revisit MSVC platform checks for ARM64EC ?

[sezero]

https://techcommunity.microsoft.com/t5/windows-os-platform-blog/getting-to-know-arm64ec-defines-and-intrinsic-functions/ba-p/2957235
https://learn.microsoft.com/en-us/windows/arm/arm64ec
https://learn.microsoft.com/en-us/windows/arm/arm64ec-abi

[sezero]

After a quick'n'dirty search for _M_X64, _M_AMD64 and _M_ARM64, I
generated the draft below. Not even compile-tested, may be wrong
and/or incomplete: Someone with access to proper hardware+compiler
should check && test things. A CI workflow may help too.

In addition, CheckCPUArchitecture.cmake and cmake'ry may need some
similar surgery.

Hope these help.

diff --git a/include/SDL3/SDL_atomic.h b/include/SDL3/SDL_atomic.h
index ece4ac5..512844b 100644
--- a/include/SDL3/SDL_atomic.h
+++ b/include/SDL3/SDL_atomic.h
@@ -250,9 +250,9 @@ typedef void (*SDL_KernelMemoryBarrierFunc)();
     #define SDL_CPUPauseInstruction() __asm__ __volatile__("or 27,27,27");
 #elif (defined(__riscv) && __riscv_xlen == 64)
     #define SDL_CPUPauseInstruction() __asm__ __volatile__(".insn i 0x0F, 0, x0, x0, 0x010");
-#elif defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_X64))
+#elif defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_X64) && !defined(_M_ARM64EC))
     #define SDL_CPUPauseInstruction() _mm_pause()  /* this is actually "rep nop" and not a SIMD instruction. No inline asm in MSVC x86-64! */
-#elif defined(_MSC_VER) && (defined(_M_ARM) || defined(_M_ARM64))
+#elif defined(_MSC_VER) && (defined(_M_ARM) || defined(_M_ARM64) || defined(_M_ARM64EC))
     #define SDL_CPUPauseInstruction() __yield()
 #elif defined(__WATCOMC__) && defined(__386__)
     extern __inline void SDL_CPUPauseInstruction(void);
diff --git a/include/SDL3/SDL_intrin.h b/include/SDL3/SDL_intrin.h
index 5d276d6..2f1e339 100644
--- a/include/SDL3/SDL_intrin.h
+++ b/include/SDL3/SDL_intrin.h
@@ -32,7 +32,7 @@
 
 /* Need to do this here because intrin.h has C++ code in it */
 /* Visual Studio 2005 has a bug where intrin.h conflicts with winnt.h */
-#if defined(_MSC_VER) && (_MSC_VER >= 1500) && (defined(_M_IX86) || defined(_M_X64))
+#if defined(_MSC_VER) && (_MSC_VER >= 1500) && (defined(_M_IX86) || defined(_M_X64) && !defined(_M_ARM64EC))
 #ifdef __clang__
 /* As of Clang 11, '_m_prefetchw' is conflicting with the winnt.h's version,
    so we define the needed '_m_prefetch' here as a pseudo-header, until the issue is fixed. */
@@ -72,7 +72,7 @@ _m_prefetch(void *__P)
 #      include <arm_neon.h>
 #      define __ARM_NEON 1 /* Set __ARM_NEON so that it can be used elsewhere, at compile time */
 #    endif
-#    if defined (_M_ARM64)
+#    if defined (_M_ARM64) || defined(_M_ARM64EC)
 #      define SDL_NEON_INTRINSICS 1
 #      include <arm64intr.h>
 #      include <arm64_neon.h>
@@ -110,7 +110,7 @@ _m_prefetch(void *__P)
 # endif
 #endif
 
-#if defined(__x86_64__) || defined(_M_X64) || defined(__i386__) || defined(_M_IX86)
+#if defined(__x86_64__) || (defined(_M_X64) && !defined(_M_ARM64EC)) || defined(__i386__) || defined(_M_IX86)
 # if ((defined(_MSC_VER) && !defined(_M_X64)) || defined(__MMX__) || defined(SDL_HAS_TARGET_ATTRIBS)) && !defined(SDL_DISABLE_MMX)
 #  define SDL_MMX_INTRINSICS 1
 #  include <mmintrin.h>
@@ -156,6 +156,6 @@ _m_prefetch(void *__P)
 #  define SDL_AVX512F_INTRINSICS 1
 #  include <immintrin.h>
 # endif
-#endif /* defined(__x86_64__) || defined(_M_X64) || defined(__i386__) || defined(_M_IX86) */
+#endif /* defined(__x86_64__) || (defined(_M_X64) && !defined(_M_ARM64EC)) || defined(__i386__) || defined(_M_IX86) */
 
 #endif /* SDL_intrin_h_ */
diff --git a/src/atomic/SDL_spinlock.c b/src/atomic/SDL_spinlock.c
index 25615ac..4d48464 100644
--- a/src/atomic/SDL_spinlock.c
+++ b/src/atomic/SDL_spinlock.c
@@ -32,7 +32,7 @@
 #include <unixlib/local.h>
 #endif
 
-#if defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_X64))
+#if defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_X64)) && !defined(_M_ARM64EC)
 #include <xmmintrin.h>
 #endif
 
@@ -80,7 +80,7 @@ SDL_bool SDL_AtomicTryLock(SDL_SpinLock *lock)
 #elif defined(HAVE_GCC_ATOMICS) || defined(HAVE_GCC_SYNC_LOCK_TEST_AND_SET)
     return __sync_lock_test_and_set(lock, 1) == 0;
 
