search

nvdla / vp

Virtual Platform for NVDLA
Other
135 stars 81 forks source link

VP run time #38

Open DrVijayK opened 5 years ago

DrVijayK commented 5 years ago

I observed that nv_full INT8 inference on VP is taking more time than with FP16 inference. NVDLA HW branch: nvdlav1, config: nv_full NVDLA SW branch: Latest with INT8 option in nvdla_compiler

Please let me know whether this is expected? And whether this will happen when running on HW in say FPGA.

guodonglai commented 5 years ago

Hi, I was doing the same nv_full testing on vp but got some error code as shown in: https://github.com/nvdla/sw/issues/143. Could you provide some details in how you managed to run nv_full on INT8? Which loadable and image are you using?

DrVijayK commented 5 years ago

follow thread nvdla/sw#140.

Hassan313 commented 4 years ago

@DrVijayK I am having the same problem. Have you found the answer to your question?