How to calculate DPFLOP ?

[zhongjuzhe]

In vec-cond2d, we have:
printf("The execution took %d cycles.\n", runtime);
printf("The performance is %ld DPFLOP/1000 cycles.\n",
(uint64_t)(1000.0 * performance));

However, in vec-sgemm_main.c
There is no DPFLOP calculation.

Could you support DPFLOP calculation (equaltion) of all benchmarks ?

Thanks.

[jerryz123]

GEMM is pretty straightforwards, the number of multiply-accumulates is just MxNxK. If you add this, I'd appreciate a PR adding the print statement