This is the mail archive of the gcc@gcc.gnu.org mailing list for the GCC 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: [tree-ssa mudflap] slow startup


Eyal Lebedinsky wrote:
> 
> Since the changes early this month Everything is very slow. I now
> checked and I see that the whole slowdown in before my main() is
> entered. It seems that everything is registered up-front when the
> DLLs are opened.

Following up on my own post, looking at a trace of my simple
program with the old and then the new compiler, I see that there
is a significant increase of 'string literal's registered. For
example, the old trace says:

mf: register ptr=40307378 size=16 type=4 name='llist.c:317
(add_item_to_list) ssa_trace__'
mf: register ptr=402da788 size=10 type=4 name='string literal'
mf: register ptr=40307368 size=16 type=4 name='llist.c:283 (sklxad)
ssa_trace__'

And the new one:

mf: register ptr=40240b98 size=16 type=4 name='llist.c:317
(add_item_to_list) ssa_trace__'
mf: register ptr=40214240 size=21 type=4 name='string literal'
mf: register ptr=4021422b size=21 type=4 name='string literal'
mf: register ptr=40214216 size=21 type=4 name='string literal'
mf: register ptr=40214201 size=21 type=4 name='string literal'
mf: register ptr=402141ec size=21 type=4 name='string literal'
mf: register ptr=402141d7 size=21 type=4 name='string literal'
mf: register ptr=402141c2 size=21 type=4 name='string literal'
mf: register ptr=402141ad size=21 type=4 name='string literal'
mf: register ptr=40214198 size=21 type=4 name='string literal'
mf: register ptr=40214183 size=21 type=4 name='string literal'
mf: register ptr=4021416e size=21 type=4 name='string literal'
mf: register ptr=40214152 size=28 type=4 name='string literal'
mf: register ptr=40214148 size=10 type=4 name='string literal'
mf: register ptr=40240b88 size=16 type=4 name='llist.c:283 (sklxad)
ssa_trace__'

As one can see, there is a long list of extra registrations,
which is typical of the differences. Looking at the .s I can
see a longer list of registered strings in the GLOBAL_I function.

Actually, it seems that mudflap is now registering its own
tracing strings, where these were not registered before. Since
a very large number of strings are generated, it may explain
the sharp increase in startup time. Is this intended?

--
Eyal Lebedinsky (eyal@eyal.emu.id.au) <http://samba.org/eyal/>


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