On 4 February 2013 02:08, Matthew Gretton-Dann matthew.gretton-dann@linaro.org wrote:

On 2 February 2013 00:13, Derek Rollend derek.rollend@gmail.com wrote:

...

Hello,

I am currently trying to determine how much I can optimize OpenCV using the ARM's VFP and the Linaro nano image. I have downloaded the arm-linux-gnueabihf-gcc & arm-linux-gnueabihf-g++ compilers, successfully cross-compiled OpenCV using those compilers (with -O3 -mfloat-abi=hard -ftree-vectorize -funroll-loops), and compiled an example OpenCV program with the arm-linux-gnueabihf-g++ compiler (with the same flags). However, when I try and run my executable within Linaro (running on a gumstix overo) I get "/lib/arm-linux-gnueabihf/libm.so.6: version `GLIBC_2.15' not found (required by /usr/lib/libopencv_core.so.2.2).

I assume that this is because on my build machine (x86 Ubuntu 10.04) where I cross compiled the OpenCV libraries I am using libc-2.15, and on the target (gumstix) the loader can't find libc-2.15. However, within the gcc-linaro-arm-linux-gnueabihf-4.7 tar ball I downloaded today, there is a libc-2.15.so file in there that I copied into /lib/ on the gumstix. Still no success though when I try running my executable.

Any recommendations on things to try? I am hesitant to just put a newer version of glibc in Linaro as it seems like that could screw a lot of things up. Should I attempt to re-build OpenCV somehow with an older version of glibc (namely libc-2.13)?

Expand the entire sysroot tarball you have onto your ARM board into some location (say /opt/gcc-linaro - but anywhere will do). Then do: LD_LIBRARY_PATH=/opt/gcc-linaro/lib/arm-linux-gnueabihf program-to-execute args

Please try to use --sysroot option. copy the entire sysroot directories (/lib, /usr/lib, /usr/include) you have onto your ARM board into some location (say /opt/gcc-linaro) on your host machine and keep the dir layout, i.e. After copy, you will have:

/opt/gcc-linaro/lib/... /opt/gcc-linaro/usr/lib/... /opt/gcc-linaro/usr/inlcude...

Then compile your project with additional gcc option: --sysroot=/opt/gcc-linaro/

-Zhenqiang

Fernan,

I'm using CMake 2.8.0. I have not tried to compile OpenCV with FFMPEG support yet, although I will be trying to do this soon. I have only successfully compiled OpenCV 2.2 using the Linaro hard float compiler. Here is my toolchain.cmake file:

set( CMAKE_SYSTEM_NAME Linux ) set( CMAKE_SYSTEM_PROCESSOR arm ) set( CMAKE_C_COMPILER arm-linux-gnueabihf-gcc ) set( CMAKE_CXX_COMPILER arm-linux-gnueabihf-g++ ) set( CMAKE_CXX_FLAGS_RELEASE "-L /opt/gcc-linaro-arm-linux-gnueabihf-4.7-2013.01-20130125_linux/arm-linux-gnueabihf/lib -O3 -mcpu=cortex-a8 -mfpu=neon -mfloat-abi=hard -ftree-vectorize -ftree-vectorizer-verbose=3 -funroll-loops" ) set( CMAKE_CXX_FLAGS_DEBUG "-L /opt/gcc-linaro-arm-linux-gnueabihf-4.7-2013.01-20130125_linux/arm-linux-gnueabihf/lib -O3 -mcpu=cortex-a8 -mfpu=neon -mfloat-abi=hard -ftree-vectorize -ftree-vectorizer-verbose=3 -funroll-loops" )

where the Linaro hard float toolchain tarball was extracted to my /opt directory. The verbose flag will tell you when the compiler was able to vectorize a loop, or why it couldn't. I need to test some of the newer feature extraction algorithms in more recent versions of OpenCV, so I'll try and update you when/if I successfully compile a newer version of OpenCV with FFMPEG support.

Good luck.

