marvell-doc/MARVELL_Toolchain_Release_Note.txt - toolchains/armada - Git at Google

 Release Note:	Marvell GCC C/C++ Compiler Toolchain
 Release:	2013Q2 (gcc version 4.6.4)
 Date:		2013/06
 Directory:

 	0. Marvell GCC Toolchain
 	1. Marvell GCC Release Website
 	2. Binary Packages
 	3. Source Code Packages
 	4. Support
 	5. Feature Listing


 --------------------------------------------------------------------------------
 0. Marvell GCC Toolchain
 --------------------------------------------------------------------------------

     The Marvell GCC C/C++ Compiler Toolchain (Marvell GCC, or MGCC) is a
 collection of open source tools include gcc C/C++ compilers, binutils, C/C++
 libraries, gdb debugger, etc.  Marvell makes some bug-fixes and tuned features
 especially for the Marvell CPUs that are ARMv5te/ARMv7-a compatible.

     Release 2013Q2 of MGCC is the alpha release with the major tool GCC 4.6.4.

     Note that Marvell only collects and tunes these packages, and all products
 or services mentioned herin may be trademarks of their respective owners.


 --------------------------------------------------------------------------------
 1. Marvell GCC Release Website
 --------------------------------------------------------------------------------

     MGCC source code and pre-built executables for some platforms can be
 download from the website:

 	http://10.19.133.165/release/gcc/

     The details will be mentioned in the following context.


 --------------------------------------------------------------------------------
 2. Binary Packages
 --------------------------------------------------------------------------------

     Download the binary toolchains from the MGCC Release Website according to
 the following rules/subdirectories:

 (1) The MGCC Major Version

     For example:

 	mgcc-4.2/
 	mgcc-4.4/
 	mgcc-4.6/

     These are according to the GCC, the most important tool, version number.

 (2) Tool Building Version

     They are always named by 'YYYYMM-NNN-VERSION'.  For example:

 	201208-1593-4.6.4/

     The newer version has fixed more bugs and add more Marvell-tuned features.

 (3) The Default Target Endian

 	big-endian/
 	little-endian/

     'big-endian' includes ARMv5's BE32 and ARMv7's BE8 versions.  The MGCC
     runtime libraries and C/C++ libraries comes from big-endian subdirectory
     are compiled with the default 'gcc -mbig-endian' option, while from
     little-endian with the default 'gcc -mlittle-endian' option.

     There may be extra MGCC toolchains for other target.  For example, the
     Android target is at:

 	android/

     Note that users can also get the MGCC source code and extra documents
     from the subdirectories:

 	src/
 	doc/

 (4) The Default Target ARM Instruction Set

     For example:

 	v5/
 	v7/

     MGCC from v5 subdirectory generates ARMv5 instruction set, while v7
     generates ARMv7 instructions set by default.

 (5) The Host Execution Environment

     For example:

 	host-arm-linux/
 	host-i686-linux/
 	host-windows-mingw32/
 	host-x86_64-linux/
 	...

     Select suitable hardware and O.S platform on which MGCC would be executed.

     For not mentioned host environment, users may download the MGCC source
     code and build MGCC on that host.  If fails to build, please ask Marvell
     Compiler Team for help.

 (6) The Binary Package

     They are always named by

 	arm<ARCH>-marvell-<ABI>-<VFP>_<VERSION>_<HOST>_<YYYYMM>.tar.bz2

     with different target features.  For example:

 	...
 	armebv5-marvell-native-softfp-4.6.4_201403.tar.bz2
 	armv5-marvell-linux-gnueabi-soft-4.6.4_i686_201403.tar.bz2
 	armebv7-marvell-eabi-hard-4.6.4_x86_64_201403.tar.bz2
 	armv7-marvell-linux-gnueabi-soft-mingw32-4.6.4_201413.zip
 	...

     <ARCH> : v5, v7, ebv5, ebv7.

 	'v5' means ARMv5 instruction set, and while 'v7' means ARMv7.

 	The prefix 'armv5' or 'armv7' is the little-endian target, 'armebv5'
 	is BE32, and 'armebv7' is BE8.

     <ABI> : native, linux-gnueabi, eabi

 	'native' is the native-compilers running on ARM Linux hosts and
 	generating the target ARM code.  The 'gcc -mabi=aapcs-linux' is
 	used by default.  The accompanying C library is glibc.  The compiled
 	programes would be executed on the Linux environment with the
 	compatible glibc shared library.

 	'linux-gnueabi' with the default 'gcc -mabi=aapcs-linux' is always for
 	the target of ARM Linux platform.  The accompanying C library is glibc.
 	The compiled programes would be executed on the ARM Linux environment
 	with the compatible glibc shared library.

 	'eabi' with the default 'gcc -mabi=aapcs' is always for the target of
 	ARM bare board or semihosting environment.  The accompanying C library
 	is Newlib.  If you are developing embedded system without any O.S.
 	support, you should select this package to develop your programs.

     <VFP> : hard, soft, softfp

 	The 'hard', 'soft' and 'softfp' mean the default 'gcc -mfloat-abi='
 	option, and the libraries, e.g. glibc and newlib, are compiled with
 	that default option.

 	'soft' tool generates non-VFP code, and the others may generate
 	VFP/Neon code.  Objects compiled by 'soft' and 'softfp' are
 	compatible, but not 'hard'.

     <VERSION> : 4.6.4, 4.8.3, ...

 	This specifies the GCC VERSION on which the MGCC toolchain executes.

     <HOST> : i686, x86_64, mingw32

 	This specifies the host on which the MGCC toolchain executes.


     After downloading the suitable package, use 'tar' or other unpacking tools
 to unpack it.  E.g.

 	$ tar jxvf armv7-marvell-linux-gnueabi-softfp-4.6.4_x86_64_201403.tar.bz2
 	$ cd armv7-marvell-linux-gnueabi-softfp-4.6.4_x86_64_201403
 	$ ls bin/				# executable files
 	$ ls arm-marvell-linux-gnueabi/libc	# sysroot

     For 'linux-gnueabi' version, users may need to update the shared libraries
 (*.so) from the MGCC sysroot on the target ARM Linux environment if the
 compiled program fails to run because of wrong *.so version.


 --------------------------------------------------------------------------------
 3. Source Code Packages
 --------------------------------------------------------------------------------

     Download the toolchain source code from the Marvell GCC Release Website,
 and untar them.

     The below listing is the source code packages and where they come from:

     marvell-gcc			Marvell-tuned GCC C/C++ Compiler Source Code
     |
     +-- script			Building scripts and documents
     |   |
     |   +-- doc
     |
     +-- src
         |
         +-- binutils-src	Marvell-tuned Binutils originally from FSF.
         |                       http://www.gnu.org/software/binutils/
         |
         +-- cloog-src		CLooG from FSF for building GCC.
         |			http://www.cloog.org/
         |			ftp://gcc.gnu.org/pub/gcc/infrastructure/
         |
         +-- expat-src		The Expat XML Parser.
         |			http://www.libexpat.org/
         |
         +-- extra-src		Extra packages as addends to the MGCC
         |			toolchains for different purposes.
         |
         +-- gcc-src		Marvell-tuned GCC originally from FSF and
         |			Linaro.
         |			http://gcc.gnu.org/
         |			http://www.linaro.org/
         |
         +-- gdb-src		GDB from FSF and Linaro.
         |			http://www.gnu.org/software/gdb/
         |			http://www.linaro.org/
         |
         +-- glibc-ports-src	GNU glibc 2.15 from CodeSourcery for ARM
 	|			porting.
         |
         +-- glibc-src		GNU glibc 2.15 from CodeSourcery for
 	|			arm-marvell-linux-gnueabi.
         |
         +-- gmp-src		The GNU Multiple Precision Arithmetic Library
         |			for building GCC.
         |			http://gmplib.org/
         |			http://ftp.gnu.org/gnu/gmp/
         |
         +-- linux-src		Header files for arm-marvell-linux-gnueabi.
         |			https://www.kernel.org/
         |
         +-- mpc-src		GNU MPC Library from FSF for building GCC.
         |			http://www.multiprecision.org/
         |			http://ftp.gnu.org/gnu/mpc/
         |
         +-- mpfr-src		GNU MPFR Library for building GCC.
         |			http://www.mpfr.org/
         |			http://ftp.gnu.org/gnu/mpfr/
         |
         +-- ncurses-src		Ncurses (new curses) library from FSF.
         |			http://www.gnu.org/software/ncurses/
         |			http://ftp.gnu.org/pub/gnu/ncurses/
 	|
         +-- newlib-src		Newlib for arm-marvell-eabi.
         |			http://www.sourceware.org/newlib/
         |
         +-- ppl-src		Parma Polyhedra Library (PPL) for building GCC.
 	|			http://bugseng.com/products/ppl/
         |
 	+-- uclibc-src		A C library for embedded Linux.
 	|			http://www.uclibc.org/
 	|
 	+-- zlib-src		zlib data compression library.
 				http://www.zlib.net/
 				http://www.gzip.org/

     Extra packages (src/extra-src) is as addends to the MGCC toolchains.  Refer
 to the documents "MARVELL_GCC_Extra_Libraries_for_Various_Platforms.txt" and
 "MARVELL_GCC_Dump_Optimization_Options_Plugin.txt" for details.

     Refer to 'marvell-gcc/script/INSTALL' for building toolchains.  The
 following commands show the supported targets and simple building commands:

 	$ cd marvell-gcc
 	$ make -f script/Makefile.marvell

     For any qeuestions and special request, please ask FAE or Marvell Compiler
 Team for help.


 --------------------------------------------------------------------------------
 4. Support
 --------------------------------------------------------------------------------

 . Internal Bug Report Email:

 	marvell-gcc-bugs@lists.marvell.com


 . Internal Bug Report Mailing List:

 	http://lists.marvell.com/mailman/listinfo/marvell-gcc-bugs

 . External Bug Report Email:

 	mrvlgccbug@gmail.com


 --------------------------------------------------------------------------------
 5. Feature Listing
 --------------------------------------------------------------------------------


 . MGCC toolchains building process for different hosts and configured with
   multiple target features.


 . Support Marvell's extension instructions.


 . Support Marvell's WMMX intrinsics and auto-vectorization.  For details,
   refer to the following document in the ${MGCC_TOOCHAINS}/script/doc/
   directory:

 	MARVELL_Using_GCC_IWMMXT_Autovect.txt
 	MARVELL_Using_GCC_IWMMXT.doc
 	MARVELL_Using_GCC_IWMMXT_Intrinsic.pdf
 	MARVELL_Using_GCC_IWMMXT.pdf
 	MARVELL_Using_GCC_IWMMXT_Vector_Class_Header_File.pdf
 	MARVELL_Using_GCC_Neon_Auto_Vectorization.pdf


 . Optimized instruction scheduling for Marvell CPUs including PJ1 and PJ4
   series.


 . Marvell-specified options.  For details, refer to:

 	${MGCC_TOOCHAINS}/script/doc/MARVELL_GCC_Options.txt


 . Refer to the file named 'ChangeLog.Marvell' in the source code directories
   for the modification. E.g.

 	${MGCC_TOOCHAINS}/src/binutils-src/ChangeLog.Marvell
 	${MGCC_TOOCHAINS}/src/gcc-src/ChangeLog.Marvell

   These ChangeLogs are only simple log from our git repertory.


 . For the changes, new features, and fixes of official gcc from 4.4 to 4.6,
   please refer to:

 	http://gcc.gnu.org/gcc-4.5/changes.html
 	http://gcc.gnu.org/gcc-4.6/changes.html
	Release Note: Marvell GCC C/C++ Compiler Toolchain
	Release: 2013Q2 (gcc version 4.6.4)
	Date: 2013/06
	Directory:

	0. Marvell GCC Toolchain
	1. Marvell GCC Release Website
	2. Binary Packages
	3. Source Code Packages
	4. Support
	5. Feature Listing





	--------------------------------------------------------------------------------
	0. Marvell GCC Toolchain
	--------------------------------------------------------------------------------

	The Marvell GCC C/C++ Compiler Toolchain (Marvell GCC, or MGCC) is a
	collection of open source tools include gcc C/C++ compilers, binutils, C/C++
	libraries, gdb debugger, etc. Marvell makes some bug-fixes and tuned features
	especially for the Marvell CPUs that are ARMv5te/ARMv7-a compatible.

	Release 2013Q2 of MGCC is the alpha release with the major tool GCC 4.6.4.

	Note that Marvell only collects and tunes these packages, and all products
	or services mentioned herin may be trademarks of their respective owners.





	--------------------------------------------------------------------------------
	1. Marvell GCC Release Website
	--------------------------------------------------------------------------------

	MGCC source code and pre-built executables for some platforms can be
	download from the website:

	http://10.19.133.165/release/gcc/

	The details will be mentioned in the following context.





	--------------------------------------------------------------------------------
	2. Binary Packages
	--------------------------------------------------------------------------------

	Download the binary toolchains from the MGCC Release Website according to
	the following rules/subdirectories:

	(1) The MGCC Major Version

	For example:

	mgcc-4.2/
	mgcc-4.4/
	mgcc-4.6/

	These are according to the GCC, the most important tool, version number.

	(2) Tool Building Version

	They are always named by 'YYYYMM-NNN-VERSION'. For example:

	201208-1593-4.6.4/

	The newer version has fixed more bugs and add more Marvell-tuned features.

	(3) The Default Target Endian

	big-endian/
	little-endian/

	'big-endian' includes ARMv5's BE32 and ARMv7's BE8 versions. The MGCC
	runtime libraries and C/C++ libraries comes from big-endian subdirectory
	are compiled with the default 'gcc -mbig-endian' option, while from
	little-endian with the default 'gcc -mlittle-endian' option.

	There may be extra MGCC toolchains for other target. For example, the
	Android target is at:

	android/

	Note that users can also get the MGCC source code and extra documents
	from the subdirectories:

	src/
	doc/

	(4) The Default Target ARM Instruction Set

	For example:

	v5/
	v7/

	MGCC from v5 subdirectory generates ARMv5 instruction set, while v7
	generates ARMv7 instructions set by default.

	(5) The Host Execution Environment

	For example:

	host-arm-linux/
	host-i686-linux/
	host-windows-mingw32/
	host-x86_64-linux/
	...

	Select suitable hardware and O.S platform on which MGCC would be executed.

	For not mentioned host environment, users may download the MGCC source
	code and build MGCC on that host. If fails to build, please ask Marvell
	Compiler Team for help.

	(6) The Binary Package

	They are always named by

	arm<ARCH>-marvell-<ABI>-<VFP>_<VERSION>_<HOST>_<YYYYMM>.tar.bz2

	with different target features. For example:

	...
	armebv5-marvell-native-softfp-4.6.4_201403.tar.bz2
	armv5-marvell-linux-gnueabi-soft-4.6.4_i686_201403.tar.bz2
	armebv7-marvell-eabi-hard-4.6.4_x86_64_201403.tar.bz2
	armv7-marvell-linux-gnueabi-soft-mingw32-4.6.4_201413.zip
	...

	<ARCH> : v5, v7, ebv5, ebv7.

