Mastodawn

Towards Feedback-to-Plan Decisions for Self-Evolving LLM Agents in CUDA Kernel Generation

Towards Feedback-to-Plan Decisions for Self-Evolving LLM Agents in CUDA Kernel Generation

Large language models (LLMs) have shown strong empirical gains as self-evolving agents for CUDA kernel generation, driven by feedback-conditioned planning across generations. However, how planning …

hgpu.org

HGPU group 3d ago

KForge: LLM-Driven Cross-Platform Kernel Generation for AI Accelerators

#CUDA #PTX #Triton #LLM #CodeGeneration #Intel

https://hgpu.org/?p=30832

KForge: LLM-Driven Cross-Platform Kernel Generation for AI Accelerators

Production inference increasingly targets a heterogeneous mix of accelerators. Agentic pipelines interleave reasoning, tool calls, and multi-agent coordination, each with distinct compute and memor…

hgpu.org

solarserver Apr 13

Start-up Hyca Heat: Klimafreundliche Prozesswärme für die Industrie:
Die Hyca-Heat-Technologie kann wahlweise industrielle Prozesswärme auf der Basis von Strom oder einem Brenngas bereitstellen. Das soll Flexibilität und Versorgