`Sys.CPU_NAME` is `"apple-m1"` on Apple M4

[danielwe]

As described in the title:

julia> Sys.CPU_NAME
"apple-m1"

julia> versioninfo()
Julia Version 1.13.0-DEV.467
Commit 41e0eb09baa (2025-04-29 15:47 UTC)
Build Info:
  Official https://julialang.org release
Platform Info:
  OS: macOS (arm64-apple-darwin24.0.0)
  CPU: 14 × Apple M4 Pro
  WORD_SIZE: 64
  LLVM: libLLVM-19.1.7 (ORCJIT, apple-m1)
  GC: Built with stock GC
Threads: 1 default, 1 interactive, 1 GC (on 10 virtual cores)

[ararslan]

My guess is that this is due to the fallback here:

julia/src/processor_arm.cpp

Lines 712 to 727 in ccef01a

	static NOINLINE std::pair<uint32_t,FeatureList<feature_sz>> _get_host_cpu()
	{
	using namespace llvm;
	char buffer[128];
	size_t bufferlen = 128;
	sysctlbyname("machdep.cpu.brand_string",&buffer,&bufferlen,NULL,0);
	StringRef cpu_name(buffer);
	if (cpu_name.find("M1") != StringRef ::npos)
	return std::make_pair((uint32_t)CPU::apple_m1, Feature::apple_m1);
	else if (cpu_name.find("M2") != StringRef ::npos)
	return std::make_pair((uint32_t)CPU::apple_m2, Feature::apple_m2);
	else if (cpu_name.find("M3") != StringRef ::npos)
	return std::make_pair((uint32_t)CPU::apple_m3, Feature::apple_m3);
	else
	return std::make_pair((uint32_t)CPU::apple_m1, Feature::apple_m1);
	}

[oscardssmith]

note that this is a lot worse on x86 since the architecture name goes back a lot further. This leads zen5 on 1.11 to be detected as "generic"

[danielwe]

The fallback is necessary because there is no apple_m4 in the enum:

julia/src/processor_arm.cpp

Lines 168 to 172 in ccef01a

	apple_a16,
	apple_m1,
	apple_m2,
	apple_m3,
	apple_s4,

[gbaraldi]

This is a fine PR if you are interested in working on it. Have to look at what LLVM does here and pattern match. Otherwise I'll get to it in a bit

[giordano]

I guess the question is: why does m4 fall back on m1 and not m3 or m2? Aren't these ISAs supersets of each other?