Comment by bhuthesh_r

I'm trying to use 16 bit floats for matrix multiplication on x86-64. I found solutions for ARM and some NVIDIA GPUs but none for any X86-64 chips. Any pointers in this direction would be helpful.