Flickernoise build instructions

From Milkymist Wiki

Jump to: navigation, search
Flickernoise screenshot
Flickernoise Video-in Performance

Contents

[edit] Build products

The Automatic or Manual build procedures create the Flickernoise VJ application for the Milkymist One. The final Flickernoise executable is an ELF binary. From that, you can either:

  • try the ELF binary in the QEMU emulator with qemu-system-lm32 -M milkymist -kernel flickernoise
  • generate a raw binary file with make flickernoise.ralf, suitable for TFTP netbooting (recommended), booting from the memory card or downloading over the serial line with flterm. Using flterm is not recommended as this method is slow because of the large size of the binary.
  • generate a FBI image with make flickernoise.fbi that you can use with the auto-upgrade feature of Flickernoise (0.2+).

[edit] Automatic build

xiangfu wrote some Makefiles to compile the lm32 toolchain and flickernoise. Please read the README.html for detailed directions and host specific requirements. Build steps are below:

  • Download the script files: 
   $ git clone git://github.com/milkymist/scripts.git
  • Read and follow the directions in the README.html file: 
   $ cd scripts  
   $ vi README

[edit] Manual build steps

Note: The automatic build procedure is the preferred way of building the Flickernoise application.

[edit] RTEMS and toolchain

First you need to follow the RTEMS build instructions. We will assume you have installed both RTEMS and the toolchain in /opt/rtems (or /opt/rtems-VERSION). Not installing RTEMS and the toolchain in the same directory tends to cause problems w.r.t. finding include files.

We recommend using the GCC 4.5.1 based toolchain, which works reasonably well. You will however need to patch GCC a bit so the base C libraries can use all the features of our LM32 implementation: 

diff -Naur gcc-4.5.1.orig/gcc/config/lm32/t-lm32 gcc-4.5.1/gcc/config/lm32/t-lm32
--- gcc-4.5.1.orig/gcc/config/lm32/t-lm32	2010-03-30 12:26:24.000000000 +0200
+++ gcc-4.5.1/gcc/config/lm32/t-lm32	2010-11-28 12:42:30.047795004 +0100
@@ -1,2 +1,2 @@
 # for multilib
-MULTILIB_OPTIONS = mmultiply-enabled mbarrel-shift-enabled
+MULTILIB_OPTIONS = mbarrel-shift-enabled mmultiply-enabled mdivide-enabled msign-extend-enabled

Once you have the toolchain and RTEMS installed:

  • make sure the toolchain binary directory (e.g. /opt/rtems/bin) is in your PATH, so that the lm32-rtems4.11-* family of commands is found. 
export PATH=/opt/rtems-4.11/bin:$PATH
  • set the RTEMS_MAKEFILE_PATH environment variable, e.g. to /opt/rtems-4.11/lm32-rtems4.11/milkymist 
export RTEMS_MAKEFILE_PATH=/opt/rtems-4.11/lm32-rtems4.11/milkymist

[edit] libpng

We use libpng 1.5.6. Sometimes, GNU/Autocrap-infected software packages happen to work almost correctly, and libpng is one of those rare cases. 

CC=lm32-rtems4.11-gcc CFLAGS="-O9 -Wall -mbarrel-shift-enabled -mmultiply-enabled -mdivide-enabled -msign-extend-enabled -I $RTEMS_MAKEFILE_PATH/lib/include -B $RTEMS_MAKEFILE_PATH/lib -specs bsp_specs -qrtems" CPPFLAGS="-DPNG_NO_FLOATING_ARITHMETIC -DPNG_NO_FLOATING_POINT" ./configure --disable-shared --prefix=$RTEMS_MAKEFILE_PATH --host=lm32-rtems4.11
make
make install

Again, move the include files: 

mv /opt/rtems-4.11/lm32-rtems4.11/milkymist/include/* /opt/rtems-4.11/lm32-rtems4.11/milkymist/lib/include

[edit] libjpeg

This is known to work with jpegsrc v8c from IJG. 

wget http://www.ijg.org/files/jpegsrc.v8c.tar.gz

CFLAGS="-O9 -Wall -mbarrel-shift-enabled -mmultiply-enabled -mdivide-enabled -msign-extend-enabled -I $RTEMS_MAKEFILE_PATH/lib/include -B $RTEMS_MAKEFILE_PATH/lib -specs bsp_specs -qrtems" ./configure --host=lm32-rtems4.11 --disable-shared --prefix=$RTEMS_MAKEFILE_PATH
make
make install
mv /opt/rtems-4.11/lm32-rtems4.11/milkymist/include/* /opt/rtems-4.11/lm32-rtems4.11/milkymist/lib/include

[edit] libcurl

Tested with curl 7.22.0. Configure with the usual Autocrap command line (as above). Special parameters are: --host=lm32-rtems4.11 --disable-shared --without-librtmp --disable-ftp --disable-ldap --disable-ldaps --disable-rtsp --disable-dict --disable-telnet --disable-tftp --disable-pop3 --disable-imap --disable-smtp --disable-gopher --disable-manual --prefix=$RTEMS_MAKEFILE_PATH

Then you need to hack the generated files:

  • at the end of lib/setup.h add #define HAVE_SOCKET 1

[edit] expat

Works with expat 2.0.1. Edit conftools/config.sub and replace ppc with lm32, then proceed like with Freetype.

