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
For embench-iot 2.0 we discussed that we would like to modernize/update the python build scripts. @wallento wanted to take a look at it if I'm correct.
Some things that we could address:
change the --target-module speed benchmarking architecture.
It might be better to use callbacks/a more flexible approach that doesn't rely on a user-constructed shell command.
use a more declarative framework such as doit
reduce usage of global data/variables
update the benchmark size computation
right now, the size is computed via hardcoded section names, however, some sections aren't accounted for: e.g. .eh_frame, .got, .init, .plt, maybe we should instead sum up all MemSiz fields of all segments that get loaded?
The text was updated successfully, but these errors were encountered:
building now uses scons and a simpler/flat configuration strucure that utilizes the command line more.
the size benchmarks now only work on ELF (no macho on embedded devices) and instead uses section flags to associate sections with metrics.
the speed benchmark now uses a simpler/more abstract interface to user modules
@jeremybennett i tried install lief on Ubuntu using apt install python3 python3-pip && pip3 install lief and it seemed to work. Are you sure we need to swap out the elf parsing library? I couldn't replicate the problem with an Ubuntu install myself.
For embench-iot 2.0 we discussed that we would like to modernize/update the python build scripts.
@wallento wanted to take a look at it if I'm correct.
Some things that we could address:
--target-module
speed benchmarking architecture..eh_frame, .got, .init, .plt
, maybe we should instead sum up all MemSiz fields of all segments that get loaded?The text was updated successfully, but these errors were encountered: