Solve the MacOS ARM compilation problem. Changes some options and add…

… new ones.
aous72 · May 1, 2024 · fbd2654 · fbd2654
1 parent 104c8a7
commit fbd2654
Show file tree

Hide file tree

Showing 16 changed files with 537 additions and 249 deletions.
diff --git a/CMakeLists.txt b/CMakeLists.txt
@@ -1,7 +1,7 @@
 cmake_minimum_required(VERSION 3.11.0)
 
 ## Library name/version
-include(ojph_libname.cmake)
+include(ojph_version.cmake)
 
 ## project
 project (openjph VERSION ${OPENJPH_VERSION} DESCRIPTION "Open source implementation of JPH" LANGUAGES CXX)
@@ -11,21 +11,63 @@ set_property(GLOBAL PROPERTY USE_FOLDERS ON)
 # Building OpenJPH
 ################################################################################################
 
+## Target architecture
+# We use the target architecture to help with arranging files in "source_group" commands.
+# The code does not use the results provided by target_arch.cmake, and relies, instead, 
+# on its own logic, which matches that in target_arch.cmake, to identify the architecture
+include(target_arch.cmake)
+target_architecture(OJPH_TARGET_ARCH)
+message(STATUS "CPU Architecture is ${OJPH_TARGET_ARCH}")
+
 ## options
-option(OJPH_DISABLE_INTEL_SIMD "Disables the use of SIMD instructions and associated files" OFF)
-option(OJPH_ENABLE_INTEL_AVX512 "enables the use of AVX512 SIMD instructions and associated files" ON)
 option(BUILD_SHARED_LIBS "Shared Libraries" ON)
 option(OJPH_ENABLE_TIFF_SUPPORT "Enables input and output support for TIFF files" ON)
 option(OJPH_BUILD_TESTS "Enables building test code" OFF)
 option(OJPH_BUILD_EXECUTABLES "Enables building command line executables" ON)
 option(OJPH_BUILD_STREAM_EXPAND "Enables building ojph_stream_expand executable" OFF)
 
+option(OJPH_DISABLE_SIMD "Disables the use of SIMD instructions -- agnostic to architectures" OFF)
+option(OJPH_DISABLE_SSE "Disables the use of SSE SIMD instructions and associated files" OFF)
+option(OJPH_DISABLE_SSE2 "Disables the use of SSE2 SIMD instructions and associated files" OFF)
+option(OJPH_DISABLE_SSSE3 "Disables the use of SSSE3 SIMD instructions and associated files" OFF)
+option(OJPH_DISABLE_SSE4 "Disables the use of SSE4 SIMD instructions and associated files" OFF)
+option(OJPH_DISABLE_AVX "Disables the use of AVX SIMD instructions and associated files" OFF)
+option(OJPH_DISABLE_AVX2 "Disables the use of AVX2 SIMD instructions and associated files" OFF)
+option(OJPH_DISABLE_AVX512 "Disables the use of AVX512 SIMD instructions and associated files" OFF)
+option(OJPH_DISABLE_NEON "Disables the use of NEON SIMD instructions and associated files" OFF)
+
+## options that are being deprecated
+if (DEFINED OJPH_DISABLE_INTEL_SIMD)
+  message(STATUS "OJPH_DISABLE_INTEL_SIMD is being deprecated. Instead, use \"OJPH_DISABLE_SIMD\", "
+                 "which is architecture agnostic. If you do not specify any, the default is "
+                 "OJPH_DISABLE_SIMD=OFF.")
+  set(OJPH_DISABLE_SIMD ${OJPH_DISABLE_INTEL_SIMD})
+  message(STATUS "OJPH_DISABLE_SIMD is set to ${OJPH_DISABLE_SIMD}")  
+  unset(OJPH_DISABLE_INTEL_SIMD)  
+endif()
+if (DEFINED OJPH_ENABLE_INTEL_AVX512)
+  message(STATUS "OJPH_ENABLE_INTEL_AVX512 is being deprecated, use \"OJPH_DISABLE_AVX512\" instead."
+                 "If you do not specify any, the default is OJPH_DISABLE_AVX512=OFF.")
+  if (OJPH_ENABLE_INTEL_AVX512)
+    set(OJPH_DISABLE_AVX512 OFF)
+  else()
+    set(OJPH_DISABLE_AVX512 ON)
+  endif()
+  message(STATUS "OJPH_DISABLE_AVX512 is set to ${OJPH_DISABLE_AVX512}")  
+  unset(OJPH_ENABLE_INTEL_AVX512)
+endif()
+
 ## Setting some of the options if EMSCRIPTEN is the compiler
+# WebAssembly SIMD is treated differently.  The SIMD flags above have no effect on the 
+# use of WASM SIMD.  This is because, for WASM, both non-SIMD and SIMD are required,
+# and therefore two sets of binaries are generated. For CPUs, one binary can carry both 
+# non-SIMD and SIMD, and the program, at run-time, can decide which path to follow, 
+# depending on what CPU instructions are available.
 if(EMSCRIPTEN)
-  set(OJPH_DISABLE_INTEL_SIMD ON)
   set(BUILD_SHARED_LIBS OFF)
   set(OJPH_ENABLE_TIFF_SUPPORT OFF)
   set(OJPH_BUILD_STREAM_EXPAND OFF)
+  set(OJPH_DISABLE_SIMD ON)
 endif()
 
