AUTO for SYMX_ENABLE_AVX2

Author: JoseAntFer

docs/source/compilation.md

@@ -120,7 +120,7 @@ Compiled<__m256d> compiled({f, df_dx}, "my_function_simd4d", codegen_dir, checks
 Compiled<__m256> compiled({f, df_dx}, "my_function_simd8f", codegen_dir, checksum);
 ```
 Note that you will need to set values of type `FLOAT` and that the result will be returned as `View<FLOAT>`.
-If your system does not support these SIMD types, set SymX's the CMake flag `SYMX_ENABLE_AVX2` to `OFF`.
+If your system does not support these SIMD types, disable them with `-DSYMX_ENABLE_AVX2=OFF` (the default `AUTO` already does this on non-x86 targets).
 
 ### Parallel evaluation
 You can evaluate a `Compiled` expression in parallel in the following way:

docs/source/setup.md

@@ -41,17 +41,19 @@ build/examples/examples   # Examples
 
 | Option | Default | Description |
 |---|---|---|
-| `SYMX_ENABLE_AVX2` | `ON` | Enable AVX2 + FMA SIMD paths |
+| `SYMX_ENABLE_AVX2` | `AUTO` | Enable AVX2 + FMA SIMD paths (`AUTO` / `ON` / `OFF`); `AUTO` enables on x86/x86_64/AMD64 |
 | `SYMX_COMPILER_PATH` | `AUTO` | Compiler used for JIT code generation at runtime |
 | `SYMX_CODEGEN_DIR` | *(empty)* | Output directory for generated files; defaults to `<build>/codegen` |
 | `SYMX_HESS_STORAGE_FLOAT` | `float` | Hessian storage precision (`float` or `double`) |
 
 
 ### AVX2 support
 SymX uses AVX2 to speedup computations in several locations, such as evaluation and linear system solves.
-If your system does not support it (e.g. Apple Silicon) use `SYMX_ENABLE_AVX2=OFF` in CMake to disable such code paths.
+`SYMX_ENABLE_AVX2` accepts `AUTO` (default), `ON`, or `OFF`.
+`AUTO` enables AVX2 on x86/x86_64/AMD64 and disables it everywhere else (e.g. Apple Silicon).
+Override explicitly with `-DSYMX_ENABLE_AVX2=ON` or `-DSYMX_ENABLE_AVX2=OFF`.
 
-If you try to compile SymX as-is, without support for AVX2 you will get something like:
+If you try to compile on a non-AVX2 system with `ON`, you will get something like:
 
 ```bash
 ../immintrin.h:14 error "This header is only meant to be used on x86 and x64 architecture"

symx/CMakeLists.txt

@@ -1,7 +1,20 @@
 cmake_minimum_required(VERSION 3.15)
 
 # SymX options
-set(SYMX_ENABLE_AVX2 ON CACHE BOOL "Flag to enable AVX2 for symx")
+set(SYMX_ENABLE_AVX2 "AUTO" CACHE STRING "Enable AVX2+FMA SIMD: AUTO, ON, or OFF. AUTO enables on x86/x86_64/AMD64.")
+set_property(CACHE SYMX_ENABLE_AVX2 PROPERTY STRINGS "AUTO" "ON" "OFF")
+string(TOUPPER "${SYMX_ENABLE_AVX2}" _SYMX_AVX2)
+if(NOT _SYMX_AVX2 MATCHES "^(AUTO|ON|OFF)$")
+    message(FATAL_ERROR "SYMX_ENABLE_AVX2 must be AUTO, ON, or OFF (got: ${SYMX_ENABLE_AVX2})")
+endif()
+if(_SYMX_AVX2 STREQUAL "AUTO")
+    if(CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86_64|X86_64|amd64|AMD64|i[3-6]86)$")
+        set(_SYMX_AVX2 "ON")
+    else()
+        set(_SYMX_AVX2 "OFF")
+    endif()
+endif()
+message(STATUS "SymX: AVX2 = ${_SYMX_AVX2} (SYMX_ENABLE_AVX2=${SYMX_ENABLE_AVX2}, processor=${CMAKE_SYSTEM_PROCESSOR})")
 set(SYMX_COMPILER_PATH "AUTO" CACHE STRING "Path to the compiler to use for runtime compilation. Set to AUTO to auto-detect.")
 set(SYMX_HESS_STORAGE_FLOAT "float" CACHE STRING "Storage type for Hessian matrices (float or double)")
 
@@ -113,7 +126,7 @@ if(OpenMP_CXX_FOUND)
 endif()
 
 # SIMD: when AVX2 is enabled its intrinsics appear in public headers
-if(${SYMX_ENABLE_AVX2})
+if(_SYMX_AVX2 STREQUAL "ON")
     if(CMAKE_CXX_COMPILER_ID STREQUAL "GNU")
         target_compile_options(symx PUBLIC -mavx2 -mfma -Wno-ignored-attributes)
     elseif(CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
@@ -130,7 +143,7 @@ if(MSVC)
 endif()
 
 # Enable/disable the use of AVX2 instructions
-if(${SYMX_ENABLE_AVX2})
+if(_SYMX_AVX2 STREQUAL "ON")
     target_compile_definitions(symx PUBLIC SYMX_ENABLE_AVX2 BSM_ENABLE_AVX2)
 endif()