Mythtv 0.27_p20130924 fails to finish the compile phase with CFLAGS="-O2 -pipe -march=bdver1 --param l1-cache-size=16 --param l1-cache-line-size=64 --param l2-cache-size=2048", but is fine with CFLAGS="-O2 -pipe -march=amdfam10" Also, mythtv 0.26-fixes ebuilds are fine with -march=bdver1 Please let me know what other information is needed to Reproducible: Always
Created attachment 362316 [details] emerge --info '=media-tv/mythtv-0.27_p20130924::gentoo'
Created attachment 362318 [details] build log # emerge -pqv '=media-tv/mythtv-0.27_p20130924::gentoo' [ebuild R ] media-tv/mythtv-0.27_p20130924 USE="alsa dvd hls ieee1394 jack libass perl python theora vaapi vdpau vorbis xml xvid (-altivec) -autostart -bluray -cec -crystalhd -debug -dvb -egl -fftw -lcd -lirc -pulseaudio -raop -xmltv" INPUT_DEVICES="-joystick"
so, apparently this is the only ebuild on my system that does not handle '--param l1-cache-size=16 --param l1-cache-line-size=64 --param l2-cache-size=2048' in some sane manner.
(In reply to kisak42 from comment #0) > Mythtv 0.27_p20130924 fails to finish the compile phase with CFLAGS="-O2 > -pipe -march=bdver1 --param l1-cache-size=16 --param l1-cache-line-size=64 > --param l2-cache-size=2048", but is fine with CFLAGS="-O2 -pipe > -march=amdfam10" > > Also, mythtv 0.26-fixes ebuilds are fine with -march=bdver1 It's caused by `--param[ space ]arg', not by -march=.
Thank you for adjusting the summary Mr. Roovers.
*** Bug 496046 has been marked as a duplicate of this bug. ***
I would firstly recommend if you really want to nail down your hardware you use something like --march=native. But due to the way they pass around CFLAGS in their build you'd really have to take this upstream but you can adjust your CFLAGS by replacing the space with an = sign. --param=l1-cache-line-size=64 should work.
Good day Mr. Goldstein, Since this bug report was last touched the default GCC version for gentoo stable has been bumped and these param options are applied when using -march=bdver1, new users should not encounter this issue unless they follow the safe cflags at http://gentoo-en.vfose.ru/wiki/Safe_Cflags/AMD#FX-8xxx.2F6xxx.2F4xxx_.28Bulldozer.29 , which relevant to gcc 4.6.x, but not 4.8.x. The safe cflag page at https://wiki.gentoo.org/wiki/Safe_CFLAGS#FX-XXXX is up to date as of this writing. (advising users to use CFLAGS="-O2 -march=bdver1 -mprefer-avx128 -mvzeroupper -pipe" on chips like my system) I am going to close this issue report at this time as resolved/worksforme.