Asynchronous Many-Task Algorithm Improves SN Transport Solver Performance on Shared Memory Systems
This paper presents a new Asynchronous Many-Task (AMT) algorithm for solving discrete ordinates SN transport problems on unstructured meshes in shared memory systems. It reviews performance bottlenecks in existing transport solvers on modern many-core architectures with high core counts, surveys solver performance across various compute hardware, and introduces the AMT algorithm with results showing improved parallelization efficiency for neutral particle transport simulations.
Key quotes
Discrete ordinates SN transport solvers on unstructured meshes pose a challenge to scale due to complex data dependencies, memory access patterns and a high-dimensional domain.
In this paper, we review the performance bottlenecks within the shared memory parallelization scheme of an existing transport solver on modern many-core architectures with high core counts.
We then present a new Asynchronous Many-Task (AMT) algorithm for shared memory parallelism, present results showing an in...
From the article
You might also wanna read
Building a Multiplayer Particle Simulation in Go: Testing Performance Limits with Millions of Particles
The article details a developer's journey to learn Go programming by creating a particle simulation that can handle millions of particles, b
ACM Transactions on Parallel Computing (topc) Vol 13, Issue 2
FlashAttention-4: Algorithm and Kernel Pipelining Co-Design for Asymmetric Hardware Scaling
Optimizing Container Image Distribution to Eliminate Cold Starts in AI Applications
The article discusses the persistent problem of container cold starts in latency-sensitive AI applications and proposes solutions for faster
cerebrium.ai·4mo ago
Optimizing the asin() Function: A Technical Follow-up on Performance Improvements
The author revisits their previous work on optimizing the asin() (arcsine) function in C/C++ after receiving feedback from online communitie
16bpp.net·3mo ago
Memory-Efficient Two-Pass Lanczos Algorithm Implementation in Rust
This technical article presents a memory-efficient implementation of the Lanczos algorithm in Rust, addressing the significant memory requir
lukefleed.xyz·7mo ago
Comments
Sign in to join the conversation.
No comments yet. Be the first.