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 fortran/52473] CSHIFT slow - inline it?


https://gcc.gnu.org/bugzilla/show_bug.cgi?id=52473

Thomas Koenig <tkoenig at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED
           Assignee|unassigned at gcc dot gnu.org      |tkoenig at gcc dot gnu.org

--- Comment #5 from Thomas Koenig <tkoenig at gcc dot gnu.org> ---
Created attachment 41394
  --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=41394&action=edit
Benchmark of straight DO loops vs. library version

Some numbers from

https://groups.google.com/forum/#!topic/comp.lang.fortran/AI0F1Vpkc3I

show that using straight DO loops could both help and hurt:

$ ./a.out
 Testing explicit DO loops
 Dim =            1  Elapsed CPU time =    2.82861114    
 Dim =            2  Elapsed CPU time =    2.93245506    
 Dim =            3  Elapsed CPU time =    2.94523525    
 Testing built-in cshift
 Dim =            1  Elapsed CPU time =    1.65619278    
 Dim =            2  Elapsed CPU time =    2.80988693    
 Dim =            3  Elapsed CPU time =    7.13671684 

I'll see what could be done using an explicit call to memcpy
for the innermost loops.

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