You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
It would need a bunch of work. Basically the way FPGAs work is that you first program them with the kernels, then you can run that same kernel over and over again. Programming takes ~hours.
You'd need to run a program in two stages:
first, run deepcl with some flag (that you'd have to create, and implement :) ), to write out the kernels, send them to the FPGA programmer
then, once the FPGA contains the kernels, you'd run deepcl again, with a different flag (that you'd have to create and implement...), to run the kernels
Typically, FPGAs suck for training, but give good power performance for runtime prediction.
Is the DeepCL can only run on GPU or APU? Could it run on FPGAs? Such as altera?
The text was updated successfully, but these errors were encountered: