1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
|
# Makefile based on default Argeo SDK conventions,
# used as a high-level driver for build, packaging, etc.
# Low-level build only uses CMake.
-include sdk.mk
include sdk/argeo-build/cmake/default.mk
include sdk/argeo-build/jpms.mk
-include native/tp/versions.mk
A2_CATEGORY=org.argeo.jjml
TP_GGML_OLDEST=v0.9.4
TP_LLAMA_OLDEST=b6641
TP_GGML_LATEST=ac0c8be49c7458bcc6eae164244d7335ce9cc184
TP_LLAMA_LATEST=b7446
##
# Run make clean / all / install for the default CMake build.
# If system ggml and/or llama.cpp libraries are found,
# they will be used to build the JNI bindings,
# otherwise the missing layer will be buit from the source submodule.
# Use make rebuild-force-to (see below) in order to force a local build.
# rebuild-force-to: To be used for "heavy" C++ development,
# that is when adding new capabilities and exploring upstream code. It allows:
# - to ensure the target binaries are built from the local sources submodules
# - to build the tools and examples, so that they can be browsed, debugged,
# and hacked in an IDE.
TARGET_NATIVE_OUTPUT_GGML=$(TARGET_NATIVE_OUTPUT)/org.argeo.tp.ggml
TARGET_NATIVE_OUTPUT_JJML=$(TARGET_NATIVE_OUTPUT)/$(A2_CATEGORY)
# Activate various features via environment variables:
GGML_BLAS ?= OFF
GGML_VULKAN ?= OFF
GGML_CUDA ?= OFF
GGML_RPC ?= OFF
GGML_OPENMP ?= OFF
GGML_CCACHE ?= ON
LLAMA_BUILD_TOOLS ?= ON
JJML_FORCE_BUILD_LLAMA_GGML ?= OFF
ifneq (,$(VCIDEInstallDir))
MSVC_IDE_BASE=$(shell cygpath -m '$(VCIDEInstallDir)\\..')
else
MSVC_IDE_BASE=$(shell cygpath -m 'C:/Program Files (x86)/Microsoft Visual Studio/2022/BuildTools/Common7/IDE/')
endif
#MSVC_CMAKE_BASE="$(MSVC_BUILD_TOOLS)/Common7/IDE/CommonExtensions/Microsoft/CMake"
MSVC_CMAKE_BASE=$(MSVC_IDE_BASE)CommonExtensions/Microsoft/CMake
MSVC_CMAKE="$(MSVC_CMAKE_BASE)/CMake/bin/cmake.exe"
ifeq ($(MSYS_VERSION),0)
JJML_CMAKE ?= $(CMAKE)
LLAMA_CURL ?= ON
else
JJML_CMAKE ?= $(MSVC_CMAKE)
LLAMA_CURL ?= OFF
# To build with MinGW compiler
# JJML_CMAKE=cmake LLAMA_CURL=ON make -f tooling.mk clean-local rebuild-force-tp
# (and add runtime to path)
endif
rebuild-force-tp:
echo CMAKE_BUILD_TYPE=$(CMAKE_BUILD_TYPE)
$(JJML_CMAKE) -B $(BUILD_BASE) . \
-DJJML_FORCE_BUILD_TP=ON \
-DJJML_FORCE_BUILD_LLAMA_GGML=$(JJML_FORCE_BUILD_LLAMA_GGML) \
-DA2_INSTALL_MODE=a2 \
-DJAVA_HOME=$(JAVA_HOME) \
\
-DCMAKE_BUILD_TYPE=$(CMAKE_BUILD_TYPE) \
-DCMAKE_SKIP_BUILD_RPATH=ON \
-DGGML_CCACHE=$(GGML_CCACHE) \
\
-DLLAMA_CURL=$(LLAMA_CURL) \
-DLLAMA_BUILD_COMMON=$(LLAMA_BUILD_TOOLS) \
-DLLAMA_BUILD_TOOLS=$(LLAMA_BUILD_TOOLS) \
-DLLAMA_BUILD_EXAMPLES=OFF \
-DLLAMA_BUILD_TESTS=OFF \
\
-DGGML_NATIVE=OFF \
-DGGML_CPU_ALL_VARIANTS=ON \
-DGGML_BACKEND_DL=ON \
\
-DGGML_OPENMP=$(GGML_OPENMP) \
-DGGML_BLAS=$(GGML_BLAS) \
-DGGML_BLAS_VENDOR=OpenBLAS \
-DGGML_VULKAN=$(GGML_VULKAN) \
-DGGML_CUDA=$(GGML_CUDA) \
-DGGML_CUDA_FORCE_MMQ=ON \
