Benchmarks Of GCC 4.5.0 Compiler Performance

Written by Michael Larabel in Software on 19 April 2010 at 01:00 AM EDT. Page 2 of 6. 44 Comments.

Beginning with measuring how long it took to build the Apache 2.2.11 web server with each GCC release, with the 4.5 release it took almost 8% more time to build than it did in GCC 4.3.0/4.4.0. This slower build process of Apache in GCC 4.5 may be attributed to the Link-Time Optimization (LTO) or other new work merged into GCC in the last year. The Link-Time Optimization is described on the GCC Wiki as "Link Time Optimization (LTO) gives GCC the capability of dumping its internal representation (GIMPLE) to disk, so that all the different compilation units that make up a single executable can be optimized as a single module. This expands the scope of inter-procedural optimizations to encompass the whole program (or, rather, everything that is visible at link time)."

Similarly, building PHP with GCC 4.5.0 was about 7% slower than in GCC 4.3.0/4.4.0.

With building ImageMagick under GCC 4.5 it ended up taking almost 11% more time to complete.

Starting with our benchmarks that are looking at the performance of the generated binaries under the past three GCC releases was 7-Zip. Right from the start, GCC 4.5.0 proved to be faster with 7-Zip than in GCC 4.3.0 or 4.4.0. GCC 4.4.0 had regressed slightly in its performance compared to GCC 4.3.0, but that was cleared up with 4.5.0 where it is now running faster than the 4.3 release.


Related Articles