Ongoing development status of some BLAS routines

Currently, ReproBLAS only supports sequential BLAS. However, it does MPI reduction primitives from which parallel reproducible BLAS can be built. Future versions of the library will include parallel reproducible BLAS. Note that in the table below, "Vendor BLAS" refers to vendor implementations of BLAS, including SIMD, multicore, etc.
Function Description Vendor BLAS reproBLAS.h reproBLAS_MPI.h
Sequential OMP MPI
{d|s|z|c}swap swap 2 vectors
{d|s|z|c}scal x = a * x
{d|s|z|c}copy x = y
{d|s|z|c}axpy y = a * x + y
{d|s}dot dot product r{d|s}dot
{z|c}dot{u|c} dot product r{z|c}dot{u|c}
{d|s|dz|sc}nrm2 second norm r{d|s|dz|sc}nrm2
{d|s|dz|sc}asum absolute value sum r{d|s|dz|sc}asum
{d|s|z|c}sum sum r{d|s|z|c}sum
i{d|s|z|c}max index of max abs
{d|s|z|c}gemv matrix vector multiplication r{d|s|z|c}gemv
{d|s|z|c}trsv solving triangular matrix problems
{d|s|z|c}ger A := alpha*x*y' + A
{d|s|z|c}gemm matrix matrix multiplication r{d|s|z|c}gemm
{d|s|z|c}trsm solving triangular matrix problems with multiple right hand sides
: under development : non reproducible : reproducible : unknown
Last update 2/21/2016