Re: [hatari-devel] Tracing constraints

[Eero Tamminen <oak@xxxxxxxxxxxxxx>]

Hi,

On 21.11.2023 20.13, J.Young wrote:
> Well for my use case, I am tracing fixed blocks of code for game 
> reverse-engineering so it's quite a limited sandbox. Probably an 
> after-trace parser will work best for that since a general-purpose 
> algorithm in hatari seems too much overhead with possible 
> ambiguous/non-resolvable situations.

Instead of tracing, you could try profiling.

From profiler disassembly output you can see executed instructions, how 
many times each subroutine gets called (regardless of whether it was 
exception handler or not), and you can also do some deductions about 
jumps based on the per-instruction run counts.

For every address that you have provided a symbol, profiler keeps track 
from where it's called, and profiler post-processor can provide 
call-graphs of that.


> I need the jmp/jsr (a0) to show the address though :).

Often statistical data provided by profiler is enough, one does not 
necessarily need to know how a specific execution round happened to 
work.  Exceptions can happen at fairly random places, unless one does 
enough runs and checks from profiler what happened on average.


Or if one wants to know whether specific exception (handler) was ever 
called during specific part of code, one can set breakpoint to trigger 
for the particular exception exception only at start of critical section 
of the code, and remove breakpoint at end of critical section.

(I.e. have breakpoints at start and end, which set and remove breakpoint 
for the exception handler.)


	- Eero