如何在Mac下通过NDK来编译Ceres Android（1.12.0）库（暂时完结）

最新推荐文章于 2021-05-07 14:46:43 发布

Summer、苏

最新推荐文章于 2021-05-07 14:46:43 发布

阅读量1.2k

点赞数 1

分类专栏： Ceres相关

本文链接：https://blog.csdn.net/haejwcalcv/article/details/99334905

版权

Ceres相关专栏收录该内容

3 篇文章 0 订阅

订阅专栏

如何在Mac下通过NDK来编译Ceres Android库（1.12.0版本）暂时完结

前言

这里先报个预告，1.12.0版本已经很老了，但由于当时在这个版本上踩了一些坑，编出可用的Android库比较不容易，所以还是写出来纪念一下。
下篇文章我会讲述如何编译最新的ceres2.0.0版本（也就是现在ceres master分支）的Android库
有兴趣可以先关注相应的github网址：https://github.com/PotterSu/ceres-android

前言

网上关于Ceres的安装教程很多，但是Ceres 安卓库真的是很少，Ceres官网对于Android版也更是一句话略过了。即使一些博客有涉及android库如何编译，但编译出的库也大多都是阉割版，很多功能(例如Suite Sparse)无法用。这就意味着你无法求解很复杂的计算，或者即使能够求解也将非常耗时。
所以本文会逐步讲述如何编译一个功能齐全并且可以在安卓手机上调用的Ceres安卓库。
如有疑问欢迎留言，尽我所知回复大家。

准备工作

在此特别鸣谢一些大佬们的之前工作
1.https://blog.csdn.net/u010203716/article/details/72955880
这篇博客提供了一个编译好的Ceres库，版本是1.12.0的，亲测可用，不过不支持Suite Sparse 和多线程求解。我们首先下载好博主提供的ceres库到本机上。
2.https://www.jianshu.com/p/2e7399a18673
3.https://github.com/kiutao/suitesparse_android_lib
求解大型矩阵的话需要用到SuiteSparse库，然而其官网的源码虽然支持很多平台，但是却没有提供Android平台的…一大堆makefile，如果自己整理会很麻烦。
好在这位github作者已经整理好了一份安卓版的SuiteSparse库，而且这个github中附带了suitesparse依赖的lapack、blas等第三方库，不用我们自己再去关联！我们只需要根据他的脚本直接编译就好，很良心！
我们把这个库down下来。
4.除此之外我们通过homebrew安装mac版的SuiteSparse。
输入brew install suite-sparse即可
(如果不知道如何安装也可以看下我上一篇博客)。
这样SuiteSparse的头文件都会存放在/usr/local/Cellar/suite-sparse/5.3.0_1/include下，后续Android.mk关联SuiteSparse的头文件会非常方便。
5.在Eigen官网上下载Eigen库的源码包并解压。
6.安装android-ndk，网上教程比较多这里不再赘述了。

如何编译带SuiteSparse的Ceres库

1.首先进入到刚刚在博客1下载好的ceres-solver-1.12.0/jni目录下，将里面的Android.mk文件替换成下面内容。
先说下对Android.mk的改动，
首先我们需要开启编译SuiteSparse的选项，正常是关闭的，所以按照博客1提供ceres库没有办法调用trust region等求解方法。我们需要将DCERES_NO_SUITESPARSE替换成DCERES_USE_SUITESPARSE标志。
开启SUITESPARSE之后，编译时会寻找SuiteSparse需要的头文件，所以需要添加suitesparse头文件路径。如果不关联会提示找不到所需的头文件。
设置SUITESPARSE_PATH := /usr/local/Cellar/suite-sparse/5.3.0_1/include
这里的include路径就是我们准备工作中通过homebrew实现安装好的mac版的suite-sparse,我们只是用到了头文件而已。
然后把SUITESPARSE_PATH添加到
LOCAL_C_INCLUDES += $(SUITESPARSE_PATH)中。
完整Android.mk如下。

