Good Day Mentors,
I hope this email finds you well.
I am writing to share my progress on Experiment-0, specifically the OpenMP
section. Here is a summary of my work so far:
1. Cloning GCC and building libgomp
-
I cloned the GCC source repository and switched to the devel/omp/gcc-15
branch.
-
I configured the build to compile only the *C and C++ compilers*, using
the --prefix option to install it in /home/fureh/gcc-install to avoid
conflicts with my global GCC installation.
-
To verify the build, I tested a simple program that created and printed
the number of OpenMP threads. The results confirmed the expected behavior
from the thesis: *the number of OpenMP threads equals the number of
logical CPUs*.
2. Compiling the OpenMP UA benchmark (NPB v3.4.4, Class A)
-
I downloaded and extracted the NPB v3.4.4 tarball and created a copy of
the configuration file.
-
I initially compiled the UA benchmark using the *global GCC compiler*.
-
I set the number of OpenMP threads via the OMP_NUM_THREADS environment
variable and successfully ran the ua.A.x binary.
3. Running the benchmark with the custom-built libgomp
-
I modified the UA benchmark config file to point to my custom GCC
compiler:
CC = /home/fureh/gcc-install/bin/gcc
CLINK = /home/fureh/gcc-install/bin/gcc
CFLAGS = -O3 -fopenmp
-
I exported the number of threads using OMP_NUM_THREADS and ran the ua.A.x
binary.
-
The benchmark successfully executed using my custom-built libgomp,
confirming that the setup works with my local GCC installation.
I am currently reading through *Chapter 4 of the thesis* and will provide
an update on the next part of *Experiment-0* soon.
Kindly let me know if I should include *screenshots or output logs* in the
next submissions to better illustrate my progress.
Best regards,
Nchang Roy
