BFloat implementation

Open jeffsetter opened this issue 4 years ago • 2 comments

Several of my applications use bfloat types, including applications intended to test the bfloat hardware on the CGRA. What is the plan for supporting bfloat16_t?

Mar 31 '21 23:03 jeffsetter

@jeffsetter for the CGRA I have no idea. I assume the PE people are the ones to ask about that. From clockworks point of view a bfloat is just a 16 wide bit vector that gets moved between compute units, so it doesn't need any special treatment.

Apr 01 '21 22:04 dillonhuff

In order to run any pass through clockwork, I include the compute file. bfloat16_t is not recognized from the beginning, therefore, I can't run through any steps of clockwork. bin/fp_conv_2_2_compute.h:15:3: error: ‘bfloat16_t’ was not declared in this scope

Apr 01 '21 22:04 jeffsetter