Compiler Optimization questions.

Kingtiger01 · 04-12-2014, 06:46 AM

(Moderators, if this is the wrong section please move it)

Code:

Compiler Group:GCC(GCC/GCJ/CPP)
Platform:x86_64{Intel(Core2duo) & AMD(Bulldozer/amdfam10)}

Hello All!

I had a few questions regarding compiler optimization.

1. Do switches effect binary/library performance? or just compile-time performance. examples:

Code:

-03 
-march=(ex:amdf10)

2. What switches are avalible for GCC, GCJ and CPP. What do they do?

3. Is there any other Compilers for C&C++ on Linux, and what are there benefits?

tronayne · 04-12-2014, 08:26 AM

By default, GCC will compile for your environment (the installed processor). It is also a cross-compiler and you can optionally specify a particular processor target.

Optimization is for run-time, not for compile time. Generally, GCC does a pretty good job of optimizing your code and you can, if you wish, chose multiple levels of optimization (which may or may not be effective depending on your code and other factors).

As to switches, well, there are a large number of them and a description of each is beyond the scope of this post; have a look at http://linux.die.net/man/1/gcc. Of course, if you're running a Linux system, the manual page (and info page) are available via the man gcc utility. Bear in mind that the optional switches are for particular purposes and you need to have a clear understanding of what they do (from the documentation); generally, it's reasonable to use the defaults.

There are other compilers available, notably the Intel compiler; however, the Intel compiler is "pay for" and it is expensive and may not be compatible with your Linux system in all ways. If you have special needs, the Intel compiler may be a benefit but for general use on a Linux platform probably not.

Hope this helps some.

Kingtiger01 · 04-19-2014, 10:50 PM

Quote:

Originally Posted by tronayne

By default, GCC will compile for your environment (the installed processor). It is also a cross-compiler and you can optionally specify a particular processor target.

Optimization is for run-time, not for compile time. Generally, GCC does a pretty good job of optimizing your code and you can, if you wish, chose multiple levels of optimization (which may or may not be effective depending on your code and other factors).

I kind of figured that, but i just wanted to clarify.

The build environment is GNU/Linux, but this is for distribution. So im trying to cross-compile for Intel(Haswell) and AMD(Bulldozer) platforms. Generic targets loose 3-51% due to run-time optimization on some applications.

Ill research more into alternative(targeted) compilers though, but i have a starting point. - Ill list this as solved within 3 weeks, as im waiting for a second opinion, if none in 3weeks then ill list solved automatically.