Skip to content

Latest commit

 

History

History
33 lines (17 loc) · 514 Bytes

Benchmarks.md

File metadata and controls

33 lines (17 loc) · 514 Bytes

Matrix Multiplication Benchmarks

All tests are done using cargo bench with criterion, no changes made, just usingblackbox and so on.

All tests are ran on IEEE754 double precision floating point numbers.

Sizes used are as following:

  • M = 999
  • N = 1000
  • P = 999

Dense Matrices

NxN @ NxN

MxN @ NxP

Sparse Matrices

NxN @ NxN

Around 24.7ms

MxN @ NxP

  • Around 22.1ms

Target specific matmuls

Not yet implemented, but X86-64 with AVX instrctions are soon to be implemented