 # This is related to how the timestamp is set for URL downloaded files.
@@ -36,15 +78,16 @@ if (${CMAKE_VERSION} VERSION_GREATER_EQUAL "3.24.0")
   endif()
 endif()
 
+## Added by Michael Smith
 set(CMAKE_CXX_FLAGS_ASAN
   "-fsanitize=address -fno-optimize-sibling-calls -fsanitize-address-use-after-scope -fno-omit-frame-pointer -g -O1"
   CACHE STRING "Flags used by the C++ compiler during AddressSanitizer builds."
   FORCE)
 
-## build type
+## Build type
 if (NOT CMAKE_BUILD_TYPE)
   set(CMAKE_BUILD_TYPE "Release")
-  message( STATUS "To use AddressSanitizer, use \"cmake .. -DCMAKE_BUILD_TYPE=asan\"" )
+  message(STATUS "To use AddressSanitizer, use \"cmake .. -DCMAKE_BUILD_TYPE=asan\"" )
 endif()
 message(STATUS "Building ${CMAKE_BUILD_TYPE}")
 
@@ -64,11 +107,34 @@ if (CMAKE_CXX_COMPILER_ID MATCHES "Clang|GNU")
   )
 endif()
 
-## The option OJPH_DISABLE_INTEL_SIMD and OJPH_ENABLE_INTEL_AVX512
-if (OJPH_DISABLE_INTEL_SIMD)
-  add_compile_definitions(OJPH_DISABLE_INTEL_SIMD)
-elseif (OJPH_ENABLE_INTEL_AVX512)
-  add_compile_definitions(OJPH_ENABLE_INTEL_AVX512)
+## Enhanced instruction options
+if (OJPH_DISABLE_SIMD)
+  add_compile_definitions(OJPH_DISABLE_SIMD)
+else()
+  if(OJPH_DISABLE_SSE)
+    add_compile_definitions(OJPH_DISABLE_SSE)
+  endif()
+  if(OJPH_DISABLE_SSE2)
+    add_compile_definitions(OJPH_DISABLE_SSE2)
+  endif()
+  if(OJPH_DISABLE_SSSE3)
+    add_compile_definitions(OJPH_DISABLE_SSSE3)
+  endif()
+  if(OJPH_DISABLE_SSE4)
+    add_compile_definitions(OJPH_DISABLE_SSE4)
+  endif()
+  if(OJPH_DISABLE_AVX)
+    add_compile_definitions(OJPH_DISABLE_AVX)
+  endif()
+  if(OJPH_DISABLE_AVX2)
+    add_compile_definitions(OJPH_DISABLE_AVX2)
+  endif()
+  if(OJPH_DISABLE_AVX512)
+    add_compile_definitions(OJPH_DISABLE_AVX512)
+  endif()
+  if(OJPH_DISABLE_NEON)
+    add_compile_definitions(OJPH_DISABLE_NEON)
+  endif()
 endif()
 
 ## Build library and applications

diff --git a/docs/compiling.md b/docs/compiling.md
@@ -109,10 +109,8 @@ The scripts, in the ```.devcontainer``` folder, will build a docker image that c
 
 # Compiling for ARM and other platforms #
 
-To compile for platforms where x86_64 SIMD instructions are not supported, such as on ARM, we need to disable SIMD instructions; this can be achieved using
+Compilation should simply work now.  The simple test code I have passes when run on MacOS ARM on GitHub.
 
