BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:[Online] Scaling CUDA-Accelerated Applications
DTSTART:20260907T070000Z
DTEND:20260909T130000Z
DTSTAMP:20260617T035800Z
UID:indico-event-214@indico.ecap.work
DESCRIPTION:Scaling CUDA-Accelerated Applications\n\nSchedule & Format\n\n
 Date: 2026\, September 7-9\nTimes:\n\nSep 7: 9:00 - 15:00 CE(S)T\nSep 8: 9
 :00 - 15:00 CE(S)T\nSep 9: 9:00 - 15:00 CE(S)T\n\n\nFormat: Three-day\nLoc
 ation: Online via Zoom\nLanguage: English\n\nRegistered participants will 
 receive the video conferencing link via email on the day before the course
 .\nFrom Zero to Multi-Node GPU Programming\nThis event is part of the From
  Zero to Multi-Node GPU Programming series. Registration is done individua
 lly for each part of the series.\n\nPart 1 - Introduction to CUDA C/C++ (2
 026\, September 3-4) (Register)\nPart 2 - Scaling CUDA-Accelerated Applica
 tions (this course) (2026\, September 7-9) (Register)\n\nInstructors\n\nDr
 . Sebastian Kuckuk\, NHR@FAU\, certified NVIDIA DLI Ambassador\nAditya Uje
 niya\, NHR@FAU\nMarkus Velten\, NHR@TUD\, certified NVIDIA DLI Ambassador\
 n\nThis course is organized by Erlangen National High Performance Computin
 g Center (NHR@FAU) in collaboration with NHR@TUD.\nCourse Description\nSca
 ling a GPU application beyond a single accelerator requires both intra-nod
 e and inter-node parallelism. This course provides a comprehensive treatme
 nt of both: part one covers CUDA streams\, multi-GPU execution within a no
 de\, and direct peer-to-peer GPU memory access\; part two extends that fou
 ndation to multi-node deployments using CUDA-aware MPI and NVSHMEM\, inclu
 ding domain decomposition and halo exchange patterns. The course uses a pr
 ogression from CPU baseline through managed memory and algorithmic partiti
 oning to full distributed execution.\nThis course was developed to replace
  the two formerly separate NVIDIA DLI courses Accelerating CUDA C++ Applic
 ations with Multiple GPUs and Scaling CUDA C++ Applications to Multiple No
 des which have been first on hold and then finally discontinued in 2025 an
 d 2026.\nPrerequisites\nKnowledge\n\nExperience with CUDA C++ GPU programm
 ing\, including memory allocation\, kernel launches\, grid-stride loops\, 
 and error handling (equivalent to the Introduction to CUDA C/C++ course)\n
 Familiarity with the Linux command line as well as compiling and running C
 UDA applications\n\nTechnical\n\nA modern web browser (for JupyterHub acce
 ss to NHR@FAU's HPC clusters)\nA local installation of NVIDIA Nsight Syste
 ms\n\nCourse Structure\n\nCPU baseline and GPU porting: managed memory\, a
 lgorithmic work partitioning\nCUDA streams and copy/compute overlap: concu
 rrent execution and Nsight Systems profiling\nMulti-GPU programming: devic
 e management\, workload indexing\, and peer-to-peer communication\nMulti-n
 ode parallelism: MPI fundamentals\, CUDA-aware MPI\, and halo exchanges\nN
 VSHMEM: symmetric memory model\, GPU-initiated transfers\, and distributed
  solvers\n\nLearning Outcomes\nAfter completing this course\, you will be 
 able to:\n\nUse concurrent CUDA streams to overlap memory transfers with G
 PU computation\nScale CUDA C++ workloads across multiple GPUs within a sin
 gle compute node\nEnable and exploit direct peer-to-peer GPU memory access
  for efficient intra-node communication\nWrite portable\, scalable SPMD co
 de using CUDA-aware MPI with inter-node GPU communication\nApply NVSHMEM f
 or GPU-initiated data transfers using the symmetric memory model\nImplemen
 t domain decomposition and halo exchange patterns for distributed GPU work
 loads\nProfile multi-GPU execution and identify performance bottlenecks wi
 th NVIDIA Nsight Systems\n\nRegistration\, Wait List and Withdrawal Policy
 \nRegistration\nPlease register at the bottom of this page. Registration i
 s open until a few days before the course starts\, or until the course is 
 fully booked.\nPrices and Eligibility\nThis course is open and free of cha
 rge for participants affiliated with academic institutions in European Uni
 on (EU) member states and Horizon 2020-associated countries.\nWait List\nI
 f the course reaches its maximum capacity\, you can request to join the wa
 it list by sending an email to nhr-training@fau.de. Please include your na
 me and university affiliation in the message.\nWithdrawal Policy\nPlease o
 nly register if you are committed to attending the course. No-shows will b
 e blacklisted and excluded from future events.\nIf you need to withdraw yo
 ur registration\, please either cancel it directly through the registratio
 n system or send an email to nhr-training@fau.de.\nAdditional Courses\nYou
  can find an up-to-date list of all courses offered by NHR@FAU at https://
 hpc.fau.de/teaching/tutorials-and-courses/.\n\nhttps://indico.ecap.work/ev
 ent/214/
LOCATION:Online
URL:https://indico.ecap.work/event/214/
END:VEVENT
END:VCALENDAR