$ mkdir ~/Projects/GCC # compiler + libs
$ mkdir ~/Projects/GCC/libs # gmp, mpfr and mpc
$ mkdir ~/Projects/GCC/libs/files # downloaded files
$ mkdir ~/Projects/GCC/libs/install # store libs objects and include

Step #1 – Download

$ cd ~/Projects/GCC/libs/files
$ wget ftp://ftp.gmplib.org/pub/gmp-5.0.1/gmp-5.0.1.tar.bz2
$ wget http://www.mpfr.org/mpfr-current/mpfr-3.0.0.tar.bz2
$ wget http://www.multiprecision.org/mpc/download/mpc-0.8.2.tar.gz

Step #2 – Unpack

$ cd ~/Projects/GCC/libs/
$ tar jxvf files/gmp-5.0.1.tar.bz2
$ tar jxvf files/mpfr-3.0.0.tar.bz2
$ tar zxvf files/mpc-0.8.2.tar.gz

Step #3 – Build: GMP

$ cd ~/Projects/GCC/libs/

$ mkdir gmp-build
$ cd gmp-build
$ ../gmp-5.0.1/configure –prefix=$(cd ../install && pwd)
$ make install

Note 1: I’m using with ABI=64
Note 2 : Maybe you’ll get some unresolved symbols to GMP and MPFR on linking time, never mind.

Step #4 – Build: MPFR

$ cd ~/Projects/GCC/libs/

$ mkdir mpfr-build
$ cd mpfr-build
$ ../mpfr-3.0.0/configure –prefix=$(cd ../install && pwd) –with-gmp=~/Projects/GCC/libs/install
$ make install

Note: Before prefix and with-xxx there are two hyphen (not one as showed). It’s an Wordpress issue, I don’t know how to avoid concatenation.

Step #5 – Build: MPC

$ cd ~/Projects/GCC/libs/

$ mkdir mpc-build
$ cd mpc-build
$ ../mpc-0.8.2/configure –prefix=$(cd ../install && pwd) –with-gmp=~/Projects/GCC/libs/install –with-mprf=~/Projects/GCC/libs/install
$ make install

Problems? If something goes wrong…

../../mpc-0.8.2/src/mpc.h:25:17: error: gmp.h: No such file or directory
../../mpc-0.8.2/src/mpc.h:26:18: error: mpfr.h: No such file or directory

I fixed by adding symbolic links:

$ cd ~/Projects/GCC/libs/mpc-0.8.2/src/
$ ln -s ../../install/include/mpf2mpfr.h .
$ ln -s ../../install/include/mpfr.h .
$ ln -s ../../install/include/gmp.h .

Step #6 – Download & Build GCC (~4.6)

I used GCC from git (fda0037801fb258a2191aba59e1e9f0df019e3b6) and I don’t know if it will work on newer versions. You’ll have to try. Sorry. [Howto: GitMirror]. Use git checkout to specify one commit.

$ cd ~/Projects/GCC/

$ git clone git://gcc.gnu.org/git/gcc.git

$ mkdir build
$ mkdir install

$ cd build
$ ../gcc/configure –prefix=$(cd ../install/ && pwd) –with-gmp=/Users/maluta/Projects/GCC/libs/install –with-mpfr=/Users/maluta/Projects/GCC/libs/install –with-mpc=/Users/maluta/Projects/GCC/libs/install –disable-checking –enable-werror –enable-languages=c

$ make
$ make install

Step #7 – Test

The binary files will be placed in ~/Projects/GCC/install/bin

$ ~/Projects/GCC/install/bin
$ ./gcc -v

Using built-in specs.
COLLECT_GCC=./gcc
COLLECT_LTO_WRAPPER=/Users/maluta/Projects/GCC/install/libexec/gcc/x86_64-apple-darwin10.4.0/4.6.0/lto-wrapper
Target: x86_64-apple-darwin10.4.0
Configured with: ../gcc/configure –prefix=/Users/maluta/Projects/GCC/install –with-gmp=/Users/maluta/Projects/GCC/libs/install –with-mpfr=/Users/maluta/Projects/GCC/libs/install –with-mpc=/Users/maluta/Projects/GCC/libs/install –disable-checking –enable-werror –enable-languages=c
Thread model: posix
gcc version 4.6.0 20100508 (experimental) (GCC)

Although I focused on OSX build/installation the steps described are the same to any architecture. Build an compiler – especially a cross-compiler (no that case) – demands time and patience to understand many particularities.

Enjoy

$ python stats_opcodes.py file

For Cortex-m3 (check previous post):

At my laptop (x86):

Download the script here.

• IBM DeveloperWorks about Inline asssembly for x86 in Linux
• Linux Documentation Project HOWTO describing GCC Inline Assembly

The syntax may be confusing, if you don’t understand read the documentation available. Each example are an C function with the special construct “asm” for inline assembly code. Examples:

Increment an value:

void inc(int *value) {
 
    asm ( "incl %0"
         :"=a"(*value)
         :"0"(*value)
        );
 
}

Exchange to numbers:

void swap(int *x, int *y) {
 
    asm(""
        : "=a"(*y),"=b"(*x)
        : "a"(*x),"b"(*y)
       );
 
}

Copy an vector:

void vector_copy(int *v_src, int *v_dst, int *count) {
 
    asm("up: lodsl;"
	"    stosl;"
	"loop up;  "
	:
	: "S"(v_src), "D"(v_dst), "c"(*count)
	: "%eax"
	);
 
}

A simple copy:

void copy(int *from, int *to) {
 
     asm ("movl %1, %%eax;"
          "movl %%eax, %0;"
          :"=&r"(*to)
          :"r"(*from));
 
}

A very simple implementation for strncpy:

void _strncpy(char *src, char *dst, int *count) {
 
    asm("cld;"
        "rep movsb;"
        :
        : "S"(src), // 'S' == %esi
	  "D"(dst), // 'D' == %edi
	  "c"(*count)); // 'c' == %ecx
}

If you like to try check this example:

wget http://github.com/maluta/junk/raw/master/asm-inline.c
gcc -Wall -ggdb asm-inline.c -o asm-inline
./asm-inline