langou

> However if this function is going to be implemented in the reference, it should actually bring in some significant benefits beyond an implicit 2xDGEMM calls if possible in my...

> Functionality-wise similar routines might be requested such as triangular-triangular multiplication that comes in Sylvester/Schur related algorithms. Hence I'm not sure halving the matmul flops is a killer feature. But...

@mgates3 very nice application! Yes, indeed, if for a given matrix A, you know U, the unitary polar factor of A, then you can recover the Hermitian polar factor by...

We need to speak about names. The PR named the routine GEMMT. The names HERKX and GEMMT are already used for this functionality in some BLAS libraries. Then HERKMM is...

If a library already provides this functionality, then this library could keep on providing this functionality under the current name that they use. And then it would be nice if...

Hi Thijs, see: [https://www.netlib.org/lapack/releases/](https://www.netlib.org/lapack/releases/) Cheers, Julien.

@ocelotStride are you able to submit a PR?

PR stands for Pull Request. Here is one website: https://opensource.com/article/19/7/create-pull-request-github that looks about right on how to do this.

Maybe this is a jargon thing but, in the upper case, LAPACK SYTRF returns a factorization of the type ``A = U**T*D*U``. Maybe we should not call this a Bunch-Kaufman...

Thanks Jay and Jeff. As Kay suggested, I have included the `-assume protect_parens` in the `make.inc.ifort` file in the INSTALL directory. Hopefully, that should do the trick. Cheers, Julien.