I will personally dollar match every pledge up to $1,000!
We obtained a performance comparable with Intel MKL on Intel Sandy Bridge CPU.
Here is the result of the DGEMM subroutine's performance on Intel Core i5-2500K Windows 7 SP1 64-bit:
Buy the following machines for the experiment and testing.
We need these machines to optimize BLAS on AMD Piledriver and Intel Haswell micro-architectures.
Although we can remotely access a AMD Piledriver server in Versatus HPC, it's better to have a Piledriver machine in local.
In addition, we didn't have Intel Haswell platform.
I have contacted an employee at Intel China. He told me it was very hard. He would try it. So far, I don't get any response from Intel.
I also try to contact AMD China. I don't get any response, too.