[compiler-rt] Don't require the XCR0 OSXSAVE bits when detecting skylake-avx512.

[fhahn]

Summary

Darwin lazily saves the AVX512 context on first use: instead of checking
that it already does to figure out if the OS supports AVX512, trust that
the kernel will do the right thing and always assume the context save
support is available.

[craig.topper]

LGTM

compiler-rt/lib/builtins/cpu_model.c

if ((ECX >> 25) & 1)
     setFeature(FEATURE_AES);
 
   // If CPUID indicates support for XSAVE, XRESTORE and AVX, and XGETBV
   // indicates that the AVX registers will be saved and restored on context
   // switch, then we have full AVX support.
   const unsigned AVXBits = (1 << 27) | (1 << 28);
   bool HasAVX = ((ECX & AVXBits) == AVXBits) && !getX86XCR0(&EAX, &EDX) &&
                 ((EAX & 0x6) == 0x6);
+#if defined(__APPLE__)
+  // Darwin lazily saves the AVX512 context on first use: trust that the OS will
+  // save the AVX512 context if we use AVX512 instructions, even the bit is not
+  // set right now.
+  bool HasAVX512Save = true;
+#else
+  // AVX512 requires additional context to be saved by the OS.
   bool HasAVX512Save = HasAVX && ((EAX & 0xe0) == 0xe0);
+#endif
 
   if (HasAVX)
     setFeature(FEATURE_AVX);
 
   bool HasLeaf7 =
       MaxLeaf >= 0x7 && !getX86CpuIDAndInfoEx(0x7, 0x0, &EAX, &EBX, &ECX, &EDX);
 
   if (HasLeaf7 && ((EBX >> 3) & 1))