HPCE IIT Madras in collaboration with NVIDIA is organizing a hackathon for students to help accelerate their applications on GPUs. Please take advantage of this event to improve performance of your research applications. Note that this is not a competition. NVIDIA experts would help improve speedup of your applications. The event is planned on January 24 + 25.

Registration
Please register for the hackathon. More details would be sent to the registered participants.

Program

The event is scheduled from 09:00--17:00 at Aryabhatta Hall (CSB 25), CSE Department, IIT Madras.

Day 1 (January 24)
  + Opening ceremony
  + Half-day presentations and training from NVIDIA
  + Half-day hands-on session from NVIDIA

Day 2 (January 25)
  + Application optimization by IITM Teams
  + Evaluation
  + Closing ceremony

Tea breaks would be at 11:00 and 15:30 on both the days.

Detailed Agenda

Topic	Description
Introduction to Heterogeneous Computing	Why hybrid Architecture? Heterogeneous Programming Models Moore's law, Dennard Scaling, Amdahl's law
CPU vs GPU Architecture	Gives fundamental difference between the throughput based vs latency based architecture
Ways to GPGPU Programming	Libraries vs Directive based vs Languages
Hands-on Directive based Programming	Introduction to OpenACC Directive based Programming – Hands-on
CUDA Languages	Introduction to CUDA C API Introduction to CUDA Memory Hierarchy Hands-On CUDA C Programming
Mini Application	Hands-On session on real world Mini-Application in domain of Weather/CFD
Matlab on GPU	GPU support on Matlab and how to write CUDA code on Matlab
IITM Application porting	Discuss strategy for porting in-house application Porting under guidance of mentor

Acknowledgment
HPCE Team thanks CSE Department for supporting us conduct the event.

Contact
rupesh@iitm.ac.in

NVIDIA Organizers
Prakash Tubakad, Bharat Kumar

IITM Organizers
Anand, Prasanna, Saravana, Gayathri, Rupesh
HPCE, IIT Madras

NVIDIA-IITM Hackathon