	'v5' means ARMv5 instruction set, and while 'v7' means ARMv7.

	The prefix 'armv5' or 'armv7' is the little-endian target, 'armebv5'
	is BE32, and 'armebv7' is BE8.

	<ABI> : native, linux-gnueabi, eabi

	'native' is the native-compilers running on ARM Linux hosts and
	generating the target ARM code. The 'gcc -mabi=aapcs-linux' is
	used by default. The accompanying C library is glibc. The compiled
	programes would be executed on the Linux environment with the
	compatible glibc shared library.

	'linux-gnueabi' with the default 'gcc -mabi=aapcs-linux' is always for
	the target of ARM Linux platform. The accompanying C library is glibc.
	The compiled programes would be executed on the ARM Linux environment
	with the compatible glibc shared library.

	'eabi' with the default 'gcc -mabi=aapcs' is always for the target of
	ARM bare board or semihosting environment. The accompanying C library
	is Newlib. If you are developing embedded system without any O.S.
	support, you should select this package to develop your programs.

	<VFP> : hard, soft, softfp

	The 'hard', 'soft' and 'softfp' mean the default 'gcc -mfloat-abi='
	option, and the libraries, e.g. glibc and newlib, are compiled with
	that default option.

	'soft' tool generates non-VFP code, and the others may generate
	VFP/Neon code. Objects compiled by 'soft' and 'softfp' are
	compatible, but not 'hard'.

	<VERSION> : 4.6.4, 4.8.3, ...

	This specifies the GCC VERSION on which the MGCC toolchain executes.

	<HOST> : i686, x86_64, mingw32

	This specifies the host on which the MGCC toolchain executes.


	After downloading the suitable package, use 'tar' or other unpacking tools
	to unpack it. E.g.

	$ tar jxvf armv7-marvell-linux-gnueabi-softfp-4.6.4_x86_64_201403.tar.bz2
	$ cd armv7-marvell-linux-gnueabi-softfp-4.6.4_x86_64_201403
	$ ls bin/ # executable files
	$ ls arm-marvell-linux-gnueabi/libc # sysroot

	For 'linux-gnueabi' version, users may need to update the shared libraries
	(*.so) from the MGCC sysroot on the target ARM Linux environment if the
	compiled program fails to run because of wrong *.so version.





	--------------------------------------------------------------------------------
	3. Source Code Packages
	--------------------------------------------------------------------------------

	Download the toolchain source code from the Marvell GCC Release Website,
	and untar them.

	The below listing is the source code packages and where they come from:

	marvell-gcc Marvell-tuned GCC C/C++ Compiler Source Code
	\|
	+-- script Building scripts and documents
	\| \|
	\| +-- doc
	\|
	+-- src
	\|
	+-- binutils-src Marvell-tuned Binutils originally from FSF.
	\| http://www.gnu.org/software/binutils/
	\|
	+-- cloog-src CLooG from FSF for building GCC.
	\| http://www.cloog.org/
	\| ftp://gcc.gnu.org/pub/gcc/infrastructure/
	\|
	+-- expat-src The Expat XML Parser.
	\| http://www.libexpat.org/
	\|
	+-- extra-src Extra packages as addends to the MGCC
	\| toolchains for different purposes.
	\|
	+-- gcc-src Marvell-tuned GCC originally from FSF and
	\| Linaro.
	\| http://gcc.gnu.org/
	\| http://www.linaro.org/
	\|
	+-- gdb-src GDB from FSF and Linaro.
	\| http://www.gnu.org/software/gdb/
	\| http://www.linaro.org/
	\|
	+-- glibc-ports-src GNU glibc 2.15 from CodeSourcery for ARM
	\| porting.
	\|
	+-- glibc-src GNU glibc 2.15 from CodeSourcery for
	\| arm-marvell-linux-gnueabi.
	\|
	+-- gmp-src The GNU Multiple Precision Arithmetic Library
	\| for building GCC.
	\| http://gmplib.org/
	\| http://ftp.gnu.org/gnu/gmp/
	\|
	+-- linux-src Header files for arm-marvell-linux-gnueabi.
	\| https://www.kernel.org/
	\|
	+-- mpc-src GNU MPC Library from FSF for building GCC.
	\| http://www.multiprecision.org/
	\| http://ftp.gnu.org/gnu/mpc/
	\|
	+-- mpfr-src GNU MPFR Library for building GCC.
	\| http://www.mpfr.org/
	\| http://ftp.gnu.org/gnu/mpfr/
	\|
	+-- ncurses-src Ncurses (new curses) library from FSF.
	\| http://www.gnu.org/software/ncurses/
	\| http://ftp.gnu.org/pub/gnu/ncurses/
	\|
	+-- newlib-src Newlib for arm-marvell-eabi.
	\| http://www.sourceware.org/newlib/
	\|
	+-- ppl-src Parma Polyhedra Library (PPL) for building GCC.
	\| http://bugseng.com/products/ppl/
	\|
	+-- uclibc-src A C library for embedded Linux.
	\| http://www.uclibc.org/
	\|
	+-- zlib-src zlib data compression library.
	http://www.zlib.net/
	http://www.gzip.org/

