Author: Dann Corbit
Date: 21:07:31 05/16/02
Go up one level in this thread
On May 16, 2002 at 06:31:27, Mark Christiaens wrote: >On May 10, 2002 at 16:34:39, Dann Corbit wrote: > > >>>C/C++ compilers compile one source-code module at a time and cannot optimize >>>over these boundaries. >> >>Not true, for a long time now. > >Hmm, which compilers do this? Let's not forget that the normal aproach when >compiling C/C++ is to compile the source files one at a time in isolation. In >those circumstances, you can't inline much at all, even if your compiler would >be willing to do so. That's back in the days of Flintstones. The Intel compilers do global optimizations and global reordering of the objects to minimize the jump distnaces. It will also do profile guided optimizations to make the most likely jump scheduled in the pipleline by default. The Microsoft .net compiler does the same thing also does whole program optimization and even the old Microsoft compilers did global optimizations. Global optimizations have been done for literally decades now by good compilers. >>> JIT-compilers for Java see the whole application and can >>>for example inline methods from other classes which were not available at >>>compile time. >> >>So can C and C++ compilers. > >???? You can't inline stuff you don't have available unless you're psychic. I knew you were going to say that.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.