> > Hello, Frediano. > > Option "-mno-unaligned-access" doesn't help. Got a lot of "Alignment trap". > No additional code such as __aeabi_uread4() has added by the compiler. No > runtime checks. Looks like it just ignored. GCC 4.8.2. > > http://sourceforge.net/p/tianocore/edk2/ci/68085d37f88aaaaad97f071e2d0e60b276592cc5/ > > Like stated here: > > http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.faqs/ka15414.html > > Would globally enabled aligned reads decrease an overall performance? > Hi, sorry, I looked more carefully at generated code and the current code is not enough to make the -mno-unaligned-access option works as expected. It does not globally enable aligned read access but only when not aligned (or better when alignment is not expected). I'll post a patch to solve this issue. Frediano > 27.11.2015, 16:56, "Frediano Ziglio" <fziglio@xxxxxxxxxx>: > >> > >> Anton, which architecture it's your machine. Could be that using some > >> gcc options would make the compiler generate only aligned access > >> instructions. > > > > Can you try to add -mno-unaligned-access option to your CFLAGS ? > > > > Frediano > > -- > Anton D. Kachalov > > ITO, System Architect > Tel: 7 (495) 739-70-00 ext.7613 > _______________________________________________ Spice-devel mailing list Spice-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/spice-devel