Quick question:
Does Tesla support double precision sine/cosine in hardware, or is it also done in software like in GeForce?
What’s the approximate ratio of double precision to single precision sine/cosine throughput in GeForce Fermi?
Quick question:
Does Tesla support double precision sine/cosine in hardware, or is it also done in software like in GeForce?
What’s the approximate ratio of double precision to single precision sine/cosine throughput in GeForce Fermi?
There is no hardware support for double precision sin/cos.
Even in single precision, if you want full precision, it will evaluate sin/cos in software.