# Ceres Solver - A fast non-linear least squares minimizer
# Copyright 2015 Google Inc. All rights reserved.
# http://ceres-solver.org/
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
#
# * Redistributions of source code must retain the above copyright notice,
#   this list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright notice,
#   this list of conditions and the following disclaimer in the documentation
#   and/or other materials provided with the distribution.
# * Neither the name of Google Inc. nor the names of its contributors may be
#   used to endorse or promote products derived from this software without
#   specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
# ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
# POSSIBILITY OF SUCH DAMAGE.
#
# Author: settinger@google.com (Scott Ettinger)
#         keir@google.com (Keir Mierle)
#
# Builds Ceres for Android, using the standard toolchain (not
# standalone). It uses LLVM's libc++ as the standard library. It is a
# modern BSD licensed implementation of the standard c++ library. We
# do this to avoid any licensing issues that may arise from using
# GCC's libstdc++ which is licensed under GPL3.
#
# Building
# --------
#
# You will have to specify the environment EIGEN_PATH to point to the
# Eigen sources when building. For example:
#
#   EIGEN_PATH=/home/keir/src/eigen-3.0.5 ndk-build -j
#
# It is also possible to specify CERES_EXTRA_DEFINES, in case you need
# to pass more definitions to the C compiler.
#
# Using the library
# -----------------
# Copy the static library:
#
#   ../obj/local/armeabi-v7a/libceres.a
#
# into your own project, then link it into your binary in your
# Android.mk file.
#
# Reducing binary size
# --------------------
# This build includes the Schur specializations, which increase the
# size of the binary. If you don't need them for your application,
# consider adding:
#
#   -DCERES_RESTRICT_SCHUR_SPECIALIZATION
#
# to the LOCAL_CFLAGS variable below.
#
# Changing the logging library
# ----------------------------
# Ceres Solver ships with a replacement for glog that provides a
# simple and small implementation that builds on Android. However, if
# you wish to supply a header only version yourself, then you may
# define CERES_GLOG_DIR to point to it.

LOCAL_PATH := $(call my-dir)


# MY_DD := $(shell $(LOCAL_PATH)/assert_ndk_version.sh)
#
# MY_VAR := $(shell $(LOCAL_PATH)/assert_ndk_version.sh "r9d" $(NDK_ROOT))
#
# # Ceres requires at least NDK version r9d to compile.
# ifneq ($(MY_VAR), true)
#   ##$(error Ceres requires NDK version r9d or greater dd=$(MY_DD))
# endif

# Ceres requires Eigen to build.
ifndef EIGEN_PATH
  $(error Ceres requires Eigen; please invoke via EIGEN_PATH=... ndk-build)
endif

EIGEN_PATH := $(EIGEN_PATH)
SUITESPARSE_PATH := /usr/local/Cellar/suite-sparse/5.3.0_1/include
CERES_INCLUDE_PATHS := $(CERES_EXTRA_INCLUDES)
CERES_INCLUDE_PATHS += $(LOCAL_PATH)/../internal
CERES_INCLUDE_PATHS += $(LOCAL_PATH)/../internal/ceres
CERES_INCLUDE_PATHS += $(LOCAL_PATH)/../include
CERES_INCLUDE_PATHS += $(LOCAL_PATH)/../config

# Use the alternate glog implementation if provided by the user.
ifdef CERES_GLOG_DIR
  CERES_INCLUDE_PATHS += $(CERES_GLOG_DIR)
else
  CERES_INCLUDE_PATHS += $(LOCAL_PATH)/../internal/ceres/miniglog
endif
CERES_SRC_PATH := ../internal/ceres

include $(CLEAR_VARS)
LOCAL_CFLAGS := -DCERES_SUITESPARSE_VERSION=\"5.3.0\"

LOCAL_C_INCLUDES := $(CERES_INCLUDE_PATHS)
LOCAL_C_INCLUDES += $(EIGEN_PATH)
LOCAL_C_INCLUDES += $(SUITESPARSE_PATH)

