Re:The Death of Optimizing Compilers
by readerrr on Saturday March 14, 2015 @01:18PM
(#9203343)
On 2.: Maybe he is talking about branch prediction vs lack of this feature in compilers. What do I mean by that is that optimizing compilers( at least in c ) don't optimize based on measurements from actual programs. A compiler that would run the program for a certain amount of time, to profile, could find the weak spots and optimize.
[ Parent
| on HN ]
Re:The Death of Optimizing Compilers
by ericfontaine on Saturday March 14, 2015 @01:41PM
(#9203442)
In addition to GCC, a bunch of other compilers including Intel and Visual Studio do profile based optimization, as well as many dynamic compilers, such as firfox javascript or Java VMs. Profil-guided opitmization has been studied for decades: https://en.wikipedia.org/wiki/Profile-guided_optimization
GCC can do that.
You are mistaken, unless you can prove me wrong. Care to elaborate?
Re:The Death of Optimizing Compilers
by thoughtpolice on Saturday March 14, 2015 @01:55PM
(#9203500)
GCC has supported `-fprofile-generate` and `-fprofile-use` for profile-guided optimization since (at least[1]) GCC 3.4, which was released over 10 years ago:
https://gcc.gnu.org/onlinedocs/gcc-3.4.0/gcc/Optimize-Option...
Firefox is an example of an application that ships production binaries with PGO.
[1] It was actually supported earlier than this, but I didn't search further; the 3.4.0 manual says "New -fprofile-generate and -fprofile-use command-line options to simplify the use of profile feedback", implying it was there before in a less convenient form.
Or rather, "I may be mistaken in my grand assertion, but I'll place the burden of proof on you".
I assume mansr refers to profile guided optimizations in gcc and many other compilers. See http://en.wikipedia.org/wiki/Profile-guided_optimization.