overhead of bfd_{get,put}*()

[David Miller]

From: Ian Lance Taylor <iant@google.com>
Date: Tue, 23 Feb 2010 08:00:55 -0800

> David Miller <davem@davemloft.net> writes:
> 
>> The top offenders were, surprisingly for me, bfd_getb64(),
>> bfd_putb64() and bfd_getb_signed_64().  And it's not because they
>> touch memory, it's the byte loads and shift/or dance they do.
> 
> This is exactly why gold does its complex template dance: to avoid
> that overhead.
> 
> Those functions will hit unaligned data in some cases.

GOLD takes more than 11 seconds for this test case, which is almost 3
seconds longer than the BFD linker :-) So it may be handling this
specific issue well, but it does have other performance problems when
there are so many relocations.

I'll investigate at some point.