#SYCL
Intel GPU nemám, ale výsledky SYCL proti Vulkanu bych si rád přečetl
November 21, 2026 at 5:46 PM
lapackpp-sycl
C++ API for the Linear Algebra PACKage (with SYCL and oneMKL)
aur.archlinux.org
November 20, 2026 at 8:18 AM
blaspp-sycl
C++ API for the Basic Linear Algebra Subroutines (with SYCL and oneMKL)
aur.archlinux.org
November 20, 2026 at 6:18 AM
Mám tu někoho s moderní Intel GPU a chutí odzkoušet lokální běh AI? Mám vlastní AI engine - na CUDA, ROCm, Vulkan mám HW, nemám odzkoušený SYCL backend na Intelu.
computer monkey
ALT: computer monkey
static.klipy.com
November 21, 2026 at 1:36 PM
See you on Friday for the last day of #SC26 for Porting and Benchmarking Chapel on Emerging RISC-V Hardware: An HPC Viability Study Workshop and Performance-Portable Exact Gaussian Process Regression in Python via SYCL and oneMath featuring HPX
November 19, 2026 at 10:02 PM
cuda was relevant Back Then. doesnt really matter nowadays in my experience. everything supports HIP/SyCL/OpenCL or even straight Vulkan
November 16, 2026 at 9:03 AM
Intel Arc Pro B60 Achieves 73% Prompt Processing Surge with llamacpp b12602 SYCL Runtime Optimization + Video

Introduction: Recent architectural benchmarks from INGOAMPT (( demonstrate that compiling `llama.cpp` at build `b12602` delivers a 69% to 73% acceleration in prompt processing (prefill)…
Intel Arc Pro B60 Achieves 73% Prompt Processing Surge with llamacpp b12602 SYCL Runtime Optimization + Video
Introduction: Recent architectural benchmarks from INGOAMPT (( demonstrate that compiling `llama.cpp` at build `b12602` delivers a 69% to 73% acceleration in prompt processing (prefill) throughput on Intel Arc Pro B60 GPUs. While auto-regressive answer-token generation (decode) remains bottlenecked by memory bandwidth, this prefill breakthrough transforms local AI agent responsiveness, Retrieval-Augmented Generation (RAG) pipelines, and air-gapped security log analysis. Understanding how to properly configure Intel oneAPI SYCL runtimes, isolate local inference APIs, and verify GPU memory boundaries is essential for modern IT and cybersecurity engineers deploying sovereign AI infrastructure.
undercodetesting.com
November 21, 2026 at 5:12 AM
GGUF models run on the hardware in front of you: Metal, CUDA, Vulkan, SYCL or CPU only. Private inference on your own hardware; a person on site. On-site assessment: privatelocalai@proton.me https://localai.hex7.com
November 19, 2026 at 11:00 AM
CMake now speaks SYCL. CMake 4.5 adds experimental, opt-in support for SYCL as a native language, working with AdaptiveCpp, DPC++, and Clang. Plus: SYCL 2020 Revision 13 adds split headers that can cut compile times.

www.khronos.org/blog/sycl-re...
#SYCL #CMake
November 12, 2026 at 6:21 PM
XeStrata Runs Qwen3.8-Flash-Next on Intel, AMD, and NVIDIA GPUs

XeStrata runs the 125B Qwen3.8-Flash-Next on ordinary PCs via SYCL, not CUDA or HIP. One code path for Intel, AMD, and NVIDIA GPUs. Free build covers Intel and AMD with open-source software only; contrib-llvm adds NVIDIA.
November 18, 2026 at 2:40 PM
Alexander Strack, Marcel Graf, Alexander Van Craen, Dirk Pfl\"uger: Performance-Portable Exact Gaussian Process Regression in Python via SYCL and oneMath https://arxiv.org/abs/2611.17565 https://arxiv.org/pdf/2611.17565 https://arxiv.org/html/2611.17565
November 16, 2026 at 6:40 AM
1250億パラメータを普通のPCで、ってところより、わたしはSYCLで書き直したって部分に引っかかりました。つまり、NVIDIAじゃないと門前払いだったところに、誰かが黙って裏口を作ったってことなんよな・・。IntelのGPUって、AIの話だと正直存在を忘れられがちだったので。ちなみに、こういう「対応してないなら自分で書く」人たちが一番強いと思ってます。・・メーカーより先にユーザーが動くの、ミニ四駆の改造みたいでちょっと好き
無料でIntel・AMD・NVIDIAのGPUを使ってQwen3.8-Flash-Nextをローカルで動かせる「XeStrata」
1250億パラメーターのAIモデル「Qwen3.8-Flash-Next」を一般的なPCで動作させるソフトウェアの「Strata」をベースにGPUの計算処理をオープンな標準規格「SYCL」で書き直すことでIntel製GPUにも対応し、Intel・AMD・NVIDIA製のGPUを共通のコードで扱えるようにした無料のオープンソースソフトウェアが「XeStrata」です。
gigazine.net
November 19, 2026 at 3:30 AM
CMake now speaks SYCL

CMake 4.5 adds experimental, opt-in support for SYCL as a native language. Less custom build logic, and it works with AdaptiveCpp, both DPC++ distributions, and upstream Clang.

Also in the post: SYCL 2020 Revision 13 adds split headers […]

[Original post on fosstodon.org]
November 12, 2026 at 6:20 PM
GGUF models run on the hardware in front of you: Metal, CUDA, Vulkan, SYCL or CPU only. Private AI on your own hardware, support on site. Ask about an on-site assessment: privatelocalai@proton.me https://localai.hex7.com
November 16, 2026 at 11:01 PM
Khronos Compute is now open. It is a new Discord community for developers working with SYCL and OpenCL, where you can ask questions, share projects and talk directly with working group members. Join us and tell us what you are building: khr.io/computediscord.
#OpenCL #SYCL #Compute #support
Join the Khronos Compute Discord Server!
Check out the Khronos Compute community on Discord - hang out with 37 other members and enjoy free voice and text chat.
khr.io
November 12, 2026 at 4:52 PM
Khronos Compute is now open. It is a new Discord community for developers working with SYCL and OpenCL, where you can ask questions, share projects and talk directly with working group members. Join us and tell us what you are building: khr.io/computediscord.
#OpenCL #SYCL #Compute #support
Join the Khronos Compute Discord Server!
Check out the Khronos Compute community on Discord - hang out with 37 other members and enjoy free voice and text chat.
khr.io
November 12, 2026 at 4:53 PM
llama.cpp can change outputs without new weights.

Gemma 4 image resizing drops letterbox padding; qwen3-coder handles think blocks when reasoning_format=none; SYCL fixes rounding when copying f32 values into quantized formats. Re-test preprocessing and backend outputs…

Read more on Kimbodo:
What Recent llama.cpp Changes Mean for Reliable Open-Model Inference
What Happened The developments documented here are primarily llama.cpp inference-engine changes, not new open-weight model releases. They include a fix for shared decision prefixes that preserves output embeddings when tasks…
kimbodo.com
November 16, 2026 at 6:21 PM
Khronos Compute is now open. It is a new Discord community for developers working with SYCL and OpenCL, where you can ask questions, share projects and talk directly with working group members. Join us and tell us what you are building: https://khr.io/computediscord.
#opencl #sycl #compute #support
November 12, 2026 at 4:53 PM