SPARTA vs Standard GEMM/SpGEMM

[fvella]

• Discussing on block dense MKL multiplication (no batch, explicit loop) compare with MKL sparseXdense multiplication. 32 bit.

• Dense matrix (rectangular, progressively larger 2,4,8,...)

• verify multiplication correctness.

• Test various small graphs (small RMAT 2^8, facebook, nonsymm small MM). Not for benchmark purposes.

• add timing.

• Flag per intel_mkl

• Add picture of density blocks

• Code clean up e.g. removing obj file and CUDA SDK dependencies