BEGIN:VCALENDAR
VERSION:2.0
PRODID:Linklings LLC
BEGIN:VTIMEZONE
TZID:America/Chicago
X-LIC-LOCATION:America/Chicago
BEGIN:DAYLIGHT
TZOFFSETFROM:-0600
TZOFFSETTO:-0500
TZNAME:CDT
DTSTART:19700308T020000
RRULE:FREQ=YEARLY;BYMONTH=3;BYDAY=2SU
END:DAYLIGHT
BEGIN:STANDARD
TZOFFSETFROM:-0500
TZOFFSETTO:-0600
TZNAME:CST
DTSTART:19701101T020000
RRULE:FREQ=YEARLY;BYMONTH=11;BYDAY=1SU
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
DTSTAMP:20260202T201803Z
LOCATION:266
DTSTART;TZID=America/Chicago:20251117T103000
DTEND;TZID=America/Chicago:20251117T110000
UID:submissions.supercomputing.org_SC25_sess218_ws_waccpd106@linklings.com
SUMMARY:Mojo: MLIR-based Performance-Portable HPC Science Kernels on GPUs 
 for the Python Ecosystem
DESCRIPTION:William Godoy (Oak Ridge National Laboratory (ORNL)); Tatiana 
 Melnichenko (University of Tennessee, Knoxville; Oak Ridge National Labora
 tory (ORNL)); and Pedro Valero-Lara, Wael Elwasif, Philip Fackler, Rafael 
 Ferreira Da Silva, Keita Teranishi, and Jeffrey Vetter (Oak Ridge National
  Laboratory (ORNL))\n\nWe explore the performance and portability of the n
 ovel Mojo language for scientific computing workloads on GPUs. As the firs
 t language based on the LLVM's Multi-Level Intermediate Representation (ML
 IR) compiler infrastructure, Mojo aims to close performance and productivi
 ty gaps by combining Python's interoperability and syntax with CUDA-like c
 ompile-time programming. We target four scientific workloads: (i) a Seven-
 point stencil (memory-bound), (ii) BabelStream (memory-bound), (iii) miniB
 UDE (compute-bound), and (iv) Hartree--Fock (compute-bound with atomic ope
 rations), and compared their performance against vendor baselines on NVIDI
 A H100 and AMD MI300A GPUs. We show that Mojo's performance is competitive
  with CUDA and HIP for memory-bound kernels, whereas gaps exist on AMD GPU
 s for atomic operations and for fast-math compute-bound kernels on both AM
 D and NVIDIA GPUs. Although the learning curve and programming requirement
 s are still fairly low-level, Mojo can close significant gaps in the fragm
 ented Python ecosystem in the convergence of scientific computing and AI.\
 n\nRecording: Livestreamed, Recorded\n\nRegistration Category: Technical P
 rogram Reg Pass, Workshop Reg Pass\n\nSession Chairs: Andreas Herten (Fors
 chungszentrum Jülich, Jülich Supercomputing Centre (JSC)); Rabab Alomairy 
 (Massachusetts Institute of Technology (MIT), King Abdullah University of 
 Science and Technology (KAUST)); and Jorge Luis Galvez Vallejo (Australian
  National University)\n\n
END:VEVENT
END:VCALENDAR