[edit] Softusb firmware

You will need gcc-avr.

Make sure the RTEMS_MAKEFILE_PATH environment variable is set (see above).

Run these commands from the top directory of the SoC distribution: 

 $ cd softusb-input
 $ make
 # make install

[edit] libFPVM

Flickernoise uses the FPVM compilation library to generate microcode for the PFPU. It is part of the SoC distribution.

You will need Lemon and RE2C to compile libFPVM.

Make sure the RTEMS_MAKEFILE_PATH environment variable is set (see above).

dont forget PATH too, ie PATH=/opt/rtems-4.11/bin:$PATH

Run these commands from the top directory of the SoC distribution: 

 $ cd software/libfpvm
 $ make
 # make install

[edit] MTK

MTK (Milkymist Toolkit) is a lightweight and portable GUI toolkit, based on Genode FX.

After you have cloned the git repository, compilation and installation are very simple: 

make milkymist
make install-milkymist

(Again, these commands requires the RTEMS_MAKEFILE_PATH environment variable)

[edit] liboscparse (lop)

Liboscparse is used to implement Opensoundcontrol (OSC). It is based on liblo, but it does not use the GNU/Autocrap "build system" and does not contain any network-dependent code (the original liblo networking code does not play nice with RTEMS).

Run the following commands (making sure the environment variables listed above are set): 

 make
 make install

[edit] Yaffs

rtems-yaffs2 YAFFS (Yet Another Flash File System) is now in its second generation and provides a fast robust file system for NAND and NOR Flash.

After you have cloned the git repository, compilation is very simple: 

make -f Makefile.rtems
make -f Makefile.rtems install

(Again, this command requires the RTEMS_MAKEFILE_PATH environment variable)

[edit] Flickernoise

Compile it with: 

 cd src
 make

Do not forget to set the RTEMS_MAKEFILE_PATH environment variable to the right directory where your RTEMS is installed.

This produces an ELF binary. From that, you can either:

  • try the ELF binary in the QEMU emulator with qemu-system-lm32 -M milkymist -kernel flickernoise
  • generate a raw binary file with make flickernoise.ralf, suitable for TFTP netbooting (recommended), booting from the memory card or downloading over the serial line with flterm. Using flterm is not recommended as this method is slow because of the large size of the binary.
  • generate a FBI image with make flickernoise.fbi that you can use with the auto-upgrade feature of Flickernoise (0.2+).
  • generate a MCS flash image with make flickernoise.mcs that you can download into the board's flash memory so that Flickernoise automatically starts at every boot.

[edit] Extra libraries (normally not needed)

[edit] freetype2

This is known to work with freetype 2.4.4. 

wget http://ufpr.dl.sourceforge.net/project/freetype/freetype2/2.4.4/freetype-2.4.4.tar.gz

CFLAGS="-O9 -Wall -mbarrel-shift-enabled -mmultiply-enabled -mdivide-enabled -msign-extend-enabled -I $RTEMS_MAKEFILE_PATH/lib/include -B $RTEMS_MAKEFILE_PATH/lib -specs bsp_specs -qrtems" ./configure --host=lm32-rtems4.11 --disable-shared --prefix=$RTEMS_MAKEFILE_PATH
make
make install
mv /opt/rtems-4.11/lm32-rtems4.11/milkymist/include/* /opt/rtems-4.11/lm32-rtems4.11/milkymist/lib/include

[edit] libgd

This works with GD 2.0.33.

RTEMS lacks the required include file bits/wordsize.h so we create one: 

mkdir /opt/rtems-4.11/lm32-rtems4.11/include/bits
cat << EOF > /opt/rtems-4.11/lm32-rtems4.11/include/bits/wordsize.h
#define __WORDSIZE 32
EOF

Edit the beginning of the Makefile to set: 

COMPILER=lm32-rtems4.11-gcc
AR=lm32-rtems4.11-ar
CFLAGS=-O9 -Wall -mbarrel-shift-enabled -mmultiply-enabled -mdivide-enabled -msign-extend-enabled -fsingle-precision-constant -I$(RTEMS_MAKEFILE_PATH)/lib/include -I/opt/rtems-4.11/lm32-rtems4.11/milkymist/lib/include/freetype2 -g -DHAVE_LIBPNG -DHAVE_LIBJPEG -DHAVE_LIBFREETYPE -DHAVE_FT2BUILD_H
LIBS=-lgd -lpng -lz -ljpeg -lfreetype -lm
INCLUDEDIRS=-I.
LIBDIRS=

Also, they seem to have forgotten gd_security.c to the list of objects, so add it to LIBOBJS.

