GCC & LLVM Clang Performance On The Intel Atom

In our earlier compiler benchmarks, we found there to not be much performance difference when it came to Blowfish on John The Ripper except when using LLVM-GCC and the GCC DragonEgg plug-in, but as those two compiler options were left out of this benchmarking roundabout, the Blowfish results are not particularly interesting. However, it is good to see here the LLVM Clang performance is right on target with GCC.

With the Dhrystone 2 performance on the BYTE Unix Benchmark there's a nice performance boost found with the most recent GCC 4.6 development snapshot compared to GCC 4.5.1 where the performance was upped by 6%. The LLVM Clang performance produced a binary running at approximately the same speed of GCC 4.5.1.

With the integer arithmetic test from the early November compiler article GCC 4.5, GCC 4.6, LLVM-GCC, DragonEgg, and Clang all failed to run this test correctly. This is still the case with GCC 4.5.1 and GCC 4.6.0-20101120 on the 32-bit Atom, except the 32-bit LLVM Clang compiler did work this time. LLVM Clang on the Intel Atom ran at approximately the same speed as GCC 4.2.4/4.3.5/4.4.5.

The floating-point arithmetic performance also was close to the same between Clang and GCC.