This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

[Bug tree-optimization/30032] sqrt(CGAMMA*PRES[j]/DENS[j]) much slower than compiting compiler



------- Comment #3 from burnus at gcc dot gnu dot org  2006-11-30 18:54 -------
Created an attachment (id=12717)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=12717&action=view)
icc -O3 -xW -S of only the function eos()

> but I think what ICC does is only calculates CS[1] instead of all of CS[i].
I don't think so. If I use:
    for(j = 0; j < NODES; j++) {
      CGAMMA = CGAMMA + CS[j];
    }
rather than
    CGAMMA = CGAMMA + CS[1];
the timings don't change. I also didn't include -ip or -ipo (inter process
analysis). With this change: icc 0m2.324s; gcc 0m7.194s.

> Also you are using VLAs so maybe ICC also converts those VLAs to normal 
> arrays. You might also want to add "-mfpmath=sse,387".
This does not really help: gcc 0m7.066s


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=30032


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]