On Fri, Sep 20, 2013 at 06:49:49PM -0700, Josh Stone wrote:
But this also
shows that the new dwarf_getfuncs slowed down about .5s either way, with
no benefit.
On irc Tom Tromey suggested to trust the CU language to win back most
speed for programs mainly written in C (like the kernel). The attached
patch does that. With this variant of the patch you get all correctness
benefits (imported/partial unit support and detection of nested functions
in C code, plus detection of defining subprograms inside other constructs
for other langauges), but with a slowdown of < 0.1s in this particular case.
Cheers,
Mark