-#elif defined(_MSC_VER) && (defined(_M_ARM) || defined(_M_ARM64))
+#elif defined(_MSC_VER) && (defined(_M_ARM) || defined(_M_ARM64) || defined(_M_ARM64EC))
     return _InterlockedExchange_acq(lock, 1) == 0;
 
 #elif defined(_MSC_VER)
@@ -185,7 +185,7 @@ void SDL_AtomicUnlock(SDL_SpinLock *lock)
 #if defined(HAVE_GCC_ATOMICS) || defined(HAVE_GCC_SYNC_LOCK_TEST_AND_SET)
     __sync_lock_release(lock);
 
-#elif defined(_MSC_VER) && (defined(_M_ARM) || defined(_M_ARM64))
+#elif defined(_MSC_VER) && (defined(_M_ARM) || defined(_M_ARM64) || defined(_M_ARM64EC))
     _InterlockedExchange_rel(lock, 0);
 
 #elif defined(_MSC_VER)
diff --git a/src/cpuinfo/SDL_cpuinfo.c b/src/cpuinfo/SDL_cpuinfo.c
index bed2a35..ce75697 100644
--- a/src/cpuinfo/SDL_cpuinfo.c
+++ b/src/cpuinfo/SDL_cpuinfo.c
@@ -176,7 +176,7 @@ static int CPU_haveCPUID(void)
         mov     has_CPUID,1         ; We have CPUID support
 done:
     }
-#elif defined(_MSC_VER) && defined(_M_X64)
+#elif defined(_MSC_VER) && defined(_M_X64) && !defined(_M_ARM64EC)
     has_CPUID = 1;
 #elif defined(__sun) && defined(__i386)
     __asm (
@@ -245,7 +245,7 @@ done:
         __asm mov c, ecx \
         __asm mov d, edx                   \
     }
-#elif (defined(_MSC_VER) && defined(_M_X64))
+#elif (defined(_MSC_VER) && defined(_M_X64) && !defined(_M_ARM64EC))
 /* Use __cpuidex instead of __cpuid because ICL does not clear ecx register */
 #define cpuid(func, a, b, c, d)      \
     {                                \
@@ -296,7 +296,7 @@ static void CPU_calcCPUIDFeatures(void)
                             : "=a"(a)
                             : "c"(0)
                             : "%edx");
-#elif defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_X64)) && (_MSC_FULL_VER >= 160040219) /* VS2010 SP1 */
+#elif defined(_MSC_VER) && !defined(_M_ARM64EC) && (defined(_M_IX86) || defined(_M_X64)) && (_MSC_FULL_VER >= 160040219) /* VS2010 SP1 */
                     a = (int)_xgetbv(0);
 #elif (defined(_MSC_VER) && defined(_M_IX86)) || defined(__WATCOMC__)
                     __asm
@@ -441,7 +441,7 @@ static int CPU_haveNEON(void)
    query the OS kernel in a platform-specific way. :/ */
 #ifdef SDL_CPUINFO_DISABLED
     return 0; /* disabled */
-#elif (defined(__WINDOWS__) || defined(__WINRT__) || defined(__GDK__)) && (defined(_M_ARM) || defined(_M_ARM64))
+#elif (defined(__WINDOWS__) || defined(__WINRT__) || defined(__GDK__)) && (defined(_M_ARM) || defined(_M_ARM64) || defined(_M_ARM64EC))
 /* Visual Studio, for ARM, doesn't define __ARM_ARCH. Handle this first. */
 /* Seems to have been removed */
 #ifndef PF_ARM_NEON_INSTRUCTIONS_AVAILABLE
diff --git a/src/stdlib/SDL_mslibc.c b/src/stdlib/SDL_mslibc.c
index 184b7e2..7951abb 100644
--- a/src/stdlib/SDL_mslibc.c
+++ b/src/stdlib/SDL_mslibc.c
@@ -762,7 +762,7 @@ void __declspec(naked) _alloca_probe_16(void)
 
 #endif /* _M_IX86 */
 
-#ifdef _M_ARM64
+#if defined(_M_ARM64) || defined(_M_ARM64EC)
 
 void __chkstk(void);
 void __chkstk() {
diff --git a/src/video/SDL_blit.h b/src/video/SDL_blit.h
index 0903026..fbad01b 100644
--- a/src/video/SDL_blit.h
+++ b/src/video/SDL_blit.h
@@ -24,7 +24,7 @@
 #define SDL_blit_h_
 
 /* pixman ARM blitters are 32 bit only : */
-#if defined(__aarch64__) || defined(_M_ARM64)
+#if defined(__aarch64__) || defined(_M_ARM64) || defined(_M_ARM64EC)
 #undef SDL_ARM_SIMD_BLITTERS
 #undef SDL_ARM_NEON_BLITTERS
 #endif

[slouken]

From the documentation:

When compiling ARM64EC, x64 intrinsic functions are supported and will be translated to ARM64EC code automatically.

I think this means we really don't have to make any SDL code changes to build and run in that environment?

As a test, I followed the instructions to add an ARM64EC target here, and everything built fine.
https://learn.microsoft.com/en-us/windows/arm/arm64ec-build

I don't have an ARM64 device to test on, so maybe someone could verify that?