GCC, LLVM Clang及Open64编译器比较.docVIP

GCC 4.6, LLVM/Clang 3.0, AMD Open64 Compiler Benchmarks For those interested by last weeks articles on the AMD Bulldozer benchmarks of the GCC, Open64, and LLVM/Clang compilers and the FX-8150 compiler tuning, but curious to know how the new LLVM/Clang 3.0 and other compilers perform on other processors, heres some new test results. These tests are looking at GCC 4.6, LLVM/Clang 3.0-RC1, and AMD Open64 compilers on Intel Sandy Bridge and AMD Shanghai hardware. The first release candidate of GCC 4.6.2, LLVM/Clang 3.0 RC1, and AMD Open64 were tested on these two non-Bulldozer setups. The CFLAGS/CXXFLAGS set outside of each tests default flags were specifying the march/mtume option to native for taking advantage of each systems native features. GCC was built with --enable-checking=release and LLVM was built with --enable-optimized and --disable-assertions, to mirror release builds. The Open64 compiler was using the 64-bit binary provided by AMD. All other testing remained at the respective defaults. The AMD Shanghai setup was a dual AMD Opteron 2384 quad-core (octal-core total) system with 4GB of RAM and a SATA HDD. The Intel Core i7 Sandy Bridge setup had an i7-2630QM (quad-core with Hyper Threading), 8GB of RAM, and a Super Talent SSD. Both systems were running Ubuntu 11.10 (x86_64) with the Linux 3.0 kernel. Results for Open64 on Sandy Bridge were not available since the AMD-sponsored compiler failed to build working binaries for any of the tests. Now lets see the results. Open64 was certainly the fastest compiler for the older dual AMD Opteron setup, but since the compiler failed on Intel Sandy Bridge hardware, its not known whether this lead of being more than twice as fast as GCC/LLVM is due to AMD-specific or general optimizations. As far as GCC vs. LLVM/Clang, the Apple-sponsored compiler was slightly slower than GCC on the Opteron setup while LLVM/Clang carried a nice lead for the Intel Sandy Bridge. The OpenMP-using Smallpt test is much slower at present wit