H1-2018 Was Certainly Eventful For The GCC Compiler
The stable GCC 8.1 debut brought with it initial C++2A support, initial Intel Cannonlake and Icelake CPU enablement, Profile Guided Optimization improvements and other optimization passes work, Intel CET, Qualcomm Saphira CPU support and other ARM CPU improvements, C17 language support, updates for the Go and Fortran languages, AMD HSA IL / BRING improvements, and a whole lot of other work that built up over the past year. GCC 8.1 was officially released in early May while all feature work is now focused on GCC 9 that should debut as stable around the end of Q1'2019.
Among the GCC 9 work that was carried out in the first half of this year included much better GCC bash completion support, dropping Intel MPX, compiler-assisted performance analysis, Ada language support improvements, AMD HSA / BRIG improvements, dropping support for older ARM architectures, and more. Of course, expect much more feature work on GCC 9 to happen during the second half of this year.
For those wondering how the GCC compiler performance is looking when using a SVN snapshot from the end of last year to the end of H1'2018, here's a look at some of the compiler benchmarks where there was a notable difference.
Fhourstones exhibited some small performance improvements on the tested systems.
SciMark2 saw slight improvements on a subset of the tested Linux systems.
Most tests though saw none or very minimal performance changes with the GCC code churn over the past six months.
The AVX-heavy Himeno saw some nice performance improvements on the Intel Core systems but Ryzen appearing to regress.
The Ebizzy server synthetic benchmark improved with the Xeon rig.
The C-Ray ray-tracer was running faster with the newer GCC code.
Over on LinuxBenchmarking.com we continue to put out new GCC benchmarks of the tests being built under the latest daily GCC SVN code on a bi-daily basis.
It will certainly be interesting to see what the GNU toolchain developers have in store for H2-2018!
3 Comments