LOCAL_CPP_EXTENSION := .cc
LOCAL_CFLAGS += $(CERES_EXTRA_DEFINES) \
                -DCERES_NO_LAPACK \
                -DCERES_USE_SUITESPARSE \
                -DCERES_NO_CXSPARSE \
                -DCERES_STD_UNORDERED_MAP


# If the user did not enable threads in CERES_EXTRA_DEFINES, then add
# CERES_NO_THREADS.
#
# TODO(sameeragarwal): Update comments here and in the docs to
# demonstrate how OpenMP can be used by the user.
ifeq (,$(findstring CERES_HAVE_PTHREAD, $(LOCAL_CFLAGS)))
  LOCAL_CFLAGS += -DCERES_NO_THREADS
endif

# LOCAL_STATIC_LIBRARIES := static_libamd static_libcamd static_libbtf static_libcholmod static_libcsparse
# 						  static_libcxsparse static_libgraphblas static_libspqr static_libsuitesparseconfig static_libumfpack

LOCAL_SRC_FILES :=  $(CERES_SRC_PATH)/single_linkage_clustering.cc \
					$(CERES_SRC_PATH)/compressed_col_sparse_matrix_utils.cc \
					$(CERES_SRC_PATH)/solver.cc \
					$(CERES_SRC_PATH)/sparse_matrix.cc \
					$(CERES_SRC_PATH)/sparse_normal_cholesky_solver.cc \
					$(CERES_SRC_PATH)/split.cc \
					$(CERES_SRC_PATH)/stringprintf.cc \
					$(CERES_SRC_PATH)/suitesparse.cc \
					$(CERES_SRC_PATH)/triplet_sparse_matrix.cc \
					$(CERES_SRC_PATH)/trust_region_minimizer.cc \
					$(CERES_SRC_PATH)/trust_region_preprocessor.cc \
					$(CERES_SRC_PATH)/trust_region_step_evaluator.cc \
					$(CERES_SRC_PATH)/trust_region_strategy.cc \
					$(CERES_SRC_PATH)/types.cc \
					$(CERES_SRC_PATH)/visibility_based_preconditioner.cc \
					$(CERES_SRC_PATH)/visibility.cc \
					$(CERES_SRC_PATH)/wall_time.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_d_d_d.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_2_2.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_2_3.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_2_4.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_2_d.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_3_3.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_3_4.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_3_6.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_3_9.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_3_d.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_4_3.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_4_4.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_4_8.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_4_9.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_4_d.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_2_d_d.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_4_4_2.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_4_4_3.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_4_4_4.cc \
					$(CERES_SRC_PATH)/generated/schur_eliminator_4_4_d.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_d_d_d.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_2_2.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_2_3.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_2_4.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_2_d.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_3_3.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_3_4.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_3_6.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_3_9.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_3_d.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_4_3.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_4_4.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_4_8.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_4_9.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_4_d.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_2_d_d.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_4_4_2.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_4_4_3.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_4_4_4.cc \
					$(CERES_SRC_PATH)/generated/partitioned_matrix_view_4_4_d.cc \
					$(CERES_SRC_PATH)/solver_utils.cc \
				   $(CERES_SRC_PATH)/array_utils.cc \
                   $(CERES_SRC_PATH)/blas.cc \
                   $(CERES_SRC_PATH)/block_evaluate_preparer.cc \
