 1 seconds.

But Other thing I tried is that , I have collected the C files( total no was arround 50) that are the depencencies along with the code for ssyev.
And now the performance is reduced.
for 128X128 matrix it will take 7.9 seconds
and for 256X256 matrix it takes arround 23 seconds.

Why there is so much difference?
It seems to me I might be mising the compiler flags, what are the other issues that c'd be responsible?

Please tell me the compiler flags/optimizations I should set while compiling those 50 C files to get the comparable performance.

[admin]

Those libraries were compiled with the /02 optimization flag.
Take a look here to see how it was built: http://icl.cs.utk.edu/lapack-for-window ... html#build
Julie

[lucky0002]

Actually I have used the same link to compile the libraries for my system.

But the thing is when I am using precompiled libraries ( compiled with /O2 flag) the runtime is less than 0.1 seconds.
And when I am compiling the source files( 50 C files that I collected ) with the same optimization flag, the runtime is very poor , as I indicated above.

What you say about this?