-    cd build
-    cmake -DCMAKE_BUILD_TYPE=Release \ -DOJPH_DISABLE_INTEL_SIMD=ON ../
-    make
+# Disabling SIMD instructions #
 
-You may need to install libtiff. As I do not have an ARM board, I tested this using QEMU for aarch64 architecture, targeting a Cortex-A57 CPU. The code worked without issues, but because the ARM platform is emulated, the whole process was slow.
+The code now employs the architecture-agnostic option `OJPH_DISABLE_SIMD`, which should include SIMD instructions wherever they are supported.  This can be achieved with `-DOJPH_DISABLE_SIMD=ON` option during CMake configuration.  Individual instruction sets can be disabled; see the options in the main CMakeLists.txt file.
diff --git a/ojph_libname.cmake → ojph_version.cmake b/ojph_libname.cmake → ojph_version.cmake
@@ -1,5 +1,5 @@
 ################################################################################################
-# Generating library name
+# Generating ojph library version number
 ################################################################################################
 
 ############################################################

diff --git a/src/apps/ojph_compress/CMakeLists.txt b/src/apps/ojph_compress/CMakeLists.txt
@@ -29,19 +29,32 @@ if(EMSCRIPTEN)
   target_link_libraries(ojph_compress_simd PRIVATE openjphsimd)
   install(TARGETS ojph_compress_simd DESTINATION bin)
 else()
-  if(NOT OJPH_DISABLE_INTEL_SIMD)
-    add_executable(ojph_compress ${SOURCES} ${OJPH_IMG_IO_SSE4} ${OJPH_IMG_IO_AVX2})
-
-    if (MSVC)
-      set_source_files_properties(${OJPH_IMG_IO_AVX2} PROPERTIES COMPILE_FLAGS "/arch:AVX2")
-    else()
-      set_source_files_properties(${OJPH_IMG_IO_SSE4} PROPERTIES COMPILE_FLAGS -msse4.1)
-      set_source_files_properties(${OJPH_IMG_IO_AVX2} PROPERTIES COMPILE_FLAGS -mavx2)
+  if (NOT OJPH_DISABLE_SIMD)
+    if (("${OJPH_TARGET_ARCH}" MATCHES "OJPH_ARCH_X86_64") OR ("${OJPH_TARGET_ARCH}" MATCHES "OJPH_ARCH_I386"))
+      if (NOT OJPH_DISABLE_SSE4)
+        list(APPEND SOURCES ${OJPH_IMG_IO_SSE4})
+        source_group("others" FILES ${OJPH_IMG_IO_SSE4})
+      endif()
+      if (NOT OJPH_DISABLE_AVX2)
+        list(APPEND SOURCES ${OJPH_IMG_IO_AVX2})
+        source_group("others" FILES ${OJPH_IMG_IO_AVX2})
+      endif()
+
+      # Set compilation flags
+      if (MSVC)
+        set_source_files_properties(${OJPH_IMG_IO_AVX2} PROPERTIES COMPILE_FLAGS "/arch:AVX2")
+      else()
+        set_source_files_properties(${OJPH_IMG_IO_SSE4} PROPERTIES COMPILE_FLAGS -msse4.1)
+        set_source_files_properties(${OJPH_IMG_IO_AVX2} PROPERTIES COMPILE_FLAGS -mavx2)
+      endif()
+    elseif ("${OJPH_TARGET_ARCH}" MATCHES "OJPH_ARCH_ARM")
+
     endif()
-  else()
-    add_executable(ojph_compress ${SOURCES})
+
   endif()
 
+  add_executable(ojph_compress ${SOURCES})
+
   if( USE_TIFF )
     target_link_libraries(ojph_compress PUBLIC openjph ${TIFF_LIBRARIES})
   else()

