[ARM64EC] Fix compilation of intrin.h in ARM64EC mode. (#87717)

intrin.h checks for x86_64. But the "x86_64" define is also defined for
ARM64EC, and we don't support all the intrinsics in ARM64EC mode. Fix
the preprocessor checks to handle this correctly. (If we actually need
some of these intrinsics in ARM64EC mode, we can revisit later.)

Not exactly sure how I didn't run into this issue before now... I think
I've built code that requires these headers, but maybe not since the
define fix landed.

Author: efriedma-quic

clang/lib/Headers/intrin.h

@@ -18,7 +18,7 @@
 #include <intrin0.h>
 
 /* First include the standard intrinsics. */
-#if defined(__i386__) || defined(__x86_64__)
+#if defined(__i386__) || (defined(__x86_64__) && !defined(__arm64ec__))
 #include <x86intrin.h>
 #endif
 
@@ -166,7 +166,7 @@ unsigned __int32 xbegin(void);
 void _xend(void);
 
 /* These additional intrinsics are turned on in x64/amd64/x86_64 mode. */
-#ifdef __x86_64__
+#if defined(__x86_64__) && !defined(__arm64ec__)
 void __addgsbyte(unsigned long, unsigned char);
 void __addgsdword(unsigned long, unsigned long);
 void __addgsqword(unsigned long, unsigned __int64);
@@ -236,7 +236,8 @@ __int64 _mul128(__int64, __int64, __int64 *);
 /*----------------------------------------------------------------------------*\
 |* movs, stos
 \*----------------------------------------------------------------------------*/
-#if defined(__i386__) || defined(__x86_64__)
+
+#if defined(__i386__) || (defined(__x86_64__) && !defined(__arm64ec__))
 static __inline__ void __DEFAULT_FN_ATTRS __movsb(unsigned char *__dst,
                                                   unsigned char const *__src,
                                                   size_t __n) {
@@ -305,7 +306,7 @@ static __inline__ void __DEFAULT_FN_ATTRS __stosw(unsigned short *__dst,
                        : "memory");
 }
 #endif
-#ifdef __x86_64__
+#if defined(__x86_64__) && !defined(__arm64ec__)
 static __inline__ void __DEFAULT_FN_ATTRS __movsq(
     unsigned long long *__dst, unsigned long long const *__src, size_t __n) {
   __asm__ __volatile__("rep movsq"
@@ -324,7 +325,7 @@ static __inline__ void __DEFAULT_FN_ATTRS __stosq(unsigned __int64 *__dst,
 /*----------------------------------------------------------------------------*\
 |* Misc
 \*----------------------------------------------------------------------------*/
-#if defined(__i386__) || defined(__x86_64__)
+#if defined(__i386__) || (defined(__x86_64__) && !defined(__arm64ec__))
 static __inline__ void __DEFAULT_FN_ATTRS __halt(void) {
   __asm__ volatile("hlt");
 }
@@ -339,7 +340,7 @@ static __inline__ void __DEFAULT_FN_ATTRS __nop(void) {
 /*----------------------------------------------------------------------------*\
 |* MS AArch64 specific
 \*----------------------------------------------------------------------------*/
-#if defined(__aarch64__)
+#if defined(__aarch64__) || defined(__arm64ec__)
 unsigned __int64 __getReg(int);
 long _InterlockedAdd(long volatile *Addend, long Value);
 __int64 _InterlockedAdd64(__int64 volatile *Addend, __int64 Value);
@@ -383,7 +384,7 @@ void __cdecl __prefetch(void *);
 /*----------------------------------------------------------------------------*\
 |* Privileged intrinsics
 \*----------------------------------------------------------------------------*/
-#if defined(__i386__) || defined(__x86_64__)
+#if defined(__i386__) || (defined(__x86_64__) && !defined(__arm64ec__))
 static __inline__ unsigned __int64 __DEFAULT_FN_ATTRS
 __readmsr(unsigned long __register) {
   // Loads the contents of a 64-bit model specific register (MSR) specified in
@@ -397,7 +398,6 @@ __readmsr(unsigned long __register) {
   __asm__ ("rdmsr" : "=d"(__edx), "=a"(__eax) : "c"(__register));
   return (((unsigned __int64)__edx) << 32) | (unsigned __int64)__eax;
 }
-#endif
 
 static __inline__ unsigned __LPTRINT_TYPE__ __DEFAULT_FN_ATTRS __readcr3(void) {
   unsigned __LPTRINT_TYPE__ __cr3_val;
@@ -413,6 +413,7 @@ static __inline__ void __DEFAULT_FN_ATTRS
 __writecr3(unsigned __INTPTR_TYPE__ __cr3_val) {
   __asm__ ("mov {%0, %%cr3|cr3, %0}" : : "r"(__cr3_val) : "memory");
 }
+#endif
 
