Support build on arm64 hosts (Apple M1)

[HexRabbit]

• add ARCH_ARM64 macro
• use CMAKE_HOST_SYSTEM_PROCESSOR to detect host architecture

[0vercl0k]

Awesome! Do you happen to know the various different possible value for ${CMAKE_HOST_SYSTEM_PROCESSOR} by any chance? Cheers

[HexRabbit]

Cmake's docs says that it use uname -m on GNU, Linux, Cygwin, Android, OpenBSD (arch is an alias) and macOS to get machine architecture.

Inside uname it's using glibc uname() wrapper which further invokes olduname or uname or newuname syscall on linux to get architecture, no matter which one glibc choose, user will get machine architecture from utsname()->machine which inherent from parent namespace.

In init_uts_ns, the machine field is initialized by UTS_MACHINE, it's defined in Makefile and equals to $ARCH, so the value could be any folder name under the arch/ of linux source code, however, some arch/<ARCH>/Makefile may overwrite the UTS_MACHINE value, and there's also COMPAT_UTS_MACHINE which overwrite the machine field before uname syscall returns.

So here's the list: (even though I cannot assure whether I'm right or not..lol)

• aarch64
• aarch64_be
• alpha
• arc
• arm
• arm64 (added from Apple M1)
• armv8b
• armv8l
• csky
• h8300
• hexagon
• i386
• i686
• ia64
• m68k
• microblaze
• mips
• mips64
• nds32
• nios2
• openrisc
• parisc
• parisc64
• powerpc
• ppc
• ppcle
• riscv32
• riscv64
• s390
• s390x
• sh
• sparc
• sparc64
• um
• x86_64
• xtensa

[HexRabbit]

I'm curious is there any dependency that forces you to limit the host architecture?

[0vercl0k]

Awesome! I don't think so at least for little endian architectures. I think this looks good, thank you for sending this out 🙏🏽

[HexRabbit]

@0vercl0k if we check the endianness first, maybe we could take this out?

rp/src/rp/platform.h

Line 11 in 7a2ffb7

#error Platform not supported.

[0vercl0k]

I think I'm more comfortable failing the build if this is an environment that I haven't tried / hasn't been tried. Is there a relevant architecture / platform that is big that you think I am missing? Cheers

…

On Tue, May 10, 2022 at 8:34 AM HexRabbit ***@***.***> wrote: @0vercl0k <https://github.com/0vercl0k> if we check the endianness first, maybe we could take this out? https://github.com/0vercl0k/rp/blob/7a2ffb789c0bf8803b31840304bc66768f56e6cf/src/rp/platform.h#L11 — Reply to this email directly, view it on GitHub <#44 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AALIOROVA7XGAKR67W5HAJDVJJ6ZVANCNFSM5VKBOUUQ> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[HexRabbit]

Yeah, you are right, without testing it could lead to bugs.