In this blog post, for my software optimization and portability course, I will go over the process of installing GCC on two distinct architectures x86_64 and AArch64. I will go over my process of installation for each and show them working as intended.
I worked on two Linux environments with x86 and AArch64 architecture specifically.
Building GCC for x86_64
Obtaining the Source Code
First, I cloned the GCC git repo from https://gcc.gnu.org/git/gcc.git.
$ git clone https://gcc.gnu.org/git/gcc.git
This step took a couple of minutes as there were a lot of individual files to download
Creating a Build Directory:
I then created a new directory outside the source directory where I would have GCC installed and gave it an appropriate name.
$ mkdir x86_64_gcc_install
$ cd x86_64_gcc_install
Configuring GCC:
Before I could make the GCC installation in this new directory I needed to configure GCC using the configure script in the git repo which would also configure GCC for the current architecture.
$ cd x86_64_gcc_install
$ $PWD/../gcc/configure --prefix=$HOME/spo600/x86_64_gcc_install/
I set the configure file from within the install directory and also included the prefix argument to tell the configure script where I want GCC to be installed
Building GCC:
The makefile was used with the -j 4 argument, which set the amount of CPU cores during the installation to 4.
$ make -j 4
The make process for GCC on x86_64 architecture took almost 2 hours with four parallel jobs.
After this, the makefile and configuration options were added to the install directory. To finish the installation and create the working program I installed the binaries to the x86_64_gcc_install/bin directory:
$ make install
Building GCC for AArch64
I began building GCC on the other environment with AArch64 architecture instead of x86_64
Initial setup
The process for obtaining the source code and creating the install directory was the same
$ git clone https://gcc.gnu.org/git/gcc.git
Again, this step took a couple of minutes as there were a lot of individual files to download
$ mkdir aarch64_gcc_install
$ cd aarch64_gcc_install
Configuring GCC:
Once again, configuring GCC using the configure script in the git repo which would also configure GCC for the current architecture with the prefix argument for configure location.
$ cd x86_64_gcc_install
$ $PWD/../gcc/configure --prefix=$HOME/spo600/x86_64_gcc_install/
After this, the makefile and configuration options were added to the install directory
Building GCC:
I then used 5 CPU cores with the -j 5 option to see if there was any difference in speed.
$ make -j 5
With this, it took just under 2 hours and was marginally faster as I had added more multiple cores.
I then as with x86, installed binaries to aarch64_gcc_install/bin:
$ make install
Confirming installation
For each installation, the runnable programs were available in the install_dir/bin directory and to show that GCC was installed for the specific architecture I ran the architecture-specific program (usually including the x86/AArch64 name). For example here was the x86_64 bin folder:
To confirm that the newly built GCC compilers were fully functional, I created a Hello World C program in each environment which had the following structure:
#include <stdio.h>
int main(){
printf("Hello x86_64!\n");
return 0;
}
The x86_64 part is replaced with AArch64 in its version
Compiling and Running x86_64
To run a program using the installed GCC I used the relative path to the bin folder in my install directory. Usually when fresh installing GCC you would set the bin folder as a $PATH variable but in this case there was already a version of GCC in the environment.
To 100% confirm that it is built on x86_64 I used the file command on the object file:
Compiling and Running AArch64
Same as above I used the relative path name with the specific AArch64 compiler.
Confirming with the file command:
Conclusion
This experience serves as a testament to the power of open-source development and the versatility of GCC as a cross-compiler tool.
Top comments (0)