Compile with: 

make libgd.a

Then copy libgd.a to /opt/rtems-4.11/lm32-rtems4.11/milkymist/lib and gd.h, gdcache.h, gd_io.h, gdfontg.h, gdfontl.h, gdfontmb.h, gdfonts.h and gdfontt.h to /opt/rtems-4.11/lm32-rtems4.11/milkymist/lib/include.

[edit] openjpeg

We use openjpeg version 1.2. 

wget http://www.openjpeg.org/openjpeg_v1_2.tar.gz

Uncompress the archive, and edit the Makefile:

  1. set CC = lm32-rtems4.11-gcc -mbarrel-shift-enabled -mmultiply-enabled -mdivide-enabled -msign-extend-enabled
  2. set AR = lm32-rtems4.11-ar
  3. for the target OpenJPEG, comment out $(SHAREDLIB)

RTEMS lacks memory.h so we create a fake one: 

echo "#include <string.h>" > libopenjpeg/memory.h

Type make to compile. Then ranlib and copy the library and its include file to the RTEMS directory: 

lm32-rtems4.11-ranlib libopenjpeg.a
cp libopenjpeg.a /opt/rtems-4.11/lm32-rtems4.11/milkymist/lib
cp libopenjpeg/openjpeg.h /opt/rtems-4.11/lm32-rtems4.11/milkymist/lib/include

[edit] jbig2dec

We use jbig2dec 0.11. 

wget http://ghostscript.com/~giles/jbig2/jbig2dec/jbig2dec-0.11.tar.gz

You will first need to re-generate config.sub using a recent version of automake. Indeed, the steaming pile of shit that constitutes GNU/Autocrap has a hardcoded list of CPUs it would support, and lm32 isn't included with many software packages. 

rm config.sub
automake --add-missing

Ignore any warning about version mismatches that automake may puke out - trying to fix it actually makes the problem worse.

Run the configure script, compile and install: 

CFLAGS="-O9 -Wall -mbarrel-shift-enabled \ 
-mmultiply-enabled -mdivide-enabled -msign-extend-enabled \
-I $RTEMS_MAKEFILE_PATH/lib/include \
-B $RTEMS_MAKEFILE_PATH/lib -specs bsp_specs -qrtems" \
./configure --host=lm32-rtems4.11 --disable-shared --prefix=$RTEMS_MAKEFILE_PATH
make
make install

Again, move the include file: 

mv /opt/rtems-4.11/lm32-rtems4.11/milkymist/include/* /opt/rtems-4.11/lm32-rtems4.11/milkymist/lib/include

[edit] mupdf

These instructions are known to work with mupdf 0.7. 

wget http://mupdf.com/download/mupdf-0.7.tar.gz

First, compile the code generation tools natively: 

gcc -o build/release/cmapdump ./mupdf/cmapdump.c  -lfreetype -ljpeg -lz -lm -Ifitz
gcc -o build/release/fontdump ./mupdf/fontdump.c  -lfreetype -ljpeg -lz -lm -Ifitz

Then, edit Makerules and:

  1. comment out the two lines at the beginning setting the OS variable
  2. set CC := lm32-rtems4.11-gcc
  3. set CFLAGS := -Ifitz -Imupdf -Wall -O9 -Wall -mbarrel-shift-enabled -mmultiply-enabled -mdivide-enabled -msign-extend-enabled -I $(RTEMS_MAKEFILE_PATH)/lib/include -I $(RTEMS_MAKEFILE_PATH)/lib/include/freetype2 -B $(RTEMS_MAKEFILE_PATH)/lib -specs bsp_specs -qrtems

Compile the MuPDF library with: 

make build=release build/release/libmupdf.a

Finally, copy the includes and binaries to the RTEMS folder: 

cp fitz/fitz.h /opt/rtems-4.11/lm32-rtems4.11/milkymist/lib/include
cp mupdf/mupdf.h /opt/rtems-4.11/lm32-rtems4.11/milkymist/lib/include
cp build/release/libmupdf.a /opt/rtems-4.11/lm32-rtems4.11/milkymist/lib

[edit] FFMPEG 

#undef __STRICT_ANSI__
#include <stdlib.h>

CFLAGS="-O9 -Wall -mbarrel-shift-enabled -mmultiply-enabled -mdivide-enabled -msign-extend-enabled -I $RTEMS_MAKEFILE_PATH/lib/include -B $RTEMS_MAKEFILE_PATH/lib -specs bsp_specs -qrtems" ./configure --cross-prefix=lm32-rtems4.11- --enable-cross-compile --target-os=none --arch=lm32 --disable-shared --disable-network --prefix=$RTEMS_MAKEFILE_PATH
Retrieved from "http://milkymist.org/wiki/index.php?title=Flickernoise_build_instructions"
Personal tools