Meet our experts at #ISC24
Jan André Reuter is one of the software developers working on the Score-P instrumentation and monitoring infrastructure. His research is focused on performance analysis of OpenMP and accelerators in general.
💡app.swapcard.com/widget/event/i…

> OpenMP improvements in Visual Studio C/C++ compiler: triangular loop collapse dlvr.it/T6Gvr3

サムネがかわいいけどRust○そうとしててこわい…

Outperforming Rayon with OpenMP
belijzajac.dev/outperforming-…

msvc++ OpenMP Cuda の組み合わせが悪いかな？　並列になってくれない

My latest #DearNewcastle #Constituency Report is now up, check out what I've been up to & who I've been meeting as your MP - in #Newcastle , UK Parliament & online! #OpenMP Read here👉🏾

chionwurahmp.com/2024/05/dear-n…

OpenMP improvements in Visual Studio C/C++ compiler: triangular loop collapse
devblogs.microsoft.com/cppblog/openmp…
#cpp
#cplusplus

GENESIS、openmpが必要なのか

OpenMPを無視して実行できるとか言ってるけど、そしたら時間計測の関数読み込まれなくてコンパイルできなくね？できないの俺だけなんかこれ

'Porting HPC Applications to AMD Instinct MI300A Using Unified Memory and OpenMP' appearing in the technical program at ISC: arxiv.org/abs/2405.00436

🗞️𝘕𝐸𝘞𝑆!

1st #ESiWACE3 - #WarmWorld Summer School: HPC for Climate and Weather Applications💻

📆27 Aug-2 Sep
📍Helsinki

💠Modeling weather/climate in HPC & AI
💠ESM data manag., eval. & vis.
💠MPI, OpenMP, hybrid & GPU program
💠Novel concepts

➡️Info: bit.ly/3QzYNzz

OpenMPによる並列化を学んだ方がいい空気がある

Thiago Teixeira Philipp Rudiger Stephan Hoyer I think it might be this comment by Sam Gross here: github.com/pytorch/pytorc…
(i.e. threads are not inherited when forking, but the OpenMP initialized state is)

From a 1024-core supercomputer in 2008 to pushing LAMMPS on Graviton3E with 8192 cores on 1B atoms, it's been an incredible HPC journey. Key was finding the optimal MPI+OpenMP config: 2 OpenMP threads at >=512 cores for best scaling. More HPC tips here: buff.ly/4b8ISQU

=>
'Porting HPC Applications to AMD Instinct MI300A Using Unified Memory and OpenMP', AMD, arXiv, May 1, 2024 arxiv.org/abs/2405.00436
OpenFOAM
H100: 1
MI300A: 4

MI300 ISA, Dec 2023 (533 pp) twitter.com/ogawa_tter/sta…
CDNA 3 White Paper, Dec 3 (6)
ESPM2 2023 twitter.com/ogawa_tter/sta…

🚀 Ready to supercharge your HPC apps? Dive into the future with AMD Instinct MI300A's unified memory and OpenMP! 🧠✨ #HPC #AIRevolution #aizona

Diving into the future of tech with AMD Instinct MI300A! 🚀 Harness the power of unified memory and OpenMP for groundbreaking HPC applications. Let's innovate together! 💡 #aizona #AI #TechTrends

コンパイラ エラー C3016 | Microsoft Learn learn.microsoft.com/ja-jp/cpp/erro…
この制約は知らなかった
> OpenMP for ステートメントのインデックス変数は、符号付きの整数型である必要があります。

Suyash Tandon This talk is designed to serve as a guide to application developers on how to program, using portable directives such as OpenMP, to leverage the tight integration of CPU and GPUs on the same package in the same memory space (i.e., the APU)

Spent all night porting OpenMP and compiling to clang15-arm64 only to find out that my library's perf would be better off without specializing in paged vectors.

I was bound with the same performance on my x86; I must be terrible at writing concurrent directives

OpenMP並列化が組み込まれたコードに何も考えずに新しい記述を足してってたらついに計算がエラー終了した。絶対そのせいじゃん。OpenMPの勉強をしてこなかったツケが回ってきたやんけ〜