*********************************
There is now a CONTENT FREEZE for Mercury while we switch to a new platform. It began on Friday, March 10 at 6pm and will end on Wednesday, March 15 at noon. No new content can be created during this time, but all material in the system as of the beginning of the freeze will be migrated to the new platform, including users and groups. Functionally the new site is identical to the old one. webteam@gatech.edu
*********************************
Kiran Ravikumar
Advisor: Prof. P. K. Yeung
will defend a doctoral thesis entitled,
Extreme-scale computing and studies of intermittency, mixing of passive scalars and stratified flows in turbulence
On
Friday, July 30 at 1:30 p.m.
https://bluejeans.com/321266573/8858
A new batched asynchronous algorithm capable of extremely large problem sizes has been developed for dense node heterogeneous architecture machines like Summit. Optimizing data copies between CPU and GPU and communication over the network while overlapping data copies and computations are key to achieving good performance. Processing data residing on the larger CPU memory in batches on the GPU helps avoids limitations on problem size. Favorable performance is obtained up to a world-leading problem size of 18432^3 (over 6 trillion grid points) on 3072 Summit nodes. A more portable implementation using OpenMP is pursued to target 32768^3 problem size on the exascale machine Frontier expected in early 2022.
Hero-sized simulations are often relatively short in time, which raises concerns regarding sampling and statistical independence. A Multiple Resolution Independent Simulations approach (MRIS) is developed to address this issue, via multiple short simulation segments evolving from lower-resolution datasets distributed over a longer physical time span. Using this approach, the effects of small-scale intermittency are studied through statistics of local averages of dissipation rate and enstrophy. The dissipation rate is further studied from a multifractal viewpoint. The MRIS approach is also used to study passive scalar intermittency and test for refined similarity hypothesis, through statistics of scalar dissipation rate at high Reynolds number. Lastly, density stratified flows are studied under both stable and unstable stratification, with anisotropy development studied through the Reynolds-stress budget.
Committee
