Whenever I try to compile any C++ program with march=native on a Macbook with a M1 chip, I get the following error when using clang:
clang: error: the clang compiler does not support '-march=native'
However, it used to work on an older Macbook with an Intel CPU. Does clang not support this architecture (yet)?
clang --version gives:
Apple clang version 12.0.0 (clang-1200.0.32.29)
Target: arm64-apple-darwin20.2.0
As far as I know this is not Apple M1 specific, it occurs also with clang for various other architectures (mostly various other arm processors).
e.g. See this bug report here https://github.com/DMOJ/judge-server/issues/303
Basically every build of clang for a new architecture must choose to provide defaults for "march=native" for the target for which the compiler is built or not; and if not then you will see this error message. Even for those arm processors who do have optimisation targets you often have to use specifically "-mcpu=xxx" instead of "-march".
e.g. For the iphone you would use -mcpu=apple-a11 -mcpu=apple-a12 etc.
However no such target has been implemented yet for Apple M1
$ clang --print-supported-cpus
Apple clang version 12.0.5 (clang-1205.0.22.9)
Target: arm64-apple-darwin20.5.0
Thread model: posix
InstalledDir: /Library/Developer/CommandLineTools/usr/bin
Available CPUs for this target:
a64fx
apple-a10
apple-a11
apple-a12
apple-a13
apple-a14
apple-a7
apple-a8
apple-a9
apple-latest
apple-s4
apple-s5
carmel
cortex-a34
cortex-a35
cortex-a53
cortex-a55
cortex-a57
cortex-a65
cortex-a65ae
cortex-a72
cortex-a73
cortex-a75
cortex-a76
cortex-a76ae
cortex-a77
cortex-a78
cortex-x1
cyclone
exynos-m3
exynos-m4
exynos-m5
falkor
generic
kryo
lightning
neoverse-e1
neoverse-n1
saphira
thunderx
thunderx2t99
thunderx3t110
thunderxt81
thunderxt83
thunderxt88
tsv110
vortex
-mcpu=apple-a14 might be the best one for M1.
In Apple clang version 13.0.0, -mcpu=apple-m1 is now available.