Create accountLogin

Recent searches

No recent searches

Search options

Only available when logged in.
hachyderm.io is one of the many independent Mastodon servers you can use to participate in the fediverse.
Hachyderm is a safe space, LGBTQIA+ and BLM, primarily comprised of tech industry professionals world wide. Note that many non-user account types have restrictions - please see our About page.

Administered by:

Server stats:

9.4K
active users

hachyderm.io: AboutStatusProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.3.7

#libCEED

0 posts0 participants0 posts today
Public
Jed Brown

When does a fluids mini-app cease to be "mini"? What separates Colorful Fluid Dynamics (CFD; h/t @kpdooty) from a tool for science?

I'd argue it's when you collect turbulent statistics and get careful about statistical convergence. Our fluids mini-app was created to drive and library integration, but has now taken this step as it transforms into an instrument to explain how RANS models fail (and what they need to succeed). With a 10x strong scaling boost so far.

#HPC
Public
Jed Brown

0.11 was released last week, providing performance-portable element algebra designed for integration in your code. These data structures move your sparse linear algebra from the dotted line to the solid line, giving a 2x performance boost even for lowest order elements (solid and fluid mechanics). This release has memory checking, backend and H(div) improvements, and various utilities, plus lots of new features in the implicit compressible viscous flow solver.
github.com/CEED/libCEED/releas

Performance plot showing efficiency to apply the Jacobian of a nonlinear elasticity operator: billions of degrees of freedom/second/GPU versus total number of degrees of freedom. The dotted line represents cuSPARSE data structures and the solid lines are libCEED "shell" data structures. The solid lines are always higher (more efficient) and larger problem sizes can be solved before running out of memory. Efficiency is 2x higher for first order elements and increases with polynomial order, while the dotted (cuSPARSE) performance decreases rapidly with polynomial order.
libCEED operator decomposition representing parallel restriction, element restriction, basis application, and qfunctions applied at quadrature points. The decomposition is logical, but fused by backends to achieve high performance from vanilla code written by the user.
ExploreLive feeds
About