Matrix multiplication. Transposed input matrix and rectangular tiles v.5i benchmark

WebGL2-compute based NxN matrix multiplication C = A x B Shader v.5 benchmark. Fixed random A, B 1024x1024 buffers are used for calculations, therefore You can set new N value (up to N=1024, multiple of TS). TS = 64 tile sizes (local_size_x,y), TSDK = 32. FLOPS = 2 N3 / time. Random CGPU and CGPU - CCPU are shown.


Intel i3-8100     ~43.5 GFLOPS (N=1024).

SGEMM in WebGL2-compute     updated 27 June 2019