| |||||||||||||
Tracing infrastructuresTracing infrastructuresPosted Sep 21, 2006 5:49 UTC (Thu) by dlang (✭ supporter ✭, #313)Parent article: Tracing infrastructures
as I read the flamefest I thought the argument was that dynamic tracepoints were better at seeing and getting hold of variables.
the dynamic tracepoint has access to the entire system environment at the time the trace event takes place, but the static tracepoint has to specify (in the compiled source) exactly what variables to pay attention to. if it specifies too many it creates additional work for GCC to store and pass variables that don't mean anything.
(Log in to post comments)
Tracing infrastructures Posted Sep 21, 2006 14:07 UTC (Thu) by ajax (guest, #7251) [Link] Global variables, yes. Local variables tend to move back and forth between registers and memory so it is hard, at any point in the code, to determine where a local variable resides at that point.
Tracing infrastructures Posted Sep 22, 2006 10:12 UTC (Fri) by nix (subscriber, #2304) [Link] Describing that is DWARF2's job, of course. GCC's generation of DWARF2 info is not perfect, but it's good enough most of the time.
|
|||||||||||||