-DGGML_CUDA_FA_ALL_QUANTS=OFF \
-DGGML_RPC=$(GGML_RPC) \
$(JJML_CMAKE) --build $(BUILD_BASE) --config $(CMAKE_BUILD_TYPE) -j $(shell nproc)
ln -f -r -s $(TARGET_NATIVE_OUTPUT_GGML)/$(shlib_prefix)*$(shlib_suffix) $(TARGET_NATIVE_OUTPUT)
ln -f -r -s $(TARGET_NATIVE_OUTPUT_JJML)/$(shlib_prefix)*$(shlib_suffix) $(TARGET_NATIVE_OUTPUT)
@$(RM) $(TARGET_NATIVE_OUTPUT_GGML)/vulkan-shaders-gen*
# Remove locally built libraries
clean-local:
$(RM) -r $(A2_OUTPUT)/org.argeo.jjml
$(RM) -r $(BUILD_BASE)
$(RM) -r $(TARGET_NATIVE_OUTPUT_JJML)
$(RM) -r $(TARGET_NATIVE_OUTPUT_GGML)
@$(RM) -v $(TARGET_NATIVE_OUTPUT)/$(shlib_prefix)ggml*$(shlib_suffix)
@$(RM) -v $(TARGET_NATIVE_OUTPUT)/$(shlib_prefix)llama*$(shlib_suffix)
@$(RM) -v $(TARGET_NATIVE_OUTPUT)/$(shlib_prefix)Java_org_argeo_jjml_*$(shlib_suffix)
@$(RM) -v $(TARGET_NATIVE_OUTPUT)/$(shlib_prefix)Java_org_argeo_jjml_*$(shlib_suffix).*
#
# DOC
#
doc-api:
$(JAVA_HOME)/bin/javadoc -Xdoclint:none \
-d doc/reference/api \
-sourcepath org.argeo.jjml/src \
-subpackages org
#
# BUILD ENVIRONMENT
#
install-deps:
ifeq ($(MSYS_VERSION),0)
else
pacman -S --needed git make mingw-w64-ucrt-x86_64-toolchain mingw-w64-ucrt-x86_64-cmake
pacman -S --needed mingw-w64-ucrt-x86_64-ccache
# Vulkan
pacman -S --needed mingw-w64-ucrt-x86_64-vulkan-devel mingw-w64-ucrt-x86_64-shaderc
endif
tp-clone:
if [ ! -d "native/tp/ggml" ]; then git clone --single-branch --branch master https://github.com/ggml-org/ggml.git native/tp/ggml; fi;
if [ ! -d "native/tp/llama.cpp" ]; then git clone --single-branch --branch master https://github.com/ggml-org/llama.cpp.git native/tp/llama.cpp; fi;
tp-checkout-oldest:
git -C native/tp/ggml fetch origin
git -C native/tp/ggml checkout $(TP_GGML_OLDEST)
git -C native/tp/llama.cpp fetch origin
git -C native/tp/llama.cpp checkout $(TP_LLAMA_OLDEST)
tp-checkout-latest:
git -C native/tp/ggml fetch origin
git -C native/tp/ggml checkout $(TP_GGML_LATEST)
git -C native/tp/llama.cpp fetch origin
git -C native/tp/llama.cpp checkout $(TP_LLAMA_LATEST)
#
# PACKAGING
#
JMOD_JJML=org.argeo.jjml
JMOD_JJML_JNI=org.argeo.jjml.jni
JMOD_GGML=org.argeo.tp.ggml.libs
JMOD_GGML_LLM=org.argeo.tp.ggml.llm.libs
jmod-jjml: a2-prepare-output
$(call a2_jmod_prepare_output,$(JMOD_JJML))
$(COPY) COPYING.LESSER NOTICE $(JMODS_BASE)/$(JMOD_JJML)/legal
# examples
mkdir -p $(JMODS_BASE)/$(JMOD_JJML)/man/examples
$(COPY) -v sdk/jbin/*.java $(JMODS_BASE)/$(JMOD_JJML)/man/examples
$(call a2_jmod_create_lib,$(JMOD_JJML))
# list content
#$(JLINK_HOME)/bin/jmod list $(JLINK_A2_JMODS)/$(JMOD_JJML).jmod
jmod-jjml-jni: a2-prepare-output
$(call a2_jmod_prepare_output,$(JMOD_JJML_JNI))
$(COPY) COPYING.LESSER NOTICE $(JMODS_BASE)/$(JMOD_JJML_JNI)/legal
$(COPY) $(TARGET_NATIVE_OUTPUT_JJML)/$(shlib_prefix)Java_org_argeo_jjml*$(shlib_suffix) \
$(JMODS_BASE)/$(JMOD_JJML_JNI)/lib
$(call a2_jmod_bare_module,$(JMOD_JJML_JNI))
