Hi Michael,
Thanks for the answer. I would like to know if someone has investigated
this issue for some benchmark or real-world cases. Is there any
write-up/report/paper on this thing?
Thanks,
Wirawan
On Tue, 6 Jan 2009, Michael Meissner wrote:
On Tue, Jan 06, 2009 at 03:28:06PM -0500, Wirawan Purwanto wrote:
Hi,
What is, and how much is, the effect of -mtune (or its implication in
-march) switch of gcc (version 4.3, that is)? Suppose I use
-march=pentium4 and then run the binary on pentium M machine (both support
MMX, SSE, and SSE2), what kind of penalty will I get? I don't have a
particular code in mind, but think of a numerical-type code with lots of
double precision math.
It really depends on what your code is, what machine you specified in the
-mtune= option, and what machine you are running on. Some code will not see an
appreciable difference, some will see a huge difference. If you are compiling
code to run on the machine you are going to be running on, consider using
-mtune=native. The default tunings are an attempt to run reasonably well on
both modern Intel and AMD machines. However, it is a compromise tuning, in
that there are some things that could be faster if compiled for a specific
machine.
--
Michael Meissner, IBM
4 Technology Place Drive, MS 2203A, Westford, MA, 01886, USA
meissner@xxxxxxxxxxxxxxxxxx
