mshachkov
Re:Intel benchmarks D270 vs. RTX PRO 6000: 1.12x throughput on Qwen3-3...
on Monday November 16, 2026 @02:51PM (#50522191)
> How does the D270 participate in larger multi-card operations?
Also interested in this, unfortunately the reported benchmark does not evaluate tp > 1 use-case.
Re:DeepSeek V4 Flash 0731
on Friday August 07, 2026 @08:10PM (#49217659)
This one works great. https://github.com/ormandj/sglang-deepseek-v4-flash-sm120
Re:Visualizing 6D Mesh Parallelism
on Thursday December 19, 2024 @07:19PM (#42467141)
Could anyone suggest a source about incorporation of "caching" into job-shop-like scheduling problem formulation?
In other words, i would like to explore a way to approach a problem of statically scheduling (heterogenous) computations and data transfers between multuple tiers of caches (disk, ram, vram) in OR fashion (e.g. formulating MILP).
Re:Scalable and Performant Data Loading
on Sunday December 01, 2024 @08:10PM (#42292116)
Could anyone suggest a source about incorporation of "caching" into job-shop-like scheduling problem formulation?
In other words, i would like to explore a way to approach a problem of statically scheduling (heterogenous) computations and data transfers between multuple tiers of caches (disk, ram, vram) in OR fashion (e.g. formulating MILP).
Cut Your Losses in Large-Vocabulary Language Models
OpenMP 6.0
StarPU: A Unified Runtime System for Heterogeneous Multicore Architectures
Vulkan video extensions for accelerated H.264 and H.265 encode
Ultra Ethernet Consortium
Libjpeg-Turbo 3.0.0
Re:Faiss: A library for efficient similarity search
on Thursday March 30, 2023 @05:25PM (#35379007)
There is a wip [0] on RAFT [1] integration to faiss as an implementation of cuda gpu backed indices, although you can use RAFT directly.
[0] https://github.com/facebookresearch/faiss/pull/2521 [1] https://github.com/rapidsai/raft
Intel Open-Sources SYCLomatic Tool for CUDA to SYCL Migration
Nvidia Composable On-Package Architecture
Asus DDR5 to DDR4 Converter Card
HSE: Heterogeneous-Memory Storage Engine
Eigen 3.4
Intel Ponte Vecchio: 45 Tflops, 5TBps HBM2e
Re:Triton: Open-Source GPU Programming for Neural Networks
on Thursday July 29, 2021 @05:30AM (#27994521)
Note, that there is a new (substantially accelerated) "autoscheduler" implementation in TVM (one of competitors in the article linked), details can be found in https://arxiv.org/abs/2006.06762
Samsung Unveils Memory Module Incorporating CXL Interconnect
Vulkan Video Acceleration Extensions
Re:The Engineering Cameras and Microphone on the Perseverance Rover (2...
on Friday February 26, 2021 @02:07AM (#26272080)
Thx for suggesstion, but the images that can be downloaded out there aren't matching the description that is given at the article (they seems to be processed in some way), nor they are sequenced and can't to be downloaded in batch. Based on article, i am expecting to get 1024x1024 monochrome images.
Re:The Engineering Cameras and Microphone on the Perseverance Rover (2...
on Thursday February 25, 2021 @07:02PM (#26269994)
I want to download the sequence of images (raw, unprocessed) taken by LCAM. Where (other then from "future" PDS release) can i download this sequence?
Re:Perseverance rover’s descent and touchdown on Mars [video]
on Tuesday February 23, 2021 @04:49AM (#26235493)
Anyone can point me to the place where i can find the original video files (that are received from the rover)?
Re:A Family of Better Random Number Generators
on Thursday October 15, 2020 @05:55AM (#24786770)
One might be interested in https://github.com/wangyi-fudan/wyhash