On Tuesday, March 10th, 2026 at 4:50 AM, vatsal darji <[email protected]> wrote:
> Subject: Re: GSoC Experiment-0 Progress Update - OpenMP Phase Complete > > Hello all mentors, > > I apologise for my previous mistake of not adding all the mentors, and thank > you for the detailed response and for pointing me to the updated project > page, the todo document, and the thesis. I have been going through all of > them carefully and they have given me a much deeper understanding of the > problem space. > > I have also started studying the libgomp functions you mentioned : > gomp_dynamic_max_threads(), gomp_team_start(), and gomp_thread_start() as > well as the barrier synchronization implementation in wait.h, bar.c, and > bar.h. The thesis, even in its placeholder form, has been an excellent > companion while reading through the codebase. > > I also wanted to update you on my progress with Experiment-0: > > I have completed the OpenMP phase: > 1. Cloned GCC from the devel/omp/gcc-15 branch and built libgomp from source. > 2. Compiled the UA benchmark (Class A) from NPB v3.4.4 against the > custom-built libgomp. > 3. Successfully ran the benchmark with OMP_NUM_THREADS=4 verification passed > with a residual error of ~3×10⁻¹⁴. > Key results: > - Time in seconds: 27.45 > - Mop/s total: 19.83 > - Verification: SUCCESSFUL > I have attached the terminal output and ldd confirmation showing that the > custom libgomp is being loaded correctly. The success message in the benchmark output simply implies that things run without any issues. So good job at confirming that the benchmark indeed uses your custom-built libgomp. > I am currently working on the Virtualization phase (QEMU/KVM setup) and will > then continue to the tracing and scheduler phase, I'll send another update > once the rest of Experiment-0 is completed. Looking forward to it. Himadri
