Comment by weebull
10 months ago
Just as a general comment on this topic...
The fact that people complain about the performance of the mechanism that enables the system to be profiled, and so performance problems be identified, is beyond ironic. Surely the epitome of premature optimisation.
It's not moronic when the people profiling and the people having performance issues are different groups. That some developer benefits from having frame pointers in libc does not mean that all users of that software also need to have frame pointers enabled.
It goes both ways. I see people with ultra-bloated applications trying to add even more bloat and force it on the rest of us too. Like the guy saying that DWARF unwinding is impractical when you have 1Gbyte code cache.
I don't see the people writing hyperfast low-level code asking for this.
My experience with profiling is anyway that you don't need that fine-grained profiling. It's main use is finding stuff like "we spend 90% of our time reallocating a string over and over to add characters one at a time". After a few of those it's just "it's a little bit slow everywhere".
> After a few of those it's just "it's a little bit slow everywhere".
And after that, you need fine-grained profiling to find multiple 1% wins and apply them repeatedly.
(I do this for a living, basically)
Like disabling frame pointers
So what are these other techniques the 2004 migration from frame pointers assumed would work for stack walking? Why don't they work today? I get that _64 has a lot more registers, so there's minimal value to +1 the register?
In 2004, the assumption made by the GCC developers was that you would be walking stacks very infrequently, in a debugger like GDB. Not sampling stacks 1000s of times a second for profiling.
im sure in ancient mesopotamia there was somebody arguing about you could brew beer faster if you stop measuring the hops so carefully but then someone else was saying yes but if you dont measure the hops carefully then you dont know the efficiency of your overall beer making process so you cant isolate the bottlenecks.
the funny thing is i am not sure if the world would actually work properly if we didn't have both of these kinds of people.