                   $(CERES_SRC_PATH)/block_jacobian_writer.cc \
                   $(CERES_SRC_PATH)/block_jacobi_preconditioner.cc \
                   $(CERES_SRC_PATH)/block_random_access_dense_matrix.cc \
                   $(CERES_SRC_PATH)/block_random_access_diagonal_matrix.cc \
                   $(CERES_SRC_PATH)/block_random_access_matrix.cc \
                   $(CERES_SRC_PATH)/block_random_access_sparse_matrix.cc \
                   $(CERES_SRC_PATH)/block_sparse_matrix.cc \
                   $(CERES_SRC_PATH)/block_structure.cc \
                   $(CERES_SRC_PATH)/callbacks.cc \
                   $(CERES_SRC_PATH)/canonical_views_clustering.cc \
                   $(CERES_SRC_PATH)/cgnr_solver.cc \
                   $(CERES_SRC_PATH)/compressed_row_jacobian_writer.cc \
                   $(CERES_SRC_PATH)/compressed_row_sparse_matrix.cc \
                   $(CERES_SRC_PATH)/conditioned_cost_function.cc \
                   $(CERES_SRC_PATH)/conjugate_gradients_solver.cc \
                   $(CERES_SRC_PATH)/coordinate_descent_minimizer.cc \
                   $(CERES_SRC_PATH)/corrector.cc \
                   $(CERES_SRC_PATH)/covariance.cc \
                   $(CERES_SRC_PATH)/covariance_impl.cc \
                   $(CERES_SRC_PATH)/dense_normal_cholesky_solver.cc \
                   $(CERES_SRC_PATH)/dense_qr_solver.cc \
                   $(CERES_SRC_PATH)/dense_sparse_matrix.cc \
                   $(CERES_SRC_PATH)/detect_structure.cc \
                   $(CERES_SRC_PATH)/dogleg_strategy.cc \
                   $(CERES_SRC_PATH)/dynamic_compressed_row_jacobian_writer.cc \
                   $(CERES_SRC_PATH)/dynamic_compressed_row_sparse_matrix.cc \
                   $(CERES_SRC_PATH)/evaluator.cc \
                   $(CERES_SRC_PATH)/file.cc \
                   $(CERES_SRC_PATH)/gradient_checker.cc \
                   $(CERES_SRC_PATH)/gradient_checking_cost_function.cc \
                   $(CERES_SRC_PATH)/gradient_problem.cc \
                   $(CERES_SRC_PATH)/gradient_problem_solver.cc \
                   $(CERES_SRC_PATH)/is_close.cc \
                   $(CERES_SRC_PATH)/implicit_schur_complement.cc \
                   $(CERES_SRC_PATH)/iterative_schur_complement_solver.cc \
                   $(CERES_SRC_PATH)/lapack.cc \
                   $(CERES_SRC_PATH)/levenberg_marquardt_strategy.cc \
                   $(CERES_SRC_PATH)/line_search.cc \
                   $(CERES_SRC_PATH)/line_search_direction.cc \
                   $(CERES_SRC_PATH)/line_search_minimizer.cc \
                   $(CERES_SRC_PATH)/linear_least_squares_problems.cc \
                   $(CERES_SRC_PATH)/linear_operator.cc \
                   $(CERES_SRC_PATH)/line_search_preprocessor.cc \
                   $(CERES_SRC_PATH)/linear_solver.cc \
                   $(CERES_SRC_PATH)/local_parameterization.cc \
                   $(CERES_SRC_PATH)/loss_function.cc \
                   $(CERES_SRC_PATH)/low_rank_inverse_hessian.cc \
                   $(CERES_SRC_PATH)/minimizer.cc \
                   $(CERES_SRC_PATH)/normal_prior.cc \
                   $(CERES_SRC_PATH)/parameter_block_ordering.cc \
                   $(CERES_SRC_PATH)/partitioned_matrix_view.cc \
                   $(CERES_SRC_PATH)/polynomial.cc \
                   $(CERES_SRC_PATH)/preconditioner.cc \
                   $(CERES_SRC_PATH)/preprocessor.cc \
                   $(CERES_SRC_PATH)/problem.cc \
                   $(CERES_SRC_PATH)/problem_impl.cc \
                   $(CERES_SRC_PATH)/program.cc \
                   $(CERES_SRC_PATH)/reorder_program.cc \
                   $(CERES_SRC_PATH)/residual_block.cc \
                   $(CERES_SRC_PATH)/residual_block_utils.cc \
                   $(CERES_SRC_PATH)/schur_complement_solver.cc \
                   $(CERES_SRC_PATH)/schur_eliminator.cc \
                   $(CERES_SRC_PATH)/schur_jacobi_preconditioner.cc \
                   $(CERES_SRC_PATH)/scratch_evaluate_preparer.cc


ifndef CERES_GLOG_DIR
LOCAL_SRC_FILES += $(CERES_SRC_PATH)/miniglog/glog/logging.cc
endif

