Update arch.cmake

cpp/cmake/arch.cmake

@@ -1,10 +1,34 @@
-if(WASM)
-    # Disable SLP vectorization on WASM as it's brokenly slow. To give an idea, with this off it still takes
-    # 2m:18s to compile scalar_multiplication.cpp, and with it on I estimate it's 50-100 times longer. I never
-    # had the patience to wait it out...
+# 🛡️ ARCHITECTURE DETECTION: Rely on CMake's built-in variables for robustness.
+# 'EMSCRIPTEN' is automatically set by the emscripten toolchain file.
+if(EMSCRIPTEN OR CMAKE_SYSTEM_NAME MATCHES "Emscripten")
+    # 🚀 WASM OPTIMIZATION
+    # Disable SLP vectorization on WASM as it causes massive compile-time regressions.
+    # (As noted: scalar_multiplication.cpp compile time explodes without this).
+    # -fno-exceptions: Standard for high-perf WASM/Crypto to reduce code size and overhead.
     add_compile_options(-fno-exceptions -fno-slp-vectorize)
-endif()
 
-if(NOT WASM AND NOT APPLE AND NOT ARM)
-    add_compile_options(-march=skylake)
+else()
+    # 🖥️ NATIVE OPTIMIZATION (Linux/macOS/Windows)
+
+    # Check if we are specifically on x86_64 architecture.
+    if(CMAKE_SYSTEM_PROCESSOR MATCHES "x86_64|amd64|AMD64")
+
+        # 🍎 APPLE SILICON / INTEL MAC HANDLING
+        # Apple's Clang usually handles architecture selection well via CMAKE_OSX_ARCHITECTURES.
+        # We explicitly avoid overriding flags on Apple unless necessary to prevent conflicts.
+        if(NOT APPLE)
+            # 🛡️ PORTABILITY FIX: Changed '-march=skylake' to '-march=native'.
+            # '-march=native': Tells the compiler to use all instructions available on THIS machine 
+            # (AVX2, BMI2, ADX, etc.). It works on both Intel and AMD.
+            #
+            # NOTE: If building Docker images for distribution, consider using '-march=x86-64-v3' 
+            # instead of 'native' to ensure binaries work on diverse servers.
+            add_compile_options(-march=native)
+        endif()
+
+    endif()
+
+    # Note: No explicit 'else' needed for ARM (M1/M2/M3 etc.) as '-march=native' 
+    # is implied or handled by the compiler defaults on those platforms usually, 
+    # or CMAKE_OSX_ARCHITECTURES handles it for macOS.
 endif()