[Bug middle-end/31249] pseudo-optimzation with sincos/cexpi

[dominiq at lps dot ens dot fr]

------- Comment #5 from dominiq at lps dot ens dot fr  2007-03-19 12:43 -------
> There is no option to turn it off.  But for !TARGET_C99_FUNCTIONS and
> !TARGET_HAS_SINCOS targets it's off.  

>From my understanding of the thread

http://gcc.gnu.org/ml/gcc/2007-03/msg00639.html

if !TARGET_64BIT, then TARGET_C99_FUNCTIONS depends on 
darwin_macosx_version_min which seems presently default to 10.1.
So TARGET_C99_FUNCTIONS seems to be set to 0 at least for a G5 under 
OSX 10.3 and a G4 under 10.4.

> Does darwin have a sincos() library function?

I don't know. If there is no answer in this PR, I can ask the 
question on gcc@gcc.gnu.org.

>From the behavior reported in PR30969, PR30980, and PR31161, it seems
that the optimization is off for gcc, but on for g++ and gfortran,
though I cannot figure out why.

> If the ABI were not stupid of course ;)

Since sin() and cos() are non trivial functions, I am very surprised 
that a wrong API makes a 50% difference.

If the API is so time consuming, why not inline sin() and cos()? 
In addition a decent optimizer should be able to eliminate redundant
part of the codes, making the use of a sincos() function not necessary,
or am I too naive?

What is the best way to collect data on the different platform
to evaluate how this optimization is really working?
I have only access to OSX and Intel or AMD64 under Linux.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=31249