# TODO use distinct version for JNI?
$(call a2_jmod_create_native,$(JMOD_JJML_JNI),$(A2_LAYER_VERSION))
jmod-ggml-libs: a2-prepare-output
$(call a2_jmod_prepare_output,$(JMOD_GGML))
$(COPY) native/tp/ggml/include/*.h $(JMODS_BASE)/$(JMOD_GGML)/include
$(COPY) native/tp/ggml/LICENSE native/tp/ggml/AUTHORS $(JMODS_BASE)/$(JMOD_GGML)/legal
ifeq ($(TARGET_OS),macos)
$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/$(shlib_prefix)ggml.0$(shlib_suffix) $(JMODS_BASE)/$(JMOD_GGML)/lib
$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/$(shlib_prefix)ggml-base.0$(shlib_suffix) $(JMODS_BASE)/$(JMOD_GGML)/lib
# When GGML_BACKEND_DL=ON, *.so are generated,
-$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/libggml-cpu*.so $(JMODS_BASE)/$(JMOD_GGML)/lib
-$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/libggml-metal.so $(JMODS_BASE)/$(JMOD_GGML)/lib
-$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/libggml-blas.so $(JMODS_BASE)/$(JMOD_GGML)/lib
# otherwise *.dylib
-$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/libggml-cpu*.dylib $(JMODS_BASE)/$(JMOD_GGML)/lib
-$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/libggml-metal.dylib $(JMODS_BASE)/$(JMOD_GGML)/lib
-$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/libggml-blas.dylib $(JMODS_BASE)/$(JMOD_GGML)/lib
else
$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/$(shlib_prefix)ggml$(shlib_suffix) $(JMODS_BASE)/$(JMOD_GGML)/lib
$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/$(shlib_prefix)ggml-base$(shlib_suffix) $(JMODS_BASE)/$(JMOD_GGML)/lib
-$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/$(shlib_prefix)ggml-cpu*$(shlib_suffix) $(JMODS_BASE)/$(JMOD_GGML)/lib
# MSVC linker libs
-$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/$(shlib_prefix)ggml.lib $(JMODS_BASE)/$(JMOD_GGML)/lib
-$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/$(shlib_prefix)ggml-base.lib $(JMODS_BASE)/$(JMOD_GGML)/lib
endif
$(call a2_jmod_bare_module,$(JMOD_GGML))
$(call a2_jmod_create_native,$(JMOD_GGML),$(GGML_VERSION))
jmod-ggml-llm-libs: a2-prepare-output
$(call a2_jmod_prepare_output,$(JMOD_GGML_LLM))
$(COPY) native/tp/llama.cpp/include/*.h $(JMODS_BASE)/$(JMOD_GGML_LLM)/include
$(COPY) native/tp/llama.cpp/LICENSE native/tp/llama.cpp/AUTHORS $(JMODS_BASE)/$(JMOD_GGML_LLM)/legal
ifeq ($(TARGET_OS),macos)
$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/$(shlib_prefix)llama.0$(shlib_suffix) $(JMODS_BASE)/$(JMOD_GGML_LLM)/lib
else
$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/$(shlib_prefix)llama$(shlib_suffix) $(JMODS_BASE)/$(JMOD_GGML_LLM)/lib
endif
ifeq ($(TARGET_OS),windows)
# MSVC linker libs
-$(COPY) $(TARGET_NATIVE_OUTPUT_GGML)/$(shlib_prefix)llama.lib $(JMODS_BASE)/$(JMOD_GGML_LLM)/lib
endif
$(call a2_jmod_bare_module,$(JMOD_GGML_LLM))
$(call a2_jmod_create_native,$(JMOD_GGML_LLM),$(LLAMA_VERSION))
#
# DISTRIBUTABLE PACKAGES
#
package-jmods: jmod-jjml jmod-jjml-jni jmod-ggml-libs jmod-ggml-llm-libs
jdk-jjml: package-jmods
$(call a2_jlink_create_jdk,jdk-jjml,$(JMOD_JJML) $(JMOD_JJML_JNI) $(JMOD_GGML) $(JMOD_GGML_LLM))
$(call a2_jlink_copy_categories,jdk-jjml,$(A2_CATEGORY))
JDK_JJML_WIN_UPGRADE_ID=d87918b9-88e7-51fb-92d5-7186ca73314b
# FIXME make it portable on non-MSYS Windows
#JDK_JJML_WIN_UPGRADE_ID=$(shell uuidgen --sha1 --namespace $(ARGEO_ENTERPRISE_NUMBER_UUID) --name jdk-jjml-$(JLINK_JAVA_RELEASE)-$(JLINK_JVM_VARIANT))
msi-jdk-jjml:
$(call a2_jpackage_create_msi,jdk-jjml,JDK $(JLINK_JAVA_RELEASE) with additional machine learning features,Argeo GmbH,$(JDK_JJML_WIN_UPGRADE_ID))
install-msi-jdk-jjml:
msiexec /i "$(BUILD_BASE)/jdk-jjml-$(JLINK_SUFFIX)-$(A2_LAYER_VERSION).msi" /passive /qn
pkg-jdk-jjml:
$(call a2_jpackage_create_pkg,jdk-jjml,JDK $(JLINK_JAVA_RELEASE) with additional machine learning features,Argeo GmbH)
install-pkg-jdk-jjml:
$(call a2_jpackage_install_pkg,jdk-jjml)
# Note: On Windows, use dumpbin.exe in order to find depedencies of a DLL
# (similar to ldd on Linux). E.g. "C:\Program Files (x86)\Microsoft Visual
# Studio\2022\BuildTools\VC\Tools\MSVC\14.44.35207\bin\Hostx64\x64\
# dumpbin.exe" /DEPENDENTS llama.dll
|
