[X86] Add MSVC-compatible intrinsics for clac, stac, lgdt and sgdt

[mkuper]

Summary

LLVM part was posted as D12277

[compnerd]

lib/Headers/Intrin.h

961

Why does this need a builtin? Is an inline assembly block using lgdt insufficient for some reason?

965

Similar to lgdt.

[mkuper]

lib/Headers/Intrin.h

961

I think using a builtin is, generally, cleaner.
I'm ok with using inline asm (and abandoning the LLVM part of the patch), if that's the more popular option.

[compnerd]

lib/Headers/Intrin.h

961

Yes, that is the preference in my experience. Please do switch to the inline asm option.

[rnk]

lib/Headers/Intrin.h

961

I guess the distinction between these operations and others is that it isn't useful for the compiler to try to reason about these instructions, in the way that it's useful for it to reason about vector intrinsics. I'm happy with either LLVM intrinsics or inline asm.

include/clang/Basic/BuiltinsX86.def

 TARGET_BUILTIN(__builtin_ia32_wrgsbase64, "vULLi", "", "fsgsbase")
 
 // FXSR
 BUILTIN(__builtin_ia32_fxrstor, "vv*", "")
 BUILTIN(__builtin_ia32_fxrstor64, "vv*", "")
 BUILTIN(__builtin_ia32_fxsave, "vv*", "")
 BUILTIN(__builtin_ia32_fxsave64, "vv*", "")
 
+// SMAP
+BUILTIN(__builtin_ia32_clac, "v", "")
+BUILTIN(__builtin_ia32_stac, "v", "")
+
+// LGDT, SGDT
+BUILTIN(__builtin_ia32_sgdt, "vv*", "")
+BUILTIN(__builtin_ia32_lgdt, "vv*", "")
+
 // ADX
 TARGET_BUILTIN(__builtin_ia32_addcarryx_u32, "UcUcUiUiUi*", "", "adx")
 TARGET_BUILTIN(__builtin_ia32_addcarryx_u64, "UcUcULLiULLiULLi*", "", "adx")
 TARGET_BUILTIN(__builtin_ia32_addcarry_u32, "UcUcUiUiUi*", "", "adx")
 TARGET_BUILTIN(__builtin_ia32_addcarry_u64, "UcUcULLiULLiULLi*", "", "adx")
 TARGET_BUILTIN(__builtin_ia32_subborrow_u32, "UcUcUiUiUi*", "", "adx")
 TARGET_BUILTIN(__builtin_ia32_subborrow_u64, "UcUcULLiULLiULLi*", "", "adx")
 

lib/Headers/Intrin.h

 float _m_to_float(__m64);
 int _m_to_int(__m64 _M);
 #endif
 
 /* Other assorted instruction intrinsics. */
 void __addfsbyte(unsigned long, unsigned char);
 void __addfsdword(unsigned long, unsigned long);
 void __addfsword(unsigned long, unsigned short);
+static __inline__ void _clac(void);
 void __code_seg(const char *);
 static __inline__
 void __cpuid(int[4], int);
 static __inline__
 void __cpuidex(int[4], int, int);
 void __debugbreak(void);
 __int64 __emul(int, int);
 unsigned __int64 __emulu(unsigned int, unsigned int);
 void __cdecl __fastfail(unsigned int);
 unsigned int __getcallerseflags(void);
 static __inline__
 void __halt(void);
 unsigned char __inbyte(unsigned short);
 void __inbytestring(unsigned short, unsigned char *, unsigned long);
 void __incfsbyte(unsigned long);
 void __incfsdword(unsigned long);
 void __incfsword(unsigned long);
 unsigned long __indword(unsigned short);
 void __indwordstring(unsigned short, unsigned long *, unsigned long);
 void __int2c(void);
 void __invlpg(void *);
 unsigned short __inword(unsigned short);
 void __inwordstring(unsigned short, unsigned short *, unsigned long);
+static __inline__ void _lgdt(void *);
 void __lidt(void *);
 unsigned __int64 __ll_lshift(unsigned __int64, int);
 __int64 __ll_rshift(__int64, int);
 void __llwpcb(void *);
 unsigned char __lwpins32(unsigned int, unsigned int, unsigned int);
 void __lwpval32(unsigned int, unsigned int, unsigned int);
 unsigned int __lzcnt(unsigned int);
 unsigned short __lzcnt16(unsigned short);

 unsigned __int64 __readfsqword(unsigned long);
 static __inline__
 unsigned short __readfsword(unsigned long);
 #endif
 static __inline__
 unsigned __int64 __readmsr(unsigned long);
 unsigned __int64 __readpmc(unsigned long);
 unsigned long __segmentlimit(unsigned long);
+static __inline__ void _sgdt(void *);
 void __sidt(void *);
 void *__slwpcb(void);
 static __inline__
+static __inline__ void _stac(void);
 void __stosb(unsigned char *, unsigned char, size_t);
 static __inline__
 void __stosd(unsigned long *, unsigned long, size_t);
 static __inline__
 void __stosw(unsigned short *, unsigned short, size_t);
 void __svm_clgi(void);
 void __svm_invlpga(void *, int);
 void __svm_skinit(int);
__readcr3(void) {
   __asm__ __volatile__ ("mov %%cr3, %0" : "=q"(__cr3_val) : : "memory");
   return __cr3_val;
 }
 
 static __inline__ void __DEFAULT_FN_ATTRS
 __writecr3(unsigned int __cr3_val) {
   __asm__ ("mov %0, %%cr3" : : "q"(__cr3_val) : "memory");
 }
+
+static __inline__ void __DEFAULT_FN_ATTRS _clac(void) {
+  __builtin_ia32_clac();
+}
+
+static __inline__ void __DEFAULT_FN_ATTRS _stac(void) {
+  __builtin_ia32_stac();
+}
+
+static __inline__ void __DEFAULT_FN_ATTRS _lgdt(void *__ptr) {
+  __builtin_ia32_lgdt(__ptr);
+}
+
+static __inline__ void __DEFAULT_FN_ATTRS _sgdt(void *__ptr) {
+  __builtin_ia32_sgdt(__ptr);
+}
+
+
 #endif
 
 #ifdef __cplusplus
 }
 #endif
 
 #undef __DEFAULT_FN_ATTRS
 
 #endif /* __INTRIN_H */
 #endif /* _MSC_VER */

test/CodeGen/builtins-x86.c

#endif
   tmp_i = __builtin_ia32_vec_ext_v2si(tmp_V2i, 0);
 
   (void) __builtin_ia32_ldmxcsr(tmp_Ui);
   tmp_Ui = __builtin_ia32_stmxcsr();
   (void)__builtin_ia32_fxsave(tmp_vp);
   (void)__builtin_ia32_fxsave64(tmp_vp);
   (void)__builtin_ia32_fxrstor(tmp_vp);
   (void)__builtin_ia32_fxrstor64(tmp_vp);
+  (void)__builtin_ia32_clac();
+  (void)__builtin_ia32_stac();
+  (void)__builtin_ia32_sgdt(tmp_vp);
+  (void)__builtin_ia32_lgdt(tmp_vp);
   tmp_V4f = __builtin_ia32_cvtpi2ps(tmp_V4f, tmp_V2i);
   tmp_V2i = __builtin_ia32_cvtps2pi(tmp_V4f);
   tmp_i = __builtin_ia32_cvtss2si(tmp_V4f);
 
   tmp_i = __builtin_ia32_rdtsc();
   tmp_i = __builtin_ia32_rdtscp(&tmp_Ui);
   tmp_LLi = __builtin_ia32_rdpmc(tmp_i);
 #ifdef USE_64