 #ifdef __cplusplus
 }

clang/lib/Headers/intrin0.h

@@ -15,7 +15,7 @@
 #ifndef __INTRIN0_H
 #define __INTRIN0_H
 
-#ifdef __x86_64__
+#if defined(__x86_64__) && !defined(__arm64ec__)
 #include <adcintrin.h>
 #endif
 
@@ -27,7 +27,7 @@ unsigned char _BitScanForward(unsigned long *_Index, unsigned long _Mask);
 unsigned char _BitScanReverse(unsigned long *_Index, unsigned long _Mask);
 void _ReadWriteBarrier(void);
 
-#if defined(__aarch64__)
+#if defined(__aarch64__) || defined(__arm64ec__)
 unsigned int _CountLeadingZeros(unsigned long);
 unsigned int _CountLeadingZeros64(unsigned _int64);
 unsigned char _InterlockedCompareExchange128_acq(__int64 volatile *_Destination,
@@ -44,7 +44,7 @@ unsigned char _InterlockedCompareExchange128_rel(__int64 volatile *_Destination,
                                                  __int64 *_ComparandResult);
 #endif
 
-#ifdef __x86_64__
+#ifdef __x86_64__ && !defined(__arm64ec__)
 unsigned __int64 _umul128(unsigned __int64, unsigned __int64,
                           unsigned __int64 *);
 unsigned __int64 __shiftleft128(unsigned __int64 _LowPart,
@@ -55,7 +55,7 @@ unsigned __int64 __shiftright128(unsigned __int64 _LowPart,
                                  unsigned char _Shift);
 #endif
 
-#if defined(__x86_64__) || defined(__i386__)
+#if defined(__i386__) || (defined(__x86_64__) && !defined(__arm64ec__))
 void _mm_pause(void);
 #endif
 
@@ -83,7 +83,7 @@ __int64 _InterlockedXor64(__int64 volatile *_Value, __int64 _Mask);
 __int64 _InterlockedAnd64(__int64 volatile *_Value, __int64 _Mask);
 #endif
 
-#if defined(__arm__) || defined(__aarch64__)
+#if defined(__arm__) || defined(__aarch64__) || defined(__arm64ec__)
 /*----------------------------------------------------------------------------*\
 |* Interlocked Exchange Add
 \*----------------------------------------------------------------------------*/

clang/test/Headers/ms-intrin.cpp

@@ -18,6 +18,16 @@
 // RUN:     -ffreestanding -fsyntax-only -Werror \
 // RUN:     -isystem %S/Inputs/include %s
 
+// RUN: %clang_cc1 -triple aarch64--windows \
+// RUN:     -fms-compatibility -fms-compatibility-version=17.00 \
+// RUN:     -ffreestanding -fsyntax-only -Werror \
+// RUN:     -isystem %S/Inputs/include %s
+
+// RUN: %clang_cc1 -triple arm64ec--windows \
+// RUN:     -fms-compatibility -fms-compatibility-version=17.00 \
+// RUN:     -ffreestanding -fsyntax-only -Werror \
+// RUN:     -isystem %S/Inputs/include %s
+
 // REQUIRES: x86-registered-target
 
 // intrin.h needs size_t, but -ffreestanding prevents us from getting it from
@@ -41,15 +51,15 @@ void f() {
   __stosd(0, 0, 0);
   __stosw(0, 0, 0);
 
-#ifdef _M_X64
+#if defined(_M_X64) && !defined(_M_ARM64EC)
   __movsq(0, 0, 0);
   __stosq(0, 0, 0);
 #endif
 
   int info[4];
   __cpuid(info, 0);
   __cpuidex(info, 0, 0);
-#if defined(_M_X64) || defined(_M_IX86)
+#if (defined(_M_X64) && !defined(_M_ARM64EC)) || defined(_M_IX86)
   _xgetbv(0);
 #endif
   __halt();