This is the mail archive of the archer@sourceware.org mailing list for the Archer project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: update on speedup project


>> This omits namespaces, which is weird and (IMO) not useful.

Roland> I wouldn't take this as an exhaustive normative explanation.
Roland> Rather, I'd infer from "fully qualified" that it means all the
Roland> qualifications a name can have in the language in question.
Roland> You can ask dwarf-discuss@lists.dwarfstd.org for clarification.

Yeah, if I pursue this I will.

>> Now I am looking at something Apple implemented: defer reading of
>> shared library symbols until they are actually needed.  I think this
>> may be a simpler way to achieve the same goal of speeding up "attach".

Roland> What does "actually needed" mean?  Until the first by-name search?
Roland> (It's clear what it means when you're not doing any by-name searches.)

I was thinking something like: do not read symbols for an objfile
until either:

1. A by-name search reaches this objfile.

2. A by-address search needs partial symbols for this objfile.
   This seems needed for backtraces.

Tom


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]