compiles with hacks, many fixmes

Author: mike dupont

CMakeLists.txt

@@ -19,7 +19,9 @@ set(LLAMA_FMA OFF)
 set(LLAMA_F16C OFF)
 set(CMAKE_CUDA_FLAGS "--verbose") #
 set(CMAKE_CUDA_ARCHITECTURES "60;61;70") # needed for f16 CUDA intrinsics
-set(CUDACXX /usr/local/cuda-12.2/bin/nvcc)
+set(CUDACXX /usr/local/cuda-12.3/bin/nvcc)
+set(CMAKE_CUDA_COMPILER /usr/local/cuda-12.3/bin/nvcc)
+set(CUDA_TOOLKIT_ROOT_DIR /usr/local/cuda-12.3)
 #GGML_USE_CUBLAS
 
 #set(CMAKE_EXE_LINKER_FLAGS -pg)
@@ -128,7 +130,7 @@ option(LLAMA_BUILD_SERVER               "llama: build server example"
 # Compile flags
 #
 
-set(CMAKE_CXX_STANDARD 11)
+set(CMAKE_CXX_STANDARD 17)
 set(CMAKE_CXX_STANDARD_REQUIRED true)
 set(CMAKE_C_STANDARD 11)
 set(CMAKE_C_STANDARD_REQUIRED true)
@@ -420,14 +422,15 @@ endif()
 
 if (LLAMA_ALL_WARNINGS)
     if (NOT MSVC)
-        set(warning_flags -Wall -Wextra -Wpedantic -Wcast-qual -Wno-unused-function)
+    # -Wpedantic
+        set(warning_flags -Wall -Wextra  -Wcast-qual -Wno-unused-function)
         set(c_flags -Wshadow -Wstrict-prototypes -Wpointer-arith -Wmissing-prototypes -Werror=implicit-int -Werror=implicit-function-declaration)
-        set(cxx_flags -Wmissing-declarations -Wmissing-noreturn)
+        set(cxx_flags -Wmissing-declarations -Wmissing-noreturn -fpermissive)
         set(host_cxx_flags "")
 
         if (CMAKE_C_COMPILER_ID MATCHES "Clang")
             set(warning_flags ${warning_flags} -Wunreachable-code-break -Wunreachable-code-return)
-            set(host_cxx_flags ${host_cxx_flags} -Wmissing-prototypes -Wextra-semi)
+            set(host_cxx_flags ${host_cxx_flags} -Wmissing-prototypes -Wextra-semi -fpermissive)
 
             if (
                 (CMAKE_C_COMPILER_ID STREQUAL "Clang"      AND CMAKE_C_COMPILER_VERSION VERSION_GREATER_EQUAL 3.8.0) OR
@@ -437,21 +440,21 @@ if (LLAMA_ALL_WARNINGS)
             endif()
         elseif (CMAKE_C_COMPILER_ID STREQUAL "GNU")
             set(c_flags ${c_flags} -Wdouble-promotion)
-            set(host_cxx_flags ${host_cxx_flags} -Wno-array-bounds)
+            set(host_cxx_flags ${host_cxx_flags} -Wno-array-bounds -fpermissive)
 
             if (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL 7.1.0)
-                set(host_cxx_flags ${host_cxx_flags} -Wno-format-truncation)
+                set(host_cxx_flags ${host_cxx_flags} -Wno-format-truncation -fpermissive)
             endif()
             if (CMAKE_CXX_COMPILER_VERSION VERSION_GREATER_EQUAL 8.1.0)
-                set(host_cxx_flags ${host_cxx_flags} -Wextra-semi)
+                set(host_cxx_flags ${host_cxx_flags} -Wextra-semi -fpermissive)
             endif()
         endif()
     else()
         # todo : msvc
     endif()
 
     set(c_flags   ${c_flags}  -save-temps --verbose  ${warning_flags})
-    set(cxx_flags ${cxx_flags}  -save-temps --verbose ${warning_flags})
+    set(cxx_flags ${cxx_flags} -fpermissive  -save-temps --verbose ${warning_flags})
     add_compile_options("$<$<COMPILE_LANGUAGE:C>:${c_flags}>"
                         "$<$<COMPILE_LANGUAGE:CXX>:${cxx_flags}>"
                         "$<$<COMPILE_LANGUAGE:CXX>:${host_cxx_flags}>")
@@ -677,13 +680,13 @@ if (GGML_USE_CPU_HBM)
 endif()
 
 add_library(ggml OBJECT
-            ggml.c
+            ggml.cpp
             ggml.h
-            ggml-alloc.c
+            ggml-alloc.cpp
             ggml-alloc.h
-            ggml-backend.c
+            ggml-backend.cpp
             ggml-backend.h
-            ggml-quants.c
+            ggml-quants.cpp
             ggml-quants.h
             ${GGML_SOURCES_CUDA} ${GGML_HEADERS_CUDA}
             ${GGML_SOURCES_OPENCL} ${GGML_HEADERS_OPENCL}