|
From: Nicholas N. <nj...@cs...> - 2006-10-21 15:30:26
Attachments:
x
|
Josef, With gcc 3.3.4 I get a lot of 'inlining failed in call to X' messages from GCC (see attachment). Some of them look like they are important for Callgrind's speed. If they are, you might like to look at Memcheck's INLINE macro which tries harder than usual to inline functions. Nick |
|
From: Josef W. <Jos...@gm...> - 2006-10-21 15:59:33
|
On Saturday 21 October 2006 17:30, Nicholas Nethercote wrote: > Josef, > > With gcc 3.3.4 I get a lot of 'inlining failed in call to X' messages from > GCC (see attachment). Some of them look like they are important for > Callgrind's speed. Yes, the ones from sim.c could be important. I do not get such warnings with GCC 4.1 here on Suse 10.1. I think I'll have to do some self-hosting experiments with/without inlining these functions. > If they are, you might like to look at Memcheck's INLINE > macro which tries harder than usual to inline functions. Ah, I didn't know about the attribute "always_inline". Thanks for the tip, Josef > > Nick > |
|
From: Julian S. <js...@ac...> - 2006-10-21 22:12:36
|
gcc-3.3 seems to be markedly less enthusiastic about inlining than other versions -- both earlier and later versions. As Josef says, later gcc's don't show this problem. But I think using the INLINE macro might be a good thing since all those warning messages make it harder to see real errors (I aim to have a completely silent "make --quiet", apart from the printing of the top level directory names). J On Saturday 21 October 2006 16:30, Nicholas Nethercote wrote: > Josef, > > With gcc 3.3.4 I get a lot of 'inlining failed in call to X' messages from > GCC (see attachment). Some of them look like they are important for > Callgrind's speed. If they are, you might like to look at Memcheck's > INLINE macro which tries harder than usual to inline functions. > > Nick |