Create accountLogin

Recent searches

No recent searches

Search options

Only available when logged in.
techhub.social is one of the many independent Mastodon servers you can use to participate in the fediverse.
A hub primarily for passionate technologists, but everyone is welcome

Administered by:

Server stats:

4.8K
active users

techhub.social: AboutStatusProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.3.8

#cuda

4 posts3 participants2 posts today
Public
GripNews

🌕 史丹佛CRFM：驚人的AI生成核心程式碼
➤ AI 核心程式碼生成技術的突破性進展
✤ crfm.stanford.edu/2025/05/28/f
史丹佛大學CRFM研究團隊意外發現，利用AI生成的純CUDA-C核心程式碼，在多個機器學習運算 (如矩陣乘法、卷積、Softmax、LayerNorm) 上的效能，甚至超越了PyTorch預先優化的核心程式碼。此成果源於簡化的合成資料生成方法，展現了AI在效能優化方面的潛力，並為未來核心程式碼生成的研究奠定了基礎。研究人員分享了方法、五個優化核心程式碼、優化路徑範例，並探討了此技術的影響。
+ 哇！AI竟然能生成比專業工程師優化的程式碼還好，這對機器學習的發展來說是個重大的進展！
+ 這個研究很有趣，證明瞭AI不僅能自動化許多任務，還能真正提升效能，這讓人對未來的發展充滿期待。
#人工智慧 #機器學習 #CUDA #程式碼生成 #效能優化

crfm.stanford.eduStanford CRFM
Public
.:\dGh/:.

That's it, I'm going against AMD for recommending computers for #AI.

I don't even know how to start running something on their NPU via Linux, or check it's running at all. Windows fares better but it's `llama.cpp` doesn't work there.

So, if you want to run AI on your computer: RTX, Mac, or don't bother at all.

#ML#LLM#LargeLanguageModels
Public
Mille e Una Avventura

Mettiamo a confronto varie tecniche di programmazione per scoprire qual è la più performante. Il classico algoritmo che approssima Pi greco con il metodo di Montecarlo viene implementato con un singolo thread, con le SIMD AVX2, multithreaded sulla CPU e, poi, su una GPU Nvidia sfuttando CUDA. Quali performance otterremo? #clanguage #cuda #multithreading #mandelbrot #cplusplus #visualstudio #programming youtube.com/watch?v=NyaGaUeDck

YouTubeCalcolo Parallelo in C++ / ProgrammazioneBy Mille e Una Avventura
Public
GripNews

🌗 RightNow AI：加速 CUDA 程式碼的最佳方案
➤ 人工智慧驅動的 CUDA 核心程式碼優化
✤ rightnowai.co/
RightNow AI 是一個利用人工智慧技術，自動分析、偵測瓶頸並最佳化 CUDA 核心程式碼的平臺。它支援所有主要的 NVIDIA 架構，提供伺服器端 GPU 分析、簡化提示生成優化 CUDA 核心程式碼等功能，能將效能提升 2-20 倍。平臺提供免費、Pro 和 Enterprise 三種方案，滿足不同規模的團隊需求，並提供常見問題解答與聯絡方式。
+ 「我一直被一個緩慢的 CUDA 核心程式碼困擾了很久，RightNow AI 在幾分鐘內就完成了最佳化，速度提升了三倍！這簡直太棒了！」
+ 「過去嘗試過三種不同的優化工具，但 RightNow AI 纔是真正能帶來改變的利器。我們的推論程式碼速度提升了 20 倍！」
#人工智慧 #CUDA #GPU 最佳化 #效能提升

www.rightnowai.coRightNow AIAutomatically profile, detect bottlenecks, and optimize your CUDA kernels for peak performance.
Public
Christine Sætre-esque

China’s few remaining weaknesses:

Manufacturing the most advanced chips is a still an issue because ”China lacks a domestic alternative to the cutting-edge lithography tools produced by ASML, a Dutch company. … chip designers [are] reliant on SMIC, a state-owned foundry. “

China may initially struggle awhile with “… the software used by coders to program chips. Nvidia’s platform, called CUDA, is still by far the best in the world. Nearly all AI developers learn how to use it. And it works only with Nvidia’s chips. Switching to an alternative is costly, because it pulls developers out of an enormous network of fellow users that can help solve problems.”

“Huawei has created a substitute for CUDA, called CANN …But the software is years behind Nvidia’s [It’s buggy and unenthusiastic used by local techies. But,] Huawei has beaten the odds before, and it may well do the same again with CANN.”


economist.com/business/2025/05

An employee packages semiconductor power devices for export at a semiconductor manufacturing enterprise.
The Economist · Huawei and other Chinese chip firms are catching up fastBy The Economist
Public
Chris Vitalos

Like many #Linux users, I use that #OS to extend old #PC #hardware lifecycle. My #Dell #Optiplex 9020 includes a Pascal based #GPU.

My issue is I cannot install a more modern GPU because of physical constraints with the chassis itself.

>Maxwell, Pascal, and Volta architectures are now feature-complete with no further enhancements planned. ... Users should plan migration .., as future toolkits will be unable to target [these] GPUs. 1/

docs.nvidia.com/cuda/cuda-tool

docs.nvidia.com1. CUDA 12.9 Release Notes — Release Notes 12.9 documentationThe Release Notes for the CUDA Toolkit.
#Nvidia#CUDA
Public
Hacker News 50

Faster sorting with SIMD CUDA intrinsics (2024)

Link: winwang.blog/posts/bitonic-sor
Discussion: news.ycombinator.com/item?id=4

winwang.blog · Faster sorting with SIMD CUDA intrinsicsFull code on Github: https://github.com/wiwa/blog-code/ Hi Link to heading Recently, I finished a batch at the Recurse Center… is what I would have said if this post were written when I intended to write it (i.e. 3 months ago). My project there focused on a questionable application of CUDA (mostly irrelevant to this post), but it got me thinking more about other GPU-friendly algorithms. Instead of my Recurse project (which I hope to write about in a later post), I want to simply begin writing about technical stuff I’ve played around with.
#cuda
Public
N-gated Hacker News

🎉 Oh joy, another thrilling journey into the depths of #CUDA #intrinsics, as our brave author heroically tackles the burning issue of sorting faster! 🚀 Because who wouldn't want a detailed dissertation on an #algorithm called "bitonic sort" that promises to revolutionize... absolutely nothing in your daily life. 😅
winwang.blog/posts/bitonic-sor #BitonicSort #TechJourney #SortingAlgorithms #HackerNews #ngated

winwang.blog · Faster sorting with SIMD CUDA intrinsicsFull code on Github: https://github.com/wiwa/blog-code/ Hi Link to heading Recently, I finished a batch at the Recurse Center… is what I would have said if this post were written when I intended to write it (i.e. 3 months ago). My project there focused on a questionable application of CUDA (mostly irrelevant to this post), but it got me thinking more about other GPU-friendly algorithms. Instead of my Recurse project (which I hope to write about in a later post), I want to simply begin writing about technical stuff I’ve played around with.
ExploreLive feeds
About