Create accountLogin

Recent searches

No recent searches

Search options

Only available when logged in.
social.tchncs.de is one of the many independent Mastodon servers you can use to participate in the fediverse.
A friendly server from Germany – which tends to attract techy people, but welcomes everybody. This is one of the oldest Mastodon instances.

Administered by:

Server stats:

3.9K
active users

social.tchncs.de: AboutProfiles directoryPrivacy policy

Mastodon: AboutGet the appKeyboard shortcutsView source codev4.3.6

#cuda

6 posts5 participants0 posts today
Public *
Alexandre Mutel

Ported salykova.github.io/sgemm-gpu to Vulkan (nice article!) 🐢

it's 2x slower than Cuda. That one was tricky to port, (e.g. need to alias shared buffer to allow LDS/STS.128), half of it with AI, 2nd half going over lines 1-by-1

Ported the Kernel 5 from seb-v.github.io/optimization/u to Vulkan 🚀

It's only 15% slower than Cuda (and it works on AMD)

In both cases, it's quite difficult to reason about SPIR-V ➡️ ISA, apart the AMD GPU Analyzer that is helping!

Digging deeper 🏗️

salykova blog · Beating cuBLAS in Single-Precision General Matrix MultiplicationThis blog post focuses on a GPU implementation of SGEMM (Single-precision GEneral Matrix Multiply) operation defined as C := alphaAB + beta*C. We’ll review the algorithm’s design and discuss optimization techniques such as inlined PTX, asynchronous memory copies, double-buffering, avoiding shared memory bank conflicts, and efficient coalesced storage through shared memory.
#vulkan#cuda
Public
Watson Tech World

One of the best interviews on AI and GPUs I've every seen was posted earlier today. Jensen Huang is really super smart in my opinion, and I think this interview is definitely worth watching. I can understand how it was a person like him who turned NVIDIA into a company with a market cap more than 1 trillion USD.

Jensen Huang on GPUs - Computerphile
youtube.com/watch?v=G6R7UOFx1bw

YouTubeJensen Huang on GPUs - ComputerphileBy Computerphile
#NVIDIA#GPU#AI
Public
Paul Houle

Just got my RSS reader YOShInOn building with uv and running under WSL2 with the Cuda libraries, despite a slight version mismatch... All I gotta do is switch it from arangodb (terrible license) to postgres, and it might have a future... With sentence_transformers running under WSL2 I might even be able to deduplicate the million images in my Fraxinus image sorter

screenshot with two windows; on left there is a summary of the output of a recommendation engine, on the right there is a list of python packages used by that same engine
#python#programming#ai
Replied in thread
Public
Giuseppe Bilotta

Even now, Thrust as a dependency is one of the main reason why we have a #CUDA backend, a #HIP / #ROCm backend and a pure #CPU backend in #GPUSPH, but not a #SYCL or #OneAPI backend (which would allow us to extend hardware support to #Intel GPUs). <doi.org/10.1002/cpe.8313>

This is also one of the reason why we implemented our own #BLAS routines when we introduced the semi-implicit integrator. A side-effect of this choice is that it allowed us to develop the improved #BiCGSTAB that I've had the opportunity to mention before <doi.org/10.1016/j.jcp.2022.111>. Sometimes I do wonder if it would be appropriate to “excorporate” it into its own library for general use, since it's something that would benefit others. OTOH, this one was developed specifically for GPUSPH and it's tightly integrated with the rest of it (including its support for multi-GPU), and refactoring to turn it into a library like cuBLAS is

a. too much effort
b. probably not worth it.

Again, following @eniko's original thread, it's really not that hard to roll your own, and probably less time consuming than trying to wrangle your way through an API that may or may not fit your needs.

6/

Public
N-gated Hacker News

AMD YOLO: because why not base your entire #business #strategy on a meme? 🚀🎉 Thanks to AMD's cultural enlightenment, they're now #shipping #boxes faster than philosophical musings on singularity! 🤯 Who knew rewriting a stack could be as easy as beating #NVIDIA at their own game? Just don't tell CUDA—it might get jealous! 😜
geohot.github.io//blog/jekyll/ #AMD #YOLO #meme #CUDA #competition #HackerNews #ngated

the singularity is nearer · AMD YOLOAMD is sending us the two MI300X boxes we asked for. They are in the mail.
Public *
Dr. Moritz Lehmann

Hot Aisle's 8x AMD #MI300X server is the fastest computer I've ever tested in #FluidX3D #CFD, achieving a peak #LBM performance of 205 GLUPs/s, and a combined VRAM bandwidth of 23 TB/s. 🖖🤯
The #RTX 5090 looks like a toy in comparison.

MI300X beats even Nvidia's GH200 94GB. This marks a very fascinating inflection point in #GPGPU: #CUDA is not the performance leader anymore. 🖖😛
You need a cross-vendor language like #OpenCL to leverage its power.

FluidX3D on #GitHub: github.com/ProjectPhysX/FluidX

FluidX3D benchmarks: the 8x AMD MI300X system leaves every other benchmarked computer behind in the dust.
how 8x MI300X show up in OpenCL
ExploreLive feeds
About