TODO

Integrate double precision complex to blas1/2/3 code base
 - Handling of CONJUGATE transpose within kernel functions
 - Per instruction set optimizations. (x86_64: AVX,FMA, arm: NEON,NEON_FMA)
 - Unit tests for BLAS1/2/3 functionality (OK)

Complex versions of lapack functionality
 - requires proper handling of conjugate tranpose in blas-functions

Per instruction set optimizations for complex numbers

Matrix I/O
 - functions for reading and writing matrices from files
 - marshalling/unmarshalling for transmission

Unit testing for all supported types (OK)

Automake
 - instruction set testing for compilation (NO)
 - use of config.h (not used, should it?) (OK)
 - library selection, all or only some versions

Missing lapack functionality
 - Eigenvalue computations (EVD, TRIDIAGONAL)
 - Condition numbers estimation

Multi-threading support  (OK)
 - replace current simplistic implementation with something better

Matrix operations
 - element-wise add, substract, multiply, div and other functions: abs, log, exp ...
 - ABS flag for gemm, gemv to compute with absolute values C = |A|*|B| + |C|