Added support for latest Android NDK which uses Clang instead of GCC

CMakeLists.txt

@@ -31,6 +31,7 @@ else()
     endif()
 endif()
 
+
 #
 # Option list
 #
@@ -57,7 +58,6 @@ option(LLAMA_AVX512 "llama: enable AVX512" OFF)
 option(LLAMA_AVX512_VBMI                "llama: enable AVX512-VBMI"                             OFF)
 option(LLAMA_AVX512_VNNI                "llama: enable AVX512-VNNI"                             OFF)
 option(LLAMA_FMA                        "llama: enable FMA"                                     ON)
-
 # in MSVC F16C is implied with AVX2/AVX512
 if (NOT MSVC)
     option(LLAMA_F16C                   "llama: enable F16C"                                    ON)
@@ -114,6 +114,7 @@ endif()
 #
 # Compile flags
 #
+
 set(CMAKE_CXX_STANDARD 11)
 set(CMAKE_CXX_STANDARD_REQUIRED true)
 set(CMAKE_C_STANDARD 11)
@@ -140,7 +141,6 @@ endif()
 
 if (APPLE AND LLAMA_ACCELERATE)
     find_library(ACCELERATE_FRAMEWORK Accelerate)
-
     if (ACCELERATE_FRAMEWORK)
         message(STATUS "Accelerate framework found")
 
@@ -155,14 +155,11 @@ if(LLAMA_BLAS)
     if (LLAMA_STATIC)
         set(BLA_STATIC ON)
     endif()
-
     if ($(CMAKE_VERSION) VERSION_GREATER_EQUAL 3.22)
         set(BLA_SIZEOF_INTEGER 8)
     endif()
-
     set(BLA_VENDOR ${LLAMA_BLAS_VENDOR})
     find_package(BLAS)
-
     if (BLAS_FOUND)
         message(STATUS "BLAS found, Libraries: ${BLAS_LIBRARIES}")
 
@@ -183,7 +180,6 @@ if(LLAMA_CUBLAS)
     cmake_minimum_required(VERSION 3.17)
 
     find_package(CUDAToolkit)
-
     if (CUDAToolkit_FOUND)
         message(STATUS "cuBLAS found")
 
@@ -238,7 +234,6 @@ endif()
 
 if (LLAMA_CLBLAST)
     find_package(CLBlast)
-
     if (CLBlast_FOUND)
         message(STATUS "CLBlast found")
 
@@ -280,6 +275,7 @@ if(LLAMA_ALL_WARNINGS)
             "$<$<COMPILE_LANGUAGE:C>:${c_flags}>"
             "$<$<COMPILE_LANGUAGE:CXX>:${cxx_flags}>"
     )
+
 endif()
 
 if (MSVC)
@@ -293,7 +289,6 @@ endif()
 if (LLAMA_LTO)
     include(CheckIPOSupported)
     check_ipo_supported(RESULT result OUTPUT output)
-
     if (result)
         set(CMAKE_INTERPROCEDURAL_OPTIMIZATION TRUE)
     else()
@@ -305,20 +300,16 @@ endif()
 # TODO: probably these flags need to be tweaked on some architectures
 #       feel free to update the Makefile for your architecture and send a pull request or issue
 message(STATUS "CMAKE_SYSTEM_PROCESSOR: ${CMAKE_SYSTEM_PROCESSOR}")
-
 if (NOT MSVC)
     if (LLAMA_STATIC)
         add_link_options(-static)
-
         if (MINGW)
             add_link_options(-static-libgcc -static-libstdc++)
         endif()
     endif()
-
     if (LLAMA_GPROF)
         add_compile_options(-pg)
     endif()
-
     if (LLAMA_NATIVE)
         add_compile_options(-march=native)
     endif()
@@ -326,47 +317,42 @@ endif()
 
