prasun3

I made a change to introduce an 'L0_filter' mode so that we can switch from warmup trace to the full trace. I am only creating this if flag `need_l0_filter_mode` is...

@abhinav92003 In my first implementation I was flushing the code cache. I am redoing the code changes to use drbbdup. In the older implementation I am creating separate traces. The...

Ok, so in your case you have one 'full trace' raw file which is post-processed to produced a trace.gz file. Then this trace.gz file is further filtered to produce a...

How is trace_for_instrs expected to work with L0_filter - is trace_for_instrs expected to count before the filtering occurs, or after? Here's a simple program that mostly hits in the L0...

> For separation of tracing windows we do trigger the end-of-bb clean call on a window change using a conditional-free scheme of writing a difference in TLS value or something...

Thanks for the debug! I added a comment here: https://github.com/DynamoRIO/dynamorio/pull/6356#issuecomment-1757043495

> @prasun3 do you have an opinion on how important this is: if real-world rep string loops are generally not very long maybe never switching in the middle of one...

> @prasun3 -- another question: what is your preference for whether each iteration of a rep string loop should have a regular instruction fetch record, or not? We had decided...

Looks like this issues was also reported here https://groups.google.com/g/DynamoRIO-Users/c/gKF6RqG6wqI

On RHEL 7, `/usr/include/elf.h` does not define EM_RISCV, but now we have `third_party/elfutils/libelf/elf.h` that does define it. Could we use the latter header instead of the former?