Add November talk

Author: diehlpk

index.html

@@ -56,6 +56,11 @@ <h2>Overview</h2>
       <td><a href="#fy273">From Anaconda to Pixi: Modernizing Python Package Management for Neutron Science at ORNL</a></td>
       <td>11/12/2025</td>
   </tr>
+  <tr>
+    <td>Abdel-Hameed (Hameed) A. Badawy</td>
+    <td><a href="#fy274">PPT-GPU 2.0: An Accurate, Scalable, Hybrid Performance Model for Modern GPGPUs</a></td>
+    <td>10/29/2025</td>
+  </tr>
   <tr>
       <td>Rene Gassmoeller</td>
       <td><a href="#fy272">What We Need to Model Planetary Interiors: The Role of Research Software in Computational Geodynamics</a></td>
@@ -200,6 +205,24 @@ <h3>From Anaconda to Pixi: Modernizing Python Package Management for Neutron Sci
 At ORNL, Dr. Zhang contributes to software infrastructure development for neutron scattering facilities at the Spallation Neutron Source (SNS) and High Flux Isotope Reactor (HFIR), with a focus on neutron imaging applications. His work includes developing BM3DORNL, an open-source library for artifact removal in neutron tomography; creating FuGNN (Fusion Graph Neural Network) for applications in neutron reflectometry and power-grid analysis; leading the refactoring effort for PLEIADES v2.0, a neutron resonance data reduction software originally developed at Los Alamos National Laboratory; and contributing to iBeatles, the data reduction engine for neutron Bragg edge strain mapping at ORNL's VENUS beamline, where he addressed packaging challenges and implemented command-line batch processing capabilities. He is an active contributor to open-source scientific computing projects and serves as a reviewer for several journals in computational materials science and neutron scattering.
 </p>
 
+<div id="fy274"></div>
+<h3>PPT-GPU 2.0: An Accurate, Scalable, Hybrid Performance Model for Modern GPGPUs</h3>
+Speaker: Abdel-Hameed (Hameed) A. Badawy<br>
+New Mexico State University<br><br>
+
+Abstract:
+<p>
+As modern GPU architectures advance to meet the demands of large-scale simulations and AI, accurate and scalable modeling tools are indispensable for performance optimization and design space exploration. This talk introduces PPT-GPU 2.0, a hybrid performance prediction toolkit designed to model NVIDIA GPUs.
+We will detail our enhancements to PPT-GPU, including advanced modeling for the Ampere architecture's redesigned shared memory, local memory, and tensor cores. Additionally, we introduce a new “balance model” to achieve more precise occupancy modeling, complementing metrics such as kernel execution time and active cycles. Furthermore, we integrated a roofline analysis framework directly within PPT-GPU 2.0. This visualization tool enables rapid identification of computational and memory bottlenecks, adding versatility to PPT-GPU 2.0 for scalable hardware-software co-design.
+Our evaluations on A100 GPUs are validated against NVIDIA Nsight Compute (Ground Truth) and AccelSim (a cycle-accurate simulator), showing average errors of 16.4% for active cycles and 19.2% for kernel execution time.
+Towards the end, time permitting, we will give a glimpse into our hardware security work, which includes using various ML techniques to insert and detect hardware trojans in digital systems.
+</p>
+
+Bio:
+<p>
+Prof. Abdel-Hameed (Hameed) A. Badawy (Senior Member, IEEE) received a B.Sc. degree (Hons.) in Electronics Engineering from Mansoura University, Egypt, focusing on computers and control systems, and an M.Sc. & Ph.D. degrees in computer engineering from the University of Maryland, College Park. He is a tenured Associate Professor with the Klipsch School of Electrical and Computer Engineering, New Mexico State University, Las Cruces, NM, USA. He is the chair of the Computer Engineering Area. Also, he is an affiliated faculty with the Los Alamos and Lawrence Berkeley National Laboratories. He has been a Visiting Research Scientist with the New Mexico Consortium. He was a Lead Research Scientist with the High-Performance Computing Laboratory at the George Washington University. His research interests include performance modeling, prediction, monitoring, and evaluation; high-performance computer architectures; hardware security; machine learning applications to computational problems; and Quantum Computing.
+</p>
+
 <div id="fy272"></div>
 <h3>What We Need to Model Planetary Interiors: The Role of Research Software in Computational Geodynamics</h3>
 Speaker: Rene Gassmoeller<br>