Can I vary the number of threads ATLAS uses dynamically?
No. The maximum number of threads to use is determined at compile time. ATLAS will never use more than this, but may use less if the problem sizes are too small to get speedup from the additional parallelism. What’s the deal with the RHS in the row-major factorization/solves? Most users are confused by the row major factorization and related solves. The right-hand side vectors are probably the biggest source of confusion. The RHS array does not represent a matrix in the mathematical sense, it is instead a pasting together of the various RHS into one array for calling convenience. As such, RHS vectors are always stored contiguously, regardless of the row/col major that is chosen. This means that ldb/ldx is always independent of NRHS, and dependant on N, regardless of the row/col major setting. Why don’t you speedup/improve ATLAS’s search There are several questions here, handled in their own sub-questions: • Why don’t you improve the type of search? • Why don’t you improve the speed of