diff --git a/src/apps/ojph_expand/CMakeLists.txt b/src/apps/ojph_expand/CMakeLists.txt
@@ -29,19 +29,32 @@ if(EMSCRIPTEN)
   target_link_libraries(ojph_expand_simd PRIVATE openjphsimd)
   install(TARGETS ojph_expand_simd DESTINATION bin)    
 else()
-  if(NOT OJPH_DISABLE_INTEL_SIMD)
-    add_executable(ojph_expand ${SOURCES} ${OJPH_IMG_IO_SSE4} ${OJPH_IMG_IO_AVX2})
-
-    if (MSVC)
-      set_source_files_properties(${OJPH_IMG_IO_AVX2} PROPERTIES COMPILE_FLAGS "/arch:AVX2")
-    else()
-      set_source_files_properties(${OJPH_IMG_IO_SSE4} PROPERTIES COMPILE_FLAGS -msse4.1)
-      set_source_files_properties(${OJPH_IMG_IO_AVX2} PROPERTIES COMPILE_FLAGS -mavx2)
+  if (NOT OJPH_DISABLE_SIMD)
+    if (("${OJPH_TARGET_ARCH}" MATCHES "OJPH_ARCH_X86_64") OR ("${OJPH_TARGET_ARCH}" MATCHES "OJPH_ARCH_I386"))
+      if (NOT OJPH_DISABLE_SSE4)
+        list(APPEND SOURCES ${OJPH_IMG_IO_SSE4})
+        source_group("others" FILES ${OJPH_IMG_IO_SSE4})
+      endif()
+      if (NOT OJPH_DISABLE_AVX2)
+        list(APPEND SOURCES ${OJPH_IMG_IO_AVX2})
+        source_group("others" FILES ${OJPH_IMG_IO_AVX2})
+      endif()
+
+      # Set compilation flags
+      if (MSVC)
+        set_source_files_properties(${OJPH_IMG_IO_AVX2} PROPERTIES COMPILE_FLAGS "/arch:AVX2")
+      else()
+        set_source_files_properties(${OJPH_IMG_IO_SSE4} PROPERTIES COMPILE_FLAGS -msse4.1)
+        set_source_files_properties(${OJPH_IMG_IO_AVX2} PROPERTIES COMPILE_FLAGS -mavx2)
+      endif()
+    elseif ("${OJPH_TARGET_ARCH}" MATCHES "OJPH_ARCH_ARM")
+
     endif()
-  else()
-    add_executable(ojph_expand ${SOURCES})
+
   endif()
 
+  add_executable(ojph_expand ${SOURCES})
+
   if( USE_TIFF )
     target_link_libraries(ojph_expand PUBLIC openjph ${TIFF_LIBRARIES})
   else()

diff --git a/src/apps/ojph_stream_expand/CMakeLists.txt b/src/apps/ojph_stream_expand/CMakeLists.txt
@@ -3,28 +3,29 @@
 
 if (OJPH_BUILD_STREAM_EXPAND)
 
-    include_directories(../common)
-    include_directories(../../core/common)
-    set(CMAKE_CXX_STANDARD 14)
-    
-    file(GLOB OJPH_STREAM_EXPAND  "*.cpp" "*.h")
-    file(GLOB OJPH_SOCKETS         "../others/ojph_sockets.cpp")
-    file(GLOB OJPH_SOCKETS_H       "../common/ojph_sockets.h")
-    file(GLOB OJPH_THREADS        "../others/ojph_threads.cpp")
-    file(GLOB OJPH_THREADS_H      "../common/ojph_threads.h")
+  include_directories(../common)
+  include_directories(../../core/common)
+  set(CMAKE_CXX_STANDARD 14)
+
+  file(GLOB OJPH_STREAM_EXPAND  "*.cpp" "*.h")
+  file(GLOB OJPH_SOCKETS         "../others/ojph_sockets.cpp")
+  file(GLOB OJPH_SOCKETS_H       "../common/ojph_sockets.h")
+  file(GLOB OJPH_THREADS        "../others/ojph_threads.cpp")
+  file(GLOB OJPH_THREADS_H      "../common/ojph_threads.h")
 
-    list(APPEND SOURCES ${OJPH_STREAM_EXPAND} ${OJPH_SOCKETS} ${OJPH_SOCKETS_H} ${OJPH_THREADS} ${OJPH_THREADS_H})
+  list(APPEND SOURCES ${OJPH_STREAM_EXPAND} ${OJPH_SOCKETS} ${OJPH_SOCKETS_H} ${OJPH_THREADS} ${OJPH_THREADS_H})
 
