On Mon, Jul 7, 2025 at 11:34 AM Jaroslav Fojtik via cfarm-users <cfarm-users@lists.tetaneutral.net> wrote: > > [...] > > On 06/07/25 09:54, Jaroslav Fojtik via cfarm-users wrote: > > > Is there any accessible computer with armhf architecture on cfarm? > > armhf isn'r a hardware type, it's the Debian name for their repo > > compiled for the armv7 with hardware VFP math instructions. > I did not know what is it. > > My first question contained this link: > https://launchpadlibrarian.net/802217984/buildlog_ubuntu-questing-armhf.wp2latex_4.15~ds-1 > _BUILDIN > > G.txt.gz > > Ubuntu-Bug: https://bugs.launchpad.net/ubuntu/+source/wp2latex/+bug/2115809 > It is clear that it is meant UBUNTU. > > -------- > > I hope that I do not need this platform anymore. > First answer from Paul was helpful. > > #I am not aware of any armhf hardware. > #However, at least some of the aarch64 systems have the necessary > #cross-compilers to build armhf executables, and can execute them natively. > #A trivial demo on cfarm117 appears below. > > I has been traced defines and it seems that main problem was unalligned > access. > Ability to access unalligned variables is signelized by this macro: > __ARM_FEATURE_UNALIGNED > > #if !defined(NATIVE_ACCESS) && !defined(HI_ENDIAN) && > !defined(__ARM_FEATURE_UNALIGNED) > #define NATIVE_ACCESS > #endif > > It is all I need.
You should ignore __ARM_FEATURE_UNALIGNED, and write portable code. Unaligned access will result in a SIGBUS on ARMv5, if I recall correctly (and __ARM_FEATURE_UNALIGNED should not be defined). However, on later architectures, like ARMv6 or ARMv7, there is sometimes a performance penalty for the unaligned access. Finally, GCC and Clang sanitizers will produce a finding for unaligned access. If you want or need access to an ARMv7 dev board, try a Raspberry Pi 2. The RPI2 is ARMv7. Jeff _______________________________________________ cfarm-users mailing list cfarm-users@lists.tetaneutral.net https://lists.tetaneutral.net/listinfo/cfarm-users
