[llama_cpp] bump version and install library and header files (#6450)

* [llama_cpp] bump version, also install library and header files

* fix umask for installed libs, bump version, add two more programs

* fix linkage of example programs on windows, bump version

Author: marcom

L/llama_cpp/build_tarballs.jl

@@ -1,7 +1,7 @@
 using BinaryBuilder, Pkg
 
 name = "llama_cpp"
-version = v"0.0.3"  # fake version number
+version = v"0.0.4"  # fake version number
 
 # url = "https://github.com/ggerganov/llama.cpp"
 # description = "Port of Facebook's LLaMA model in C/C++"
@@ -16,19 +16,22 @@ version = v"0.0.3"  # fake version number
 # 0.0.1           20.03.2023       master-074bea2    https://github.com/ggerganov/llama.cpp/releases/tag/master-074bea2
 # 0.0.2           21.03.2023       master-8cf9f34    https://github.com/ggerganov/llama.cpp/releases/tag/master-8cf9f34
 # 0.0.3           22.03.2023       master-d5850c5    https://github.com/ggerganov/llama.cpp/releases/tag/master-d5850c5
+# 0.0.4           25.03.2023       master-1972616    https://github.com/ggerganov/llama.cpp/releases/tag/master-1972616
 
 sources = [
-    # 2023.03.22, https://github.com/ggerganov/llama.cpp/releases/tag/master-d5850c5
-    # fake version = 0.0.3
+    # fake version = 0.0.4
     GitSource("https://github.com/ggerganov/llama.cpp.git",
-              "d5850c53ca179b9674b98f35d359763416a3cc11"),
+              "19726169b379bebc96189673a19b89ab1d307659"),
     DirectorySource("./bundled"),
 ]
 
 script = raw"""
 cd $WORKSPACE/srcdir/llama.cpp*
 
 atomic_patch -p1 ../patches/cmake-remove-mcpu-native.patch
+if [[ "${target}" == *-w64-mingw32* ]]; then
+    atomic_patch -p1 ../patches/windows-examples-fix-missing-ggml-link.patch
+fi
 
 EXTRA_CMAKE_ARGS=
 if [[ "${target}" == *-linux-* ]]; then
@@ -40,25 +43,49 @@ cmake .. \
     -DCMAKE_INSTALL_PREFIX=$prefix \
     -DCMAKE_TOOLCHAIN_FILE=${CMAKE_TARGET_TOOLCHAIN} \
     -DCMAKE_BUILD_TYPE=RELEASE \
+    -DBUILD_SHARED_LIBS=ON \
+    -DLLAMA_BUILD_TESTS=OFF \
+    -DLLAMA_BUILD_EXAMPLES=ON \
+    -DLLAMA_OPENBLAS=OFF \
     -DLLAMA_NATIVE=OFF \
     $EXTRA_CMAKE_ARGS
 make -j${nproc}
 
 # `make install` doesn't work (2023.03.21)
 # make install
-for prg in main quantize; do
+
+# executables
+for prg in embedding main perplexity quantize; do
     install -Dvm 755 "./bin/${prg}${exeext}" "${bindir}/${prg}${exeext}"
 done
 
+# libs
+for lib in libllama; do
+    if [[ "${target}" == *-w64-mingw32* ]]; then
+        install -Dvm 755 "./bin/${lib}.${dlext}" "${libdir}/${lib}.${dlext}"
+    else
+        install -Dvm 755 "./${lib}.${dlext}" "${libdir}/${lib}.${dlext}"
+    fi
+done
+
+
+# header files
+for hdr in llama.h ggml.h; do
+    install -Dvm 644 "../${hdr}" "${includedir}/${hdr}"
+done
+
 install_license ../LICENSE
 """
 
 platforms = supported_platforms(; exclude = p -> arch(p) ∉ ["i686", "x86_64", "aarch64"])
 platforms = expand_cxxstring_abis(platforms)
 
 products = [
+    ExecutableProduct("embedding", :embedding),
     ExecutableProduct("main", :main),
+    ExecutableProduct("perplexity", :perplexity),
     ExecutableProduct("quantize", :quantize),
+    LibraryProduct("libllama", :libllama),
 ]
 
 dependencies = Dependency[

L/llama_cpp/bundled/patches/windows-examples-fix-missing-ggml-link.patch

@@ -0,0 +1,40 @@
+diff --git a/examples/embedding/CMakeLists.txt b/examples/embedding/CMakeLists.txt
+index 88c425d..def5b83 100644
+--- a/examples/embedding/CMakeLists.txt
++++ b/examples/embedding/CMakeLists.txt
+@@ -1,4 +1,4 @@
+ set(TARGET embedding)
+ add_executable(${TARGET} embedding.cpp)
+-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
++target_link_libraries(${TARGET} PRIVATE common llama ggml ${CMAKE_THREAD_LIBS_INIT})
+ target_compile_features(${TARGET} PRIVATE cxx_std_11)
+diff --git a/examples/main/CMakeLists.txt b/examples/main/CMakeLists.txt
+index b2dcc29..aa1f794 100644
+--- a/examples/main/CMakeLists.txt
++++ b/examples/main/CMakeLists.txt
+@@ -1,4 +1,4 @@
+ set(TARGET main)
+ add_executable(${TARGET} main.cpp)
+-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
++target_link_libraries(${TARGET} PRIVATE common llama ggml ${CMAKE_THREAD_LIBS_INIT})
+ target_compile_features(${TARGET} PRIVATE cxx_std_11)
+diff --git a/examples/perplexity/CMakeLists.txt b/examples/perplexity/CMakeLists.txt
+index 5836df8..9bd8e37 100644
+--- a/examples/perplexity/CMakeLists.txt
++++ b/examples/perplexity/CMakeLists.txt
+@@ -1,4 +1,4 @@
+ set(TARGET perplexity)
+ add_executable(${TARGET} perplexity.cpp)
+-target_link_libraries(${TARGET} PRIVATE common llama ${CMAKE_THREAD_LIBS_INIT})
++target_link_libraries(${TARGET} PRIVATE common llama ggml ${CMAKE_THREAD_LIBS_INIT})
+ target_compile_features(${TARGET} PRIVATE cxx_std_11)
+diff --git a/examples/quantize/CMakeLists.txt b/examples/quantize/CMakeLists.txt
+index fb27d45..17a995b 100644
+--- a/examples/quantize/CMakeLists.txt
++++ b/examples/quantize/CMakeLists.txt
+@@ -1,4 +1,4 @@
+ set(TARGET quantize)
+ add_executable(${TARGET} quantize.cpp)
+-target_link_libraries(${TARGET} PRIVATE llama ${CMAKE_THREAD_LIBS_INIT})
++target_link_libraries(${TARGET} PRIVATE llama ggml ${CMAKE_THREAD_LIBS_INIT})
+ target_compile_features(${TARGET} PRIVATE cxx_std_11)