Head of the Lab
Asst. Prof. Didem Unat
Affliate - Lawrence Berkeley National Laboratory, CA, USA
Ph.D University of California, San Diego, USA
Contact: dunat AT ku.edu.tr
Muhammad Nufail Farooqi
Contact: mfarooqi14 AT ku.edu.tr
MS : National University of Science and Technology, Pakistan
My research focuses on design and development of a programming model and runtime system for Adaptive Mesh Refinement (AMR) applications executing on emerging parallel architectures in the exascale era. The goal of the programming model is to enable legacy AMR applications to hide communication cost behind computation with modest amount of programming effort and exploit asynchrony as much as possible.
Hassan Salehe Matar
Contact: hmatar AT ku.edu.tr
Bachelor: Middle East Technical University, Turkey
My research focuses on designing and improving algorithms and tools for detecting concurrency errors in parallel programs. I previously worked on improving performance of race detection algorithms using Intel(r)-TSX and interned at Intel in Munich for 1.5 years.
Contact: nahmad16 AT ku.edu.tr
MS: Beihang University, Beijing, China
My research focuses on the parallelization and optimizations of unstructured meshes on emerging multicore architectures. I particularly deal with data locality issues for sparse data structures on the chip and work on reservoir simulations as a motivating group of applications.
Muhammad Aditya Sasongko
Contact: msasongko17 AT ku.edu.tr
MS: Marmara University, Istanbul, Turkey, BS: Middle East Technical University, Ankara, Turkey
My research focuses on tool development for shared memory multicore architectures. Recently I am working on a tool that automatically generates communication matrices for threads.
Pirah Noor Soomro
Contact: psoomro16 AT ku.edu.tr
Bachelor: Mehran University of Engineering and Technology, Pakistan
I am particularly interested in reducing communication overhead and data access latency in shared memory programming models. I am working on a thread binding tool, in which I aim to automatically bind threads of a parallel application based on application behavior and underlying hardware topology.
Contact: bbastem AT ku.edu.tr
Bachelor: Koç University, Turkey
I am working on design and implementation of tiling based multi-process multi-GPU execution using TiDA. TiDA is a tiling data abstraction which enhances data locality and manages parallelism on NUMA nodes. I'm currently working on extending current API to execute applications on heterogenous systems with GPUs.
Contact: mlaghari16 AT ku.edu.tr
Bachelor: Lahore University of Management Science (LUMS), Pakistan
My primary research focuses on object placement on heterogeneous memory systems, particularly I study the Intel Knights Landing architecture. I am developing algorithms and necessary tools to guide the programmer to place objects on different types of memories to improve performance.
Contact: ddikbayir17 AT ku.edu.tr
Bachelor: Ankara I.D Bilkent University, Turkey
I am currently working on parallelisation and optimization of machine learning algorithms on GPU. I particularly focus on the KNET Deep Learning Framework developed at Koc University. My goal is to reduce the communication overhead and memory accesses of machine learning core kernels.
Bachelor: Koç University, Turkey
First Position: KTH Royal Institute of Technology - MSc Embedded Systems, Sweden
Metin Muradoğlu (Koç University, Mechanical Engineering)
John Shalf (Lawrence Berkeley National Laboratory, USA)
Ann Almgren (Lawrence Berkeley National Laboratory, USA)
Xing Cai (Simula Research Laboratory, Norway)
Deniz Yüret (Koç University)
Anshu Dubey (Argonne National Laboratory)
Milind Chabbi (Baidu Research, USA)