Re: [PATCH] perf tools: Allow multiple threads or processes in record, stat, top

[David Ahern <dsahern@gmail.com>]

On 02/08/2012 10:36 PM, Namhyung Kim wrote:
> Oh, I meant this:
> 
> static struct thread_map *thread_map__new_by_pid_str(const char *pid_str)
> {
> 	...
> 	while (*str) {
> 		...
> 		tmp = thread_map__new_by_pid(pid);
> 		...
> 	}
> 	...
> }
> 
> static struct thread_map *thread_map__new_by_tid_str(const char *tid_str)
> {
> 	...
> 	while (*str) {
> 		...
> 		tmp = thread_map__new_by_tid(tid);
> 		...
> 	}
> 	...
> }


I'm still not seeing the code simplification here. For example,
new_by_tid(tid) allocates threads, sets the first element (it only
supports one) and returns it. In new_by_tid_str we loop over a CSV,
allocating/re-allocating on additional and then setting the current i-th
tid as opposed to the 0th element.

Similarly, for pid.

Personally, I liked my first patch better which dropped the current
tid/pid functions; they are only used by perf-test and it can be updated
to convert getpid() to a string and use the new API.

David