 if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "arm" OR ${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64")
     message(STATUS "ARM detected")
-
     if (MSVC)
         # TODO: arm msvc?
     else()
         if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "aarch64")
             # Apple M1, M2, etc.
             # Raspberry Pi 3, 4, Zero 2 (64-bit)
-            if(NOT DEFINED ANDROID_NDK)
+
+            # latest Android NDK uses CLang now, which does not support the native flag
+            if(DEFINED ANDROID_NDK)
+            else()
                 add_compile_options(-mcpu=native)
             endif()
         endif()
-
         if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "armv6")
             # Raspberry Pi 1, Zero
-            if(NOT DEFINED ANDROID_NDK)
             add_compile_options(-mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access)
         endif()
-        endif()
-
         if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "armv7")
             # Raspberry Pi 2
-            if(NOT DEFINED ANDROID_NDK)
+            add_compile_options(-mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access)
-                add_compile_options(-mfpu=neon-fp-armv8 -mfp16-format=ieee -mno-unaligned-access -funsafe-math-optimizations)
         endif()
-        endif()
-
         if (${CMAKE_SYSTEM_PROCESSOR} MATCHES "armv8")
             # Raspberry Pi 3, 4, Zero 2 (32-bit)
-            if(NOT DEFINED ANDROID_NDK)
             add_compile_options(-mfp16-format=ieee -mno-unaligned-access)
+
+            if(DEFINED ANDROID_NDK)
+            add_compile_options(-march=armv8.4a+dotprod)
             endif()
         endif()
     endif()
 elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "^(x86_64|i686|AMD64)$")
     message(STATUS "x86 detected")
-
     if (MSVC)
         if (LLAMA_AVX512)
             add_compile_options($<$<COMPILE_LANGUAGE:C>:/arch:AVX512>)
             add_compile_options($<$<COMPILE_LANGUAGE:CXX>:/arch:AVX512>)
-
             # MSVC has no compile-time flags enabling specific
             # AVX512 extensions, neither it defines the
             # macros corresponding to the extensions.
@@ -375,7 +361,6 @@ elseif(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^(x86_64|i686|AMD64)$")
                 add_compile_definitions($<$<COMPILE_LANGUAGE:C>:__AVX512VBMI__>)
                 add_compile_definitions($<$<COMPILE_LANGUAGE:CXX>:__AVX512VBMI__>)
             endif()
-
             if (LLAMA_AVX512_VNNI)
                 add_compile_definitions($<$<COMPILE_LANGUAGE:C>:__AVX512VNNI__>)
                 add_compile_definitions($<$<COMPILE_LANGUAGE:CXX>:__AVX512VNNI__>)
@@ -391,28 +376,22 @@ elseif(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^(x86_64|i686|AMD64)$")
         if (LLAMA_F16C)
             add_compile_options(-mf16c)
         endif()
-
         if (LLAMA_FMA)
             add_compile_options(-mfma)
         endif()
-
         if (LLAMA_AVX)
             add_compile_options(-mavx)
         endif()
-
         if (LLAMA_AVX2)
             add_compile_options(-mavx2)
         endif()
-
         if (LLAMA_AVX512)
             add_compile_options(-mavx512f)
             add_compile_options(-mavx512bw)
         endif()
-
         if (LLAMA_AVX512_VBMI)
             add_compile_options(-mavx512vbmi)
         endif()
-
         if (LLAMA_AVX512_VNNI)
             add_compile_options(-mavx512vnni)
         endif()
@@ -420,7 +399,6 @@ elseif(${CMAKE_SYSTEM_PROCESSOR} MATCHES "^(x86_64|i686|AMD64)$")
 elseif (${CMAKE_SYSTEM_PROCESSOR} MATCHES "ppc64")
     message(STATUS "PowerPC detected")
     add_compile_options(-mcpu=native -mtune=native)
-
     #TODO: Add  targets for Power8/Power9 (Altivec/VSX) and Power10(MMA) and query for big endian systems (ppc64/le/be)
 else()
     message(STATUS "Unknown architecture")
@@ -429,6 +407,7 @@ endif()
 #
 # Build libraries
 #
+
 add_library(ggml OBJECT
             ggml.c
             ggml.h
@@ -471,9 +450,11 @@ if(GGML_SOURCES_CUDA)
     set_property(TARGET llama PROPERTY CUDA_ARCHITECTURES OFF)
 endif()
 
+
 #
 # programs, examples and tests
 #
+
 if (LLAMA_BUILD_TESTS AND NOT CMAKE_JS_VERSION)
     include(CTest)
     add_subdirectory(tests)