Sign Up to our social questions and Answers Engine to ask questions, answer people’s questions, and connect with other people.
Login to our social questions & Answers Engine to ask questions answer people’s questions & connect with other people.
Lost your password? Please enter your email address. You will receive a link and will create a new password via email.
Please briefly explain why you feel this question should be reported.
Please briefly explain why you feel this answer should be reported.
Please briefly explain why you feel this user should be reported.
Working with an undisclosed API, I found a function that can set the number of “GPRs” allocated to a different phase of the graphics pipeline.
My guess is that GPR means “General Purpose Register”, but I could not find really detailed documentation about their usage in widespread GPUs (also, their number, etc).
Does it mean something else ?
Provided you have correctly guessed the meaning of “GPR”:
For NVIDIA GPUs, the official source of such information used to be Appendix A to the CUDA Programming Guide. But if you don’t feel like downloading a PDF just to look up the number of registers, go to Wikipedia and scroll down to the line “Number of 32-bit registers per multiprocessor”.
In CUDA, registers are allocated dynamically. Without going much into details, the less registers your CUDA code needs, the more resident threads you can potentially have per multiprocessor, thus increasing utilization.