Devin Matthews

This PR adds new BLAS-like level 3 operations which include a diagonal matrix "sandwiched" between two other matrices: - `bli_gemdm`: `C = alpha*A*D*B + beta*C` - `bli_gemdmt`: `C = alpha*tri[ul](A*D*B)...

Refactor the control tree infrastructure

22

This is a major refactor that places the vast majority of the logic, physical block sizes, preferences, kernel/variant pointers, etc. under the direct purview of the control tree. This refactor...

See https://github.com/devinamatthews/tblis/pull/56 for an example of where GCC can sometimes be detected as the wrong compiler based on the symlink used to invoke it. The solution there is to also...

Add skew-symmetric BLAS operations

21

This PR adds a number of level-2 and level-3 skew-symmetric (and skew-hermitian) BLAS operations, defining the essential operations of a "Skew-BLAS" interface. These operations have been added as full 1st-class...

Support compiler names with spaces

1

In some cases, especially when using autotools, the value of `CC` or `CXX` may have spaces in it. For example, in my project `CXX` is detected as `g++ -std=c++17`. Should...

Fix bug with piledriver gemm kernels.

1

First, just enable BLAS tests and see if it fails.

`bli_scalv` has inconsistent semantics

1

The operations `bli_(inv)?scal[vdm]` are alone in having a `conjalpha` parameter (in the typed API and kernel definitions), while also taking an input vector (matrix diagonal, matrix) `x` which does *not*...

Cannot use complex `bli_scal2s` with `&x == &y`

1

The level0 macros `bli_scal2s`, `bli_scal2js`, `bli_scal2ris`, and `bli_scal2jris` (collectively, `y = alpha*conj?(x)`) all boil down to the same implementation which, for complex numbers, writes the real part of `y` before...

Temporary C microtile is too small sometimes

1

A temporary C microtile is used in various places, such as on the diagonal of a symmetric matrix (GEMMT) where care must be taken not to write to the unstored...