LOCAL_MODULE := ceres
include $(BUILD_STATIC_LIBRARY)

2.将jni目录下的Application.mk替换成如下内容，这里我自定义配置了编译器和c++标准(gnustl_static和c++0x)，这是因为我的调用程序就是gnustl_static编译的，不然即使编译好了库也没法调用。
可能你需要的是(c++_static等)，要根据你的实际需求来。

# Ceres Solver - A fast non-linear least squares minimizer
# Copyright 2015 Google Inc. All rights reserved.
# http://ceres-solver.org/
#
# Redistribution and use in source and binary forms, with or without
# modification, are permitted provided that the following conditions are met:
#
# * Redistributions of source code must retain the above copyright notice,
#   this list of conditions and the following disclaimer.
# * Redistributions in binary form must reproduce the above copyright notice,
#   this list of conditions and the following disclaimer in the documentation
#   and/or other materials provided with the distribution.
# * Neither the name of Google Inc. nor the names of its contributors may be
#   used to endorse or promote products derived from this software without
#   specific prior written permission.
#
# THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
# AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
# IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
# ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
# LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
# CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
# SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
# INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
# CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
# ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
# POSSIBILITY OF SUCH DAMAGE.

APP_BUILD_SCRIPT := $(call my-dir)/Android.mk
APP_PROJECT_PATH := $(call my-dir)

APP_CPPFLAGS := -frtti -fexceptions -std=c++0x
APP_OPTIM := release

# Use libc++ from LLVM. It is a modern BSD licensed implementation of
# the standard C++ library.
APP_STL := gnustl_static
APP_ABI := armeabi-v7a

3.准备好Android.mk和Application.mk后就可以开始编译啦，我们打开终端进入到your/path/to/ceres-solver-1.12.0/jni路径下
输入我们准备工作中事先下载好的Eigen源码包路径和你的NDK路径
EIGEN_PATH=/your/path/to/eigen /your/path/to/ndk-build
不知道ndk-build位置，可以在终端输入which ndk-build，这里以我的Eigen和NDK路径为例，
EIGEN_PATH=/Users/subowen/WorkDir/SDKDir/eigen /Users/subowen/WorkDir/SDKDir/NDKDir/android-ndk-r16b/ndk-build
在这里插入图片描述
如果一切顺利的话会出现上面的编译各个文件的log。同时在/your/path/to/ceres-solver-1.12.0/obj/local/armeabi-v7a/目录下会生成编译成功的libceres.a静态库。

添加编译suitesparse的安卓so库

上述只是成功编译了libceres.a的静态库，但是由于我们用到了suitesparse，所以必然也需要suitesparse的库作为依赖，libceres.a才能成功运行。
那么如何编译SuiteSparse呢？
这里再次感谢https://github.com/kiutao/suitesparse_android_lib 的作者的整理。
我们从github上下载好后。按照作者的readme提示，需要对SuiteSparse库进行交叉编译。
这里有两个步骤：
1.首先交叉编译clapack生成相应的库，该库后续会被suitesparse用到。
首先进入clapack文件夹。
在这里插入图片描述
新建build文件夹并进入

这里我们新建个make_clapck.sh文件（其实原作者已经提供了这个sh脚本用来交叉编译CMakeLists.txt得到安卓so库，不过我们需要根据自己的需求进行修改）。
打开新建的make_clapck.sh文件输入如下编译选项

cmake -DCMAKE_TOOLCHAIN_FILE=/Users/subowen/WorkDir/SDKDir/NDKDir/android-ndk-r16b/build/cmake/android.toolchain.cmake 
-DANDROID_NDK=/Users/subowen/WorkDir/SDKDir/NDKDir/android-ndk-r16b/ 
-DANDROID_TOOLCHAIN="clang" 
-DANDROID_STL=gnustl_static 
-DCMAKE_BUILD_TYPE=Release 
-DBUILD_ANDROID=TRUE 
-DANDROID_ABI="armeabi-v7a" 
-DANDROID_PLATFORM=android-21 
-DANDROID_ARM_NEON=TRUE  
-DEIGEN3_INCLUDE_DIR="/usr/local/Cellar/eigen/3.3.7/include/eigen3" 
-DEIGEN3_VERSION_OK=ON 
-DANDROID_DISABLE_FORMAT_STRING_CHECKS=TRUE 
-DSHARED="SHARED" .. && cmake --build .