-    source_group("main"        FILES ${OJPH_STREAM_EXPAND})
-    source_group("others"      FILES ${OJPH_SOCKETS} ${OJPH_THREADS})
-    source_group("common"      FILES ${OJPH_SOCKETS_H} ${OJPH_THREADS_H})
+  source_group("main"        FILES ${OJPH_STREAM_EXPAND})
+  source_group("others"      FILES ${OJPH_SOCKETS} ${OJPH_THREADS})
+  source_group("common"      FILES ${OJPH_SOCKETS_H} ${OJPH_THREADS_H})
 
-    add_executable(ojph_stream_expand ${SOURCES})
-    if(MSVC)
-        target_link_libraries(ojph_stream_expand PUBLIC openjph ws2_32)
-    else()
-        target_link_libraries(ojph_stream_expand PUBLIC openjph pthread)
-    endif(MSVC)
+  add_executable(ojph_stream_expand ${SOURCES})
+  if(MSVC)
+      target_link_libraries(ojph_stream_expand PUBLIC openjph ws2_32)
+  else()
+      target_link_libraries(ojph_stream_expand PUBLIC openjph pthread)
+  endif(MSVC)
+
+  install(TARGETS ojph_stream_expand DESTINATION bin)
 
-    install(TARGETS ojph_stream_expand DESTINATION bin)
 endif(OJPH_BUILD_STREAM_EXPAND)
diff --git a/src/apps/others/ojph_img_io.cpp b/src/apps/others/ojph_img_io.cpp
@@ -471,39 +471,49 @@ namespace ojph {
         converter = gen_cvrt_32b3c_to_16ub3c_be;
     }
 
-#ifndef OJPH_DISABLE_INTEL_SIMD
+#ifndef OJPH_DISABLE_SIMD
 
-    if (get_cpu_ext_level() >= X86_CPU_EXT_LEVEL_SSE41) {
-      if (bytes_per_sample == 1) {
-        if (num_components == 1) 
-          converter = sse41_cvrt_32b1c_to_8ub1c;
-        else
-          converter = sse41_cvrt_32b3c_to_8ub3c;
-      }
-      else {
-        if (num_components == 1) 
-          converter = sse41_cvrt_32b1c_to_16ub1c_be;
-        else
-          converter = sse41_cvrt_32b3c_to_16ub3c_be;
-      }
-    }
+  #if (defined(OJPH_ARCH_X86_64) || defined(OJPH_ARCH_I386))
 
-    if (get_cpu_ext_level() >= X86_CPU_EXT_LEVEL_AVX2) {
-      if (bytes_per_sample == 1) {
-        if (num_components == 1) 
-          converter = avx2_cvrt_32b1c_to_8ub1c;
-        else
-          converter = avx2_cvrt_32b3c_to_8ub3c;
+    #ifndef OJPH_DISABLE_SSE4
+      if (get_cpu_ext_level() >= X86_CPU_EXT_LEVEL_SSE41) {
+        if (bytes_per_sample == 1) {
+          if (num_components == 1) 
+            converter = sse41_cvrt_32b1c_to_8ub1c;
+          else
+            converter = sse41_cvrt_32b3c_to_8ub3c;
+        }
+        else {
+          if (num_components == 1) 
+            converter = sse41_cvrt_32b1c_to_16ub1c_be;
+          else
+            converter = sse41_cvrt_32b3c_to_16ub3c_be;
+        }
       }
-      else {
-        if (num_components == 1) 
-          converter = avx2_cvrt_32b1c_to_16ub1c_be;
-        else
-          { } // did not find an implementation better than sse41
+    #endif // !OJPH_DISABLE_SSE4
+
+    #ifndef OJPH_DISABLE_AVX2
+      if (get_cpu_ext_level() >= X86_CPU_EXT_LEVEL_AVX2) {
+        if (bytes_per_sample == 1) {
+          if (num_components == 1) 
+            converter = avx2_cvrt_32b1c_to_8ub1c;
+          else
+            converter = avx2_cvrt_32b3c_to_8ub3c;
+        }
+        else {
+          if (num_components == 1) 
+            converter = avx2_cvrt_32b1c_to_16ub1c_be;
+          else
+            { } // did not find an implementation better than sse41
+        }
       }
-    }
+    #endif // !OJPH_DISABLE_AVX2
+
+  #elif defined(OJPH_ARCH_ARM)
+
+  #endif // !(defined(OJPH_ARCH_X86_64) || defined(OJPH_ARCH_I386))
 
-#endif
+#endif // !OJPH_DISABLE_SIMD
   }
 
   ////////////////////////////////////////////////////////////////////////////