-Derek

On Mon, Feb 11, 2013 at 3:07 PM, Fernan fhledesma@gmail.com wrote:

Derek Rollend <derek.rollend@...> writes:

...

Hello, I am currently trying to determine how much I can optimize OpenCV using

the

...

ARM's VFP and the Linaro nano image. I have downloaded the arm-linux-gnueabihf-gcc & arm-linux-gnueabihf-g++ compilers, successfully cross-compiled OpenCV using those compilers (with -O3 -mfloat-abi=hard -ftree-vectorize -funroll-loops), and compiled an example OpenCV program

with

...

the arm-linux-gnueabihf-g++ compiler (with the same flags).

Hello Derek, I'm trying to cross compile OpenCV (2.4.0) for ARM. I'm using Linaro toolchain and CMake (http://processors.wiki.ti.com/index.php/Building_OpenCV_for_ARM_Cortex-A8), but I get the following error: .../ffmpeg_codecs.hpp:78:36:fatal error: libavformat/avformat.h: No such file or directory compilation terminated. I have installed libavformat-dev and configured CMake again, but I get the same error. Which version of CMake you are using? How you write de toolchain.cmake? How do you set the (-O3 -mfloat-abi=hard -ftree-vectorize -funroll-loops) C/CXX flags? I'm using CMake2.8.7. Thanks in advance.

---

linaro-toolchain mailing list linaro-toolchain@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-toolchain

Add the Linaro toolchain bin directory to your PATH environment variable:

export PATH="/opt/gcc-linaro-arm-linux-gnueabihf-4.7-2013.01-20130125_linux/bin:$PATH"

or whichever directory you have extracted the Linaro HF tarball to.

-Derek

On Mon, Feb 11, 2013 at 4:04 PM, Fernan fhledesma@gmail.com wrote:

Derek Rollend <derek.rollend@...> writes:

...

Fernan, I'm using CMake 2.8.0. I have not tried to compile OpenCV with FFMPEG

support yet, although I will be trying to do this soon. I have only successfully compiled OpenCV 2.2 using the Linaro hard float compiler. Here is my toolchain.cmake file:

...

set( CMAKE_SYSTEM_NAME Linux ) set( CMAKE_SYSTEM_PROCESSOR arm ) set( CMAKE_C_COMPILER arm-linux-gnueabihf-gcc ) set( CMAKE_CXX_COMPILER arm-linux-gnueabihf-g++ ) set( CMAKE_CXX_FLAGS_RELEASE "-L /opt/gcc-linaro-arm-linux-gnueabihf-4.7-2013.01-20130125_linux/ arm-linux-gnueabihf/lib -O3 -mcpu=cortex-a8 -mfpu=neon -mfloat-abi=hard -ftree-vectorize -ftree-vectorizer-verbose=3 -funroll-loops" ) set( CMAKE_CXX_FLAGS_DEBUG "-L /opt/gcc-linaro-arm-linux-gnueabihf- 4.7-2013.01-20130125_linux/arm-linux-gnueabihf/lib -O3 -mcpu=cortex-a8 -mfpu=neon -mfloat-abi=hard -ftree-vectorize -ftree-vectorizer-verbose=3 -funroll-loops" )

Thank you very much Derek, some questions arises reading your response, e.g. How Cmake knows the PATH of arm-linux-gnueabihf-g++?

But now I will try to cross-compile again using your toolchain.cmake.

---

linaro-toolchain mailing list linaro-toolchain@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-toolchain

I added the lib path explicitly so I knew that arm-linux-gnueabihf-g++ was definitely linking against the correct hard float libraries. Perhaps this is unnecessary, but I just wanted to be sure. I've been using OpenCV 2.2 because that was the version that was known to compile successfully from that TI wiki page you sent earlier. I am now trying with 2.4.0, I'll let you know if I succeed.

It's definitely not easy. All the tools seem to be there, it's piecing everything together correctly that is the tough part. Push on!

On Tue, Feb 12, 2013 at 1:35 PM, Fernan fhledesma@gmail.com wrote:

Hi Derek,

Derek Rollend <derek.rollend@...> writes:

...

