Trying Link-Time Optimizations On GCC 4.10
This weekend I ran some quick and dirty link-time optimization (LTO) benchmarks from a GCC 4.10 compiler snapshot from earlier this month. Here's the results.
For those unfamiliar with GCC's Link-Time Optimization abilities, read this earlier article and benchmarks from two years ago when testing GCC 4.7 after its LTO abilities were into shape. LTO has the ability to yield faster binaries of large programs by applying optimizations at link-time across multiple source files, but at the cost of longer build times and using greater system resources. With GNU developers continuing to optimize the compiler's LTO performance, out of curiosity I ran some quick benchmarks this weekend of a 4.10 snapshot.
With just being a quick, one-page article, the results aren't too particularly exciting and mostly a data dump of some benchmark numbers, but as GCC 4.10 continues to be developed over the months ahead and gains new abilities ahead of its release that will most likely come in H1'2015, further benchmarks will materialize on Phoronix.
While LTO holds lots of hope, don't get too excited as it won't yield performance improvements for all applications.
For many applications, the performance improvements of LTO are rather small.
The build times are certainly longer...
Find more of these GCC 4.10 LTO tests via this result file but they are rather basic and not particularly exciting -- the more interesting GCC 4.10 compiler tests will come later with a proper, multi-page write-up.
For those unfamiliar with GCC's Link-Time Optimization abilities, read this earlier article and benchmarks from two years ago when testing GCC 4.7 after its LTO abilities were into shape. LTO has the ability to yield faster binaries of large programs by applying optimizations at link-time across multiple source files, but at the cost of longer build times and using greater system resources. With GNU developers continuing to optimize the compiler's LTO performance, out of curiosity I ran some quick benchmarks this weekend of a 4.10 snapshot.
With just being a quick, one-page article, the results aren't too particularly exciting and mostly a data dump of some benchmark numbers, but as GCC 4.10 continues to be developed over the months ahead and gains new abilities ahead of its release that will most likely come in H1'2015, further benchmarks will materialize on Phoronix.
While LTO holds lots of hope, don't get too excited as it won't yield performance improvements for all applications.
For many applications, the performance improvements of LTO are rather small.
The build times are certainly longer...
Find more of these GCC 4.10 LTO tests via this result file but they are rather basic and not particularly exciting -- the more interesting GCC 4.10 compiler tests will come later with a proper, multi-page write-up.
30 Comments