	Extra packages (src/extra-src) is as addends to the MGCC toolchains. Refer
	to the documents "MARVELL_GCC_Extra_Libraries_for_Various_Platforms.txt" and
	"MARVELL_GCC_Dump_Optimization_Options_Plugin.txt" for details.

	Refer to 'marvell-gcc/script/INSTALL' for building toolchains. The
	following commands show the supported targets and simple building commands:

	$ cd marvell-gcc
	$ make -f script/Makefile.marvell

	For any qeuestions and special request, please ask FAE or Marvell Compiler
	Team for help.





	--------------------------------------------------------------------------------
	4. Support
	--------------------------------------------------------------------------------

	. Internal Bug Report Email:

	marvell-gcc-bugs@lists.marvell.com


	. Internal Bug Report Mailing List:

	http://lists.marvell.com/mailman/listinfo/marvell-gcc-bugs

	. External Bug Report Email:

	mrvlgccbug@gmail.com



	--------------------------------------------------------------------------------
	5. Feature Listing
	--------------------------------------------------------------------------------


	. MGCC toolchains building process for different hosts and configured with
	multiple target features.


	. Support Marvell's extension instructions.


	. Support Marvell's WMMX intrinsics and auto-vectorization. For details,
	refer to the following document in the ${MGCC_TOOCHAINS}/script/doc/
	directory:

	MARVELL_Using_GCC_IWMMXT_Autovect.txt
	MARVELL_Using_GCC_IWMMXT.doc
	MARVELL_Using_GCC_IWMMXT_Intrinsic.pdf
	MARVELL_Using_GCC_IWMMXT.pdf
	MARVELL_Using_GCC_IWMMXT_Vector_Class_Header_File.pdf
	MARVELL_Using_GCC_Neon_Auto_Vectorization.pdf


	. Optimized instruction scheduling for Marvell CPUs including PJ1 and PJ4
	series.


	. Marvell-specified options. For details, refer to:

	${MGCC_TOOCHAINS}/script/doc/MARVELL_GCC_Options.txt


	. Refer to the file named 'ChangeLog.Marvell' in the source code directories
	for the modification. E.g.

	${MGCC_TOOCHAINS}/src/binutils-src/ChangeLog.Marvell
	${MGCC_TOOCHAINS}/src/gcc-src/ChangeLog.Marvell

	These ChangeLogs are only simple log from our git repertory.


	. For the changes, new features, and fixes of official gcc from 4.4 to 4.6,
	please refer to:

	http://gcc.gnu.org/gcc-4.5/changes.html
	http://gcc.gnu.org/gcc-4.6/changes.html