这里根据你的实际情况替换NDK的路径DCMAKE_TOOLCHAIN_FILE和DANDROID_NDK，
以及DEIGEN3_INCLUDE_DIR路径和一些安卓平台版本等等。
之后运行脚本即可sh make_clapck.sh即可编译出SuiteSparse所需要的libblas.a、liblapack.a和libf2c.a，下一步编译suitesparse.so我们会用到这三个.a静态库。

2.接下来我们需要开始编译suitesparse.so了！
这里我们首先需要修改suitespase文件夹下的CMakeLists.txt文件，原始作者提供的CMakeLists没法找到需要的clapack库，所以我们将红框内的内容替换为刚刚编译好的clapack库路径。
在这里插入图片描述
接下来同样，创建在suitesparse文件夹下创建build文件夹，进入build文件夹并新建make_suitesparse.sh脚本

在make_suitesparse.sh中填写如下内容：

export CLAPACKDIR={/Users/subowen/WorkDir/SDKDir/suitesparse_android_lib-master/clapack}
cmake 
-DCMAKE_TOOLCHAIN_FILE=/Users/subowen/WorkDir/SDKDir/NDKDir/android-ndk-r16b/build/cmake/android.toolchain.cmake -DANDROID_NDK=/Users/subowen/WorkDir/SDKDir/NDKDir/android-ndk-r16b 
-DANDROID_TOOLCHAIN="clang" 
-DANDROID_STL=gnustl_static 
-DCMAKE_BUILD_TYPE=Release 
-DBUILD_ANDROID=TRUE 
-DANDROID_ABI="armeabi-v7a" 
-DANDROID_PLATFORM=android-21 
-DANDROID_ARM_NEON=TRUE  
-DEIGEN3_INCLUDE_DIR="/usr/local/Cellar/eigen/3.3.7/include/eigen3" 
-DEIGEN3_VERSION_OK=ON 
-DANDROID_DISABLE_FORMAT_STRING_CHECKS=TRUE 
-DSHARED="SHARED" .. && cmake --build .

同样上述的NDK路径Eigen路径和一些选项要根据你自己的需要灵活替换，出现问题了试试替换ANDOIRD_STL和ANDROID_TOOLCHAIN或许可以解决。
接下来运行sh make_suitesparse.sh即可开始编译。
成功后会在build路径的lib文件夹下生成一大堆so文件。
在这里插入图片描述
这些so文件都需要关联到你的工程中!要不libceres.a依然无法调用。
也就是说最终你的安卓工程下要有libceres.a以及上图所有的so。
所以也就是说你的工程有10多个库文件，同时你的Android.mk文件要配置这些动态库和静态库的路径。
这很不优雅，所以下一个博客我会介绍如何编译最新的功能齐全的ceres2.0.0版本，同时只需要一个libceres.so就可以搞定!不在需要上图一大坨的so文件！

Summer、苏

关注

1
点赞
踩
3

收藏

觉得还不错? 一键收藏
2
评论
如何在Mac下通过NDK来编译Ceres Android（1.12.0）库（暂时完结）

如何在Mac下通过NDK来编译Ceres Android库前言准备工作修改Android.mk内容前言网上关于Ceres的安装教程很多，但是Ceres 安卓库真的是很少，Ceres官网对于Android版也更是一句话略过了。即使一些博客有涉及android库如何编译，但编译出的库也大多都是阉割版，很多功能(例如Suite Sparse)无法用。这就意味着你无法求解很复杂的计算，或者即使能够求解...
复制链接

扫一扫