Several of our samples include the same code to increment a counter from
the cache (most suboptimal). Given that we now support private and shared
caches, the possibility of eflags analysis and register liveness analysis,
inc vs. add flag/perf effects etc. writing a maximally efficient in cache
increment is fairly non-trivial. Doing more complicated stats gathering is
even harder. As such convenience routines for efficient stats gathering
would likely be appreciated. A good place to start would be to replicate
the functionality of our internal STATS_* macros with client available
in-cache versions.
From derek.br...@gmail.com on February 24, 2009 14:42:49
this was PR 315346
Several of our samples include the same code to increment a counter from the cache (most suboptimal). Given that we now support private and shared caches, the possibility of eflags analysis and register liveness analysis, inc vs. add flag/perf effects etc. writing a maximally efficient in cache increment is fairly non-trivial. Doing more complicated stats gathering is even harder. As such convenience routines for efficient stats gathering would likely be appreciated. A good place to start would be to replicate the functionality of our internal STATS_* macros with client available in-cache versions.
Original issue: http://code.google.com/p/dynamorio/issues/detail?id=54