Re: [PATCH] x86/build/64: Prevent native builds from generating APX instructions
David Laight <[email protected]>
| Newsgroups | org.kernel.vger.stable,org.kernel.vger.linux-kernel |
|---|---|
| Message-ID | <20260802100808.754400af@pumpkin> |
On Sat, 1 Aug 2026 17:00:36 -0700 "Chang S. Bae" <[email protected]> wrote: > On 8/1/2026 4:44 AM, David Laight wrote: > > > > Much like when AVX support was added, -march-native needs to check that the > > operating system is saving the registers not that the cpu supports them. > > This GCC bug looks relevant: > > "__builtin_cpu_supports avx does not verify OS supports it" > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=85100 Yes, except I thought it was much older... > The resulting commit 059cc8aca774 ("i386: Enable AVX/AVX512 features > only if supported by OSXSAVE") switched the compiler to read XCR0 before > set_feature(). > > Similarly, the APX enablement commit e686416b6217 ("[APX_EGPR] Initial > support for APX_F") extends the XCR0 check to determine whether APX is > usable. So, the compiler seemingly won't emit APX instructions without > OS setting XCR0.APX. Does that mean that this patch is completely unnecessary? David > > Thanks, > Chang