Re: [9fans] Good enough approximation for ape/pcc

[Ronald G Minnich <rminnich@lanl.gov>]

Brantley Coile wrote:

> Or they might just like writing stuff for the compiler, personal
> entertainment.

bingo.

I know one or two of the guys, or used to. It was all about optimize,
measure, optimize some more. The big picture got lost.

It's humorous to see the inline madness, on machines with 512KB caches yet:
[rminnich@q linux-2.6.16]$ grep -r -i 'static inline' . |wc
   21262  142677 1810997

yep, 21262 instances of 'static inline'. Even better:

[rminnich@q linux-2.6.16]$ grep -r -i 'static inline' include/ | wc
    8249   54789  689506

so, like, 13K of them are NOT in .h. Which is worse, the fact that the
code is replicated all over the place, via .h, or there are .c files
with static inline in them? yikes.

I was talking to the afterburner guys some months ago.

them. "How many inb are there in Plan 9?"
me. "Oh, one, maybe two"
them. "What about the inline and asm stuff?"
me. "The, uh, what? Oh yeah, people used to do that stuff in the 70s and
... oh .. the 21st century too ... hmmm ... they like to bang their head
between two blocks I guess."

There's lots of other bad stuff that happened too. Linker-sets, for
example. Or, oh, the stuff like:
if (likely(a != b)) {eatme();}

The world went mad.

ron