### Using DGEEV CULA for small matrices burst mode

Posted:

**Thu Jan 26, 2012 8:20 am**I looked at your benchmarks for Non-Symmetric Eigenvalue Routines and I have a question regarding the matrix sizes. What's the performance comparison for matrices smaller than 512?

I am interesting in computing dgeev for a 10x10 matrix in burst mode.

1. Does CULA support burst mode?

2. How does CULA compare to MKL and magma-1.1.0?

magma-1.1.0 performs very poorly for matrix sizes below 1024x1024. BUt there's no mention of burst mode.

I could not find any CULA documentation on burst mode for small matrices. Please let me know if this feature is added in R13.

Also, my organization bought the license for CULA r12, is R13 a free upgrade?

Thanks in advance.

I am interesting in computing dgeev for a 10x10 matrix in burst mode.

1. Does CULA support burst mode?

2. How does CULA compare to MKL and magma-1.1.0?

magma-1.1.0 performs very poorly for matrix sizes below 1024x1024. BUt there's no mention of burst mode.

I could not find any CULA documentation on burst mode for small matrices. Please let me know if this feature is added in R13.

Also, my organization bought the license for CULA r12, is R13 a free upgrade?

Thanks in advance.