No Speedup Using dsyr on Core 2 Duo Running Linux

[Lawrence_Austen]

Hi,

My code is along the lines of the following. The dimensions are A(500,500) x(500)

for ( int i = 0 ; i < 100000 ; i++)
cblas_dsyr(...A,x,...)
cout << A << endl ;

I was hoping to get a nice speedup running on my multicore Core 2 Duo (T7100) laptop, but 1 core is at 100% and no processing at all is being done with the second core. I've set both cores for maximum performance and I've used the MKL Link Line Advisor. and I've tried many variants of the various environment variables (MKL_NUM_THREADS, etc.).

Any suggestions on how I can get some speedup out of cblas_dsyr on multiple cores?

Thanks,
Lawrence

[Lawrence_Austen]

Quoting - Lawrence Austen

Hi,

I cranked the dimension of my problem up to 5000 and I get a nearly perfect speedup. So I guess this is just a memory bottlenecking type problem.

Cheers,
Lawrence


Hi,

My code is along the lines of the following. The dimensions are A(500,500) x(500)

for ( int i = 0 ; i < 100000 ; i++)
cblas_dsyr(...A,x,...)
cout << A << endl ;

I was hoping to get a nice speedup running on my multicore Core 2 Duo (T7100) laptop, but 1 core is at 100% and no processing at all is being done with the second core. I've set both cores for maximum performance and I've used the MKL Link Line Advisor. and I've tried many variants of the various environment variables (MKL_NUM_THREADS, etc.).

Any suggestions on how I can get some speedup out of cblas_dsyr on multiple cores?

Thanks,
Lawrence