How do I set up the graph analysis module when running multiple K-Schedulers at the same time about k-scheduler HOT 4 CLOSED

1. Create a directory for each fuzzing repetition. In your cases, that is rep1, rep2, ..., rep10.
2. Put gen_dyn_weight.py, afl_fuzz_kscheduler along with other graph data files in every directory.
3. Open 20 terminals or write a python script to run 10 fuzzers and 10 corresponding gen_dyn_weight.py. Make sure you enter the correct directory before running afl-fuzz-kscheduler or gen_dyn_weight.py

from k-scheduler.

Thanks, but I met another problem when I tried to run K-scheduler. I followed your advice to put python script, graph files and afl_fuzz_kscheduler in every directory. But when I ran k-scheduler, the stats showed that the exec speed of fuzzer is 0.
exec speed : 0.00/sec (zzzz...)
The output of python script is :
covered 47 node, generate new graph.
My directory:
afl-fuzz_kscheduler border_edges child_node dyn_katz_cent graph_data_pack parent_node readelf.ll afl_out_cent cfg_out_readelf cur_coverage gen_dyn_weight.py katz_cent readelf_fix.ll signal
The number of edges:
wc -l katz_cent: 42456 katz_cent
I don't understand why fuzzer doesn't execute input after running for a few seconds.
Look forward to your reply. Thank you very much

from k-scheduler.

By the way, the input seeds that I used is the default seeds provided by afl.

from k-scheduler.

1. It seems the binary is not properly built. Please check a detailed tutorial on how to build binary with kscheduler on binutils (including readelf) in https://github.com/Dongdongshe/K-Scheduler/blob/main/afl_integration/build_example/README(non_wrapper%20based%20program).md.
2. The input seed is orthogonal to kscheduler since it is a light-weight data-driven solution without any ML model training.

from k-scheduler.