You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
<a href="https://github.com/ParCoreLab/" class="text-xl font-semibold font-sans visited:text-teal-700">Unified Communication Library</a>
592
+
<a href="https://github.com/ParCoreLab/Uniconn" class="text-xl font-semibold font-sans visited:text-teal-700">Unified Communication Library</a>
564
593
</div>
565
594
<p class="text-lg">We're undertaking the design of an API for a unified communication library to streamline device-to-device communication within the CPU-free model by aiming to optimize communication efficiency across diverse devices. We are also investigating how the available communication libraries for a system perform under different
566
595
message sizes and communication patterns. Thus, we ex-
567
596
tensively benchmark current communication methods for
568
597
single-process, multi-threaded, and multi-process codes. More details about the project will be available soon. The related paper is under preparation.</p>
598
+
599
+
<p>
600
+
<a href="https://github.com/ParCoreLab/Uniconn" class="text-xl font-semibold font-sans visited:text-teal-700">More details and git repository of the project.</a>
<divclass="card text-lg"> Mohamed Wahib, Muhammed Abdullah Soyturk, Didem Unat (2025) <ahref="https://arxiv.org/pdf/2505.14864">Balanced and Elastic End-to-end Training of Dynamic LLMs</a>. <spanclass="italic"> ACM publication is pending</span>. <aclass="italic"downloadhref="./assets/preprint-pdfs/SC25_____Balanced_and_Elastic_End_to_end_Training_of_Dynamic_LLMs-4.pdf">preprint pdf</a> </div>
669
+
670
+
<divclass="card text-lg">Didem Unat, Anshu Dubey, Emmanuel Jeannot, John Shalf (2025) <ahref="https://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=10990038">The Persistent Challenge of Data locality in Post-Exascale Era</a>. In <spanclass="italic">Computing in Science & Engineering</span>. <aclass="italic"downloadhref="./assets/preprint-pdfs/Data_locality_CiSE___Camera_ready-4.pdf">preprint pdf</a> </div>
671
+
672
+
<divclass="card text-lg"> James D. Trotter, Sinan Ekmekçibaşı, Doğan Sağbili, Johannes Langguth, Xing Cai, Didem Unat (2025) CPU- and GPU-initiated Communication Strategies for Conjugate
673
+
Gradient Methods on Large GPU Clusters. In SC ’25: Proceedings of the International Conference for High Performance Computing, Networking, Storage and Analysis. <aclass="italic"downloadhref="./assets/preprint-pdfs/SC25_Inno4Scale_aCG.pdf">preprint pdf</a>
674
+
</div>
675
+
676
+
<divclass="card text-lg"> Doǧan Sağbili, Sinan Ekmekçibaşı, Khaled Z. Ibrahim, Tan Nguyen, Didem Unat (2025) UNICONN: A Uniform High-Level Communication
677
+
Library for Portable Multi-GPU Programming
678
+
<ahref="https://docs.google.com/presentation/d/1Tw4Yl8SLUjSDQwgHEITXthKg_KbjhVUZ3b9he2QTlj4/edit?usp=sharing">(presentation)</a>. In Cluster ’25: Proceedings of the IEEE International Conference on Cluster Computing (IEEE Cluster 2025). <aclass="italic"downloadhref="./assets/preprint-pdfs/Cluster_2025_______uniconn_paper__ieee_.pdf">preprint pdf</a>
679
+
</div>
680
+
681
+
<divclass="card text-lg"> Ilyas Turimbetov, Mohamed Wahib, Didem Unat (2025) <ahref="https://dl.acm.org/doi/10.1145/3721145.3730426">A Device-Side Execution Model for Multi-GPU Task
682
+
Graphs</a> <ahref="https://docs.google.com/presentation/d/1po87zQeUQb5l12AXB5RMSuod-o8yPZw32kEBtczr-v0/edit?usp=sharing">(presentation)</a>. In ICS ’25: Proceedings of the 39th ACM International Conference on Supercomputing. <aclass="italic"downloadhref="./assets/preprint-pdfs/ICS25______CPU_free_Task_Graph_Execution.pdf">preprint pdf</a>
683
+
</div>
684
+
685
+
<divclass="card text-lg"> Fatih Taşyaran, Osman Yasal, José A Morgado, Aleksandar Ilic, Didem Unat, Kamer Kaya (2024) <ahref="https://dl.acm.org/doi/10.1109/SCW63240.2024.00193">P-MoVE: performance monitoring and visualization with encoded knowledge</a> <ahref="https://docs.google.com/presentation/d/13DueK7-flbTfcH_1vKseGdIUCeZYGr0sKDs6s5HZEgo/edit?usp=sharing">(presentation)</a>. In <spanclass="italic">SC24-W: Workshops of the International Conference for High Performance Computing, Networking, Storage, and Analysis</span>. <aclass="italic"downloadhref="./assets/preprint-pdfs/RealTimeVisualization___SC_Workshop__.pdf">preprint pdf</a> </div>
686
+
687
+
<divclass="card text-lg">Didem Unat, Ilyas Turimbetov, Mohammed Kefah Taha Issa, Doğan Sağbili, Flavio Vella, Daniele De Sensi, Ismayil Ismayilov (2024) <ahref="https://arxiv.org/pdf/2409.09874">The landscape of gpu-centric communication</a>. <spanclass="italic">Under review</span>. <aclass="italic"href="https://arxiv.org/pdf/2409.09874">preprint pdf</a> </div>
<divclass="card text-lg"> Javid Baydamirli, Tal Ben Nun, Didem Unat (2024) <ahref="https://ieeexplore.ieee.org/abstract/document/10820747">Autonomous Execution for Multi-GPU Systems:
692
+
Compiler Support</a>. In SC24-W: Workshops of the International Conference for High Performance Computing, Networking, Storage and Analysis. <aclass="italic"downloadhref="./assets/preprint-pdfs/P3HPC_____Autonomous_Execution_for_Multi_GPU_Systems__Compiler_Support-2 (1).pdf">preprint pdf</a>
693
+
</div>
635
694
<divclass="card text-lg"> Javid Baydamirli, Tal Ben Nun, Didem Unat (2024) <ahref="https://sc24.supercomputing.org/proceedings/workshops/workshop_pages/ws_p3hpc108.html">Autonomous Execution for Multi-GPU Systems:
636
695
Compiler Support</a> <ahref="https://sc24.conference-program.com/presentation/?id=ws_p3hpc108&sess=sess751">(presentation)</a>. In the 2024 International Workshop on Performance, Portability, and Productivity in HPC. <aclass="italic"downloadhref="./assets/preprint-pdfs/sc24-workshop-autonomous-execution-for-multi-gpu-systems-compiler-support.pdf">preprint pdf</a>
0 commit comments