Add the Linaro toolchain bin directory to your PATH environment variable: export

PATH="/opt/gcc-linaro-arm-linux-gnueabihf-4.7-2013.01-20130125_linux

...

...

/bin:$PATH"

ok. I thought there was a CMake variable like CMAKE_C_COMPILER_PATH. I use this http://www.cmake.org/Wiki/CMake_FAQ#How_do_I_use_a_different_compiler.3F (method3).

...

On Mon, Feb 11, 2013 at 4:04 PM, Fernan

fhledesma@gmail.com wrote:

...

Derek Rollend <derek.rollend <at> ...> writes:

...

Fernan, I'm using CMake 2.8.0. I have not tried to compile OpenCV with FFMPEG support

yet, although I will be trying to do this soon. I have only successfully compiled OpenCV 2.2 using the Linaro hard float compiler. Here is my toolchain.cmake file:

...

set( CMAKE_SYSTEM_NAME Linux ) set( CMAKE_SYSTEM_PROCESSOR arm ) set( CMAKE_C_COMPILER arm-linux-gnueabihf-gcc ) set( CMAKE_CXX_COMPILER arm-linux-gnueabihf-g++ ) set( CMAKE_CXX_FLAGS_RELEASE "-L /opt/gcc-linaro-arm-linux-gnueabihf-4.7-2013.01-20130125_linux/

...

arm-linux-gnueabihf/lib -O3 -mcpu=cortex-a8 -mfpu=neon -mfloat-abi=hard -ftree-vectorize -ftree-vectorizer-verbose=3 -funroll-loops" ) set( CMAKE_CXX_FLAGS_DEBUG "-L /opt/gcc-linaro-arm-linux-gnueabihf- 4.7-2013.01-20130125_linux/arm-linux-gnueabihf/lib -O3 -mcpu=cortex-a8 -mfpu=neon -mfloat-abi=hard -ftree-vectorize -ftree-vectorizer-verbose=3 -funroll-loops" )

I have just successfully compiled but without FFMPEG too, that is the problem.

but first, I will ask you some question:

Why you need to add the link path /opt/gcc-linaro-arm-linux-gnueabihf- 4.7-2013.01-20130125_linux/arm-linux-gnueabihf/lib?

Why you use OpenCV2.2? The lastest stable release is OpenCV2.4.3.

You see, I'm a newbie about cross compiling, I'll have to study concepts of cross-compilation, I thought it was easy.

---

linaro-toolchain mailing list linaro-toolchain@lists.linaro.org http://lists.linaro.org/mailman/listinfo/linaro-toolchain

Christophe Lyon <christophe.lyon@...> writes:

Regarding your problems with ffmpeg, I'm sorry that I don't know how cmake works, but after spending some time trying to understand why cross-building fails, here are my remarks:

• installing packages such as libavcodec-dev installs the version for

your host, not for your ARM target

Hi Christophe,

thanks for replay.

ok I will take it into account.

• therefore, you should add ffmpeg to OpenCV, maybe using the 3rdparty

directory, but I don't know how it is supposed to be used

I too don't know. Then Cross compiling is not so simple like this:

http://processors.wiki.ti.com/index.php/Building_OpenCV_for_ARM_Cortex-A8

• on my system, I have noticed that cmake says 'Could NOT find JPEG),

then "Media I/O: JPEG: libjpeg (ver 62)" and I can see that g++ is called with -I<path-to-OpenCV>/3rdparty/libjpeg, which is confusing

Same thing happens to me with FFMPEG

• similarly, it looks like cmake is calling pkg-config which most

probably queries the host, not the target FS

• you have to make sure that the build process will find or build all

the dependencies for ARM, which probably means it has to build ffmpeg

• I don't know if all these are peculiarities of cmake or OpenCV, but

it makes me wonder whether OpenCV is really supporting cross-compilation?

• did you ask OpenCV community?

I could see OpenCV community recomend other ways to do what I want.

Sorry to ask more questions

:)

Christophe.

Best regards.