Execution tracing for call graph generation

Page 3/3
1 | 2 |

By yzi

Champion (439)

yzi's picture

12-02-2019, 16:43

I think that instrumentation-based profiling is a second-best choice compared to profiling unmodified code, and source code instrumentation is even worse than binary instrumentation. If you have all the code location info, per-address instruction and clock cycle counts, _and_ even jump/call information, it should be possible to reconstruct everything without instrumenting anything.

By Manuel

Ascended (14995)

Manuel's picture

12-02-2019, 20:38

Sandy Brand wrote:

Do you have any performance issues using larger TCL sctips, btw? I found that, while my approach sort of worked, such a heavy use of TCL scripting and writing data to output ports had a tendency to lock-up the emulator.

That shouldn't happen. Can you give a concrete example, so we can reproduce and investigate that?

Page 3/3
1 | 2 |
My MSX profile