Incorporate Parser Acceleration
Testing notes:
- Run
make example
first in themaster
branch and thenmake clean-out
on the23-parser-acceleration
branch. Runmake example
again on the23-parser-acceleration
branch. Verify that results are the same. - When running
make example
above each time, runwatch -n 1 nvidia-smi
on another terminal session to view the GPU utilization of the system. Verify that only 1 GPU is used in themaster
branch and with a very lower propotion of the available GPU space while all the available GPUs are used in the current branch with better utilization of the available GPU space. - Confirm that execution speed is faster in the current branch. (In the
northbay
system with 2 GPUs, the total parse time reduced from 386.19 seconds to 66.62 seconds and the total pipeline time from 459.89 seconds to 140.26 seconds.
Closes #23 (closed)
Edited by Ayush Kumar Shah