Use _M_X64 only without mixing with _M_AMD64

[pps83]

No description provided.

[Cyan4973]

Note: CI errors seem unrelated to this PR, and should disappear after rebasing on top of latest dev branch.

[Cyan4973]

_M_X64 seems indeed redundant with _M_AMD64,
so I agree with the simplification keeping only one.

Is there a specific reason you would prefer _M_X64 ?
I was under the impression that _M_AMD64 was more common.

[pps83]

I was under the impression that _M_AMD64 was more common.

No, _M_X64 is more common.

[t-mat]

Some notes for future contributors.

As for MSVC

MACRO Revisited

https://learn.microsoft.com/en-us/archive/blogs/reiley/macro-revisited
_M_AMD64 and _M_X64 are equivalent.

Not sure about icc though.

It proves nothing, but according to sourcegraph.com:

• _M_AMD64 -> 67.7k results
• _M_X64 -> 152.1k results

Also, it seems microsoft/STL uses _M_X64 as a canonical symbol for x86-64.

• https://github.com/microsoft/STL/blob/4119eb8436e2a084d8364509fcec3ee643e6d220/stl/inc/atomic#L214

#elif defined(_M_IX86) || defined(_M_X64) // ^^^ ARM64/ARM64EC/HYBRID_X86_ARM64 / x86/x64 vvv

[pps83]

Not sure about icc though.

if that's old intel compiler, then likely they would blacklist anything mentioning amd :)
new intel compiler uses clang front-end

[pps83]

ping @Cyan4973 seems like this is ok to merge as well