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:20260522T150111Z
LOCATION:C2/3/4 Ballroom
DTSTART;TZID=America/Chicago:20181115T083000
DTEND;TZID=America/Chicago:20181115T170000
UID:submissions.supercomputing.org_SC18_sess327_spost111@linklings.com
SUMMARY:Accelerating 2D FFT: Exploit GPU Tensor Cores through Mixed-Precis
 ion
DESCRIPTION:Xiaohe Cheng (Hong Kong University of Science and Technology) 
 and Anumeena Sorna (National Institute of Technology, Tiruchirappalli)\n\n
 The two-dimensional Fourier Transform is a widely-used computational kerne
 l in many HPC applications. The popular NVIDIA cuFFT library provides a si
 mple interface to compute 2D FFT on GPUs, but it's yet to utilize the rece
 nt hardware advancement in half-precision floating-point arithmetic. In th
 is poster, we propose a mixed-precision method to accelerate 2D FFT by exp
 loiting the FP16 matrix-multiply-and-accumulate units on the newest GPU ar
 chitecture, known as tensor cores. We achieve a balance between speed and 
 accuracy by dynamically splitting the single-precision input data into two
  half-precision operands and performing FFT separately. We present a CUDA-
 based implementation that achieves 3-digit more accuracy than half-precisi
 on cuFFT. We also demonstrate the stability and scalability of our approac
 h and conclude that it attains high accuracy with tolerable splitting over
 head.\n\nRegistration Category: Tech Program Reg Pass, Exhibits Reg Pass\n
 \n
END:VEVENT
END:VCALENDAR