PPubMed20 Apr 2017Cited 48×

The Dynamo package for tomography and subtomogram averaging: components for MATLAB , GPU computing and EC2 Amazon Web Services Dynamo package for tomography and subtomogram averaging

Daniel Castaño-Díez

Abstract

A technical description of the Dynamo software package for subtomogram averaging is provided. Details are given on advanced MATLAB libraries, parallelization strategies, the use of GPUs and accessibility though the Amazon cloud computing services.

§ The Valyu brief

The paper, in five minutes

Read from the full text via Valyu Contents

Dynamo provides a MATLAB‑based, GPU‑accelerated, cloud‑ready software suite for cryo‑electron tomography subtomogram averaging.

What’s new

i.Advanced MATLAB libraries with MEX integration for high‑performance computation.
ii.GPU implementations (compact CUDA and modular gpuArray) accelerating alignment up to ~236× speedup.
iii.Cloud deployment via a pre‑configured AWS EC2 AMI enabling pay‑as‑you‑go GPU resources.
iv.Data‑management and annotation tools through the Catalogue module.
v.Multi‑level parallelization (fine‑grain GPU, coarse‑grain MPI, multicore) supporting single‑node and cluster workflows.

How they did it

The Dynamo package is built primarily in MATLAB, with performance‑critical sections compiled as C++/CUDA MEX files. Projects encapsulate input data, numerical settings, and computation resources, allowing execution on CPUs, GPUs, or AWS EC2 instances via command‑line or GUI interfaces.

What they found

GPU speedup up to 236× (K80, 256³ cube) compared with CPU.
On‑screen refresh of 1000×1000 matrices at 36 fps on a 2.9 GHz MacBook.
Cloud instance p2.16xlarge provides 8 K80 cards (16 GPUs) with 32 CPUs for ~14 US$ per hour.
Cost per GPU hour ≈0.88 US$, yielding 15–40× speedup and 4–10× cheaper compute time than CPU‑only instances.

Caveats

✕Full modular GPU version requires a compatible CUDA version and MATLAB Parallel Computing Toolbox.
✕Standalone version incurs initialization delays (up to 1 min) and limited workspace sharing.
✕Cluster MPI parallelization depends on atomic mkdir or external lockfile, which may need manual installation.
✕Performance gains vary with algorithm; non‑compute‑intensive tasks see limited GPU benefit.

§ The Valyu brief

Reading the full paper and taking notes. This takes a few seconds…

§ Ask this paper

Ask a question about this paper

Valyu reads the full text and answers from what the paper actually says.

§ Related work in other archives

The <i> Dynamo </i> package for tomography and subtomogram averaging: components for <i> MATLAB </i> , GPU computing and EC2 Amazon Web Services <i> Dynamo </i> package for tomography and subtomogram averaging

The paper, in five minutes

Ask a question about this paper

Similar papers in all four archives

Similar papers in all four archives

The paper, in five minutes