Jeff Hammond

Field: Next time a vendor offers to donate hardware, you might ask for a big SSD so you can setup a virtual machine for every Linux distro. Just a thought.

It might be interesting to see if simd pragmas cause anything better to happen with the reference kernel. I’ve got a list of all of those, in addition to the...

I am interested in helping implement MKL extensions in BLIS to ensure compatibility. It is ideal if there is a GitHub issue for each function or collection of related functions....

See http://www.netlib.org/lapack/explore-html/d8/db4/cblas__xerbla_8c_source.html. BLIS is matching the behavior of Netlib's implementation. However, I also see http://www.netlib.org/lapack/explore-html/d6/d80/a01076_source.html so there may have been a change from 3.8.0 to 3.9.0. OpenBLAS appears to do...

If BLIS defines XERBLA and friends as weak symbols, they can be replaced dynamically at runtime, using the time-tested methods of the MPI profiling interface.

Ah, it seems that `BLIS_TREE_BARRIER` does not get defined in frame/thread/bli_thrcomm_openmp.h when it is defined in the above config file.

This is a trivial bug where struct members need to be updated to track the usage, when the preprocessor switch is on. ``` frame/thread/bli_thrcomm_openmp.c:119:8: error: ‘thrcomm_t’ {aka ‘struct thrcomm_s’} has...

@mratsim Can you run https://github.com/jeffhammond/vpu-count to see if that detects AVX-512 2x FMA correctly?

@fgvanzee @devinamatthews @loveshack I have to wonder if addressing my comment here (https://github.com/flame/blis/pull/351/files/94b34d38f6dffc074a4f12a1936c0ddba51f47ee..5597169dda1c4cb0c447cb589f8d5c2a5418a259#diff-81ef49aa7330af78381263dcf0acbea8) would solve this problem.

@mratsim For context, the reason BLIS detects your CPU as Haswell is that it _incorrectly_ thinks it has only 1 FMA unit. The handful of SKX processors that have only...