On 2019-05-31 07:12 +0100, Ken Moffat via blfs-dev wrote: > The threatened details of the first part of my investigation into > the general subject of tuning (for packages which I normally build > on my desktop systems) are now uploaded to > http://www.linuxfromscratch.org/~ken/tuning/ - this is mostly just a > list of packages and versions, in the order in which I build them. > Currently using '-O2 -march=native' or '-O3 -march=native' where the > package defaults to -O3, and taking steps to ensure that -g is > removed. And a set of notes for differences in how I've achieved > this. > > No, I doubt that -march=native is generally worth using, but it > does show if my CFLAGS or CXXFLAGS are getting used. > > Also, some comments on a few packages, e.g. how to get perl to use > flags, and how anything using qmake will use the flags qt uses, or > has been forced to use. > > My next stage is to look at what should be the first of the cheap > hardening options, -D_FORTIFY_SOURCE=2 : I have applied that to > everything using gcc/g++ from the start of LFS chapter 6 **EXCEPT > GLIBC** with currently one annoyance in c-ares (it wants defines in > CPPFLAGS) and a total failure in boost, which is where I'm currently > stalled.
I'm building my system with "-O3 -march=native -fipa-pta -fgraphite- identity -floop-nest-optimize -falign-functions=32". It's safe for most packages (all packages following the ISO C/C++ standard, theoratically) but also breaks several packages. I'll try adding "-fno-common" next time (building LFS-8.5-rc with Glibc-2.30) and write a note like you. -- http://lists.linuxfromscratch.org/listinfo/blfs-dev FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
