[LLVMbugs] [Bug 24326] Auto-detect CPU on native compilation

[bugzilla-daemon at llvm.org]

https://llvm.org/bugs/show_bug.cgi?id=24326

Renato Golin <renato.golin at linaro.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|---                         |INVALID

--- Comment #7 from Renato Golin <renato.golin at linaro.org> ---
(In reply to comment #6)
> When we build on x86, the default targets for i386 and x86-64 are the
> lowest-common-denominator versions of these platforms.  No detected features
> of the host CPU are used without -march=native and the generated binary is
> expected to run on any i386 or x86-64 CPU.  If we set -match=i686 in CFLAGS
> and some build systems ignore it, then this results in slightly suboptimal
> builds, but not in broken packages.

Good point.


> If we're building for ARMv6, we want ARMv6 binaries that will run on any
> ARMv6 or newer CPU, not whatever the host happens to be.

That's not exactly true with Clang (v8A has NEON by default). But that's for
another discussion.


> In addition, deterministic builds are increasingly important for package
> verification and we don't want *anything* other than explicit arguments
> passed to the compiler to affect the binary contents.  Some things are
> unavoidable there, but let's not introduce more.

Yes, that's the best point against my proposal, so I'm marking this bug as
invalid.

Thanks!

-- 
You are receiving this mail because:
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-bugs/attachments/20150731/8bfe2e90/attachment.html>