I've now completed a real "tuning" run of my experiment (i.e. specifying -falign-functions=32 -malign-data=cacheline because these have been said to be beneficial for haswell processors, which is what I have been using.
Summary in http://www.linuxfromscratch.org/~ken/tuning/tuning-4-alignment-tests.txt and updated spreadsheet of the tests I was able to run 10 times on the different builds: (desktop-runtime-comparisons.ods in the same directory). Conclusion for this experiment: case not proven. I hope to get one more run done ("turn all the optimization up to -O3") but I will not be playing with alignment again. Also, I had hoped to examine LTO (and I might take a stab at the first part of that - my previous attempt at recompiling gcc was a disaster in its testsuite, but maybe that was because I had not understood the problem of using -D_FORTIFY_SOURCE=2 on gcc at that point). However, the big gains touted for LTO (smaller programs) come at a compile time cost - for those of us who frequently rebuild I'm not at all sure that the cost will be worthwhile. ĸen -- This is magic for grown-ups; it has to be hard because we know there's no such thing as a free goblin. -- Pratchett, Stewart & Cohen - The Science of Discworld II -- http://lists.linuxfromscratch.org/listinfo/blfs-dev FAQ: http://www.linuxfromscratch.org/blfs/faq.html Unsubscribe: See the above information page
