Mozilla Spots Massive Performance Regression In GCC
Mozilla developers on the GCC mailing list have been expressing what they describe as a "massive performance regression" and "what might be the biggest compiler-upgrade-related performance difference we've seen at Mozilla." The Mozilla developers have upgraded from GCC 4.3 to GCC 4.5, which was released in April, and now they are experiencing massive slowdowns.
Taras Glek of Mozilla reports that with this newest release of the GNU Compiler Collection they are experiencing 4-19% slowdowns on average with their automated benchmarks on their 32-bit and 64-bit Linux builds. The only case where there wasn't a slowdown was with the 64-bit SunSpider JavaScript benchmark where it sped up by 8% when being built under GCC 4.5. Some of these Mozilla slowdowns after the GCC upgrade are discussed here and here. The start of the mailing list discussion can be found here.
Because of these performance shortcomings on GCC 4.5, Mozilla is reconsidering their switch to this newest compiler release. At least because Mozilla has an automated benchmarking infrastructure in place, they were able to spot this issue timely themselves rather than waiting for us or others to spot the problem.
Initial thoughts on these major regressions caused by GCC are the Mozilla developers building Firefox with flags for size optimizations rather than speed and possible regressions from code inlining. Mozilla was originally switching to GCC 4.5 due to its newly-introduced plug-in support and the ability to build Firefox with PGO (Profile-Guided Optimizations).
Following the release of GCC 4.5.0 in mid-April we published our own GCC 4.5 benchmarks and GCC vs. Clang/LLVM benchmarks.
Taras Glek of Mozilla reports that with this newest release of the GNU Compiler Collection they are experiencing 4-19% slowdowns on average with their automated benchmarks on their 32-bit and 64-bit Linux builds. The only case where there wasn't a slowdown was with the 64-bit SunSpider JavaScript benchmark where it sped up by 8% when being built under GCC 4.5. Some of these Mozilla slowdowns after the GCC upgrade are discussed here and here. The start of the mailing list discussion can be found here.
Because of these performance shortcomings on GCC 4.5, Mozilla is reconsidering their switch to this newest compiler release. At least because Mozilla has an automated benchmarking infrastructure in place, they were able to spot this issue timely themselves rather than waiting for us or others to spot the problem.
Initial thoughts on these major regressions caused by GCC are the Mozilla developers building Firefox with flags for size optimizations rather than speed and possible regressions from code inlining. Mozilla was originally switching to GCC 4.5 due to its newly-introduced plug-in support and the ability to build Firefox with PGO (Profile-Guided Optimizations).
Following the release of GCC 4.5.0 in mid-April we published our own GCC 4.5 benchmarks and GCC vs. Clang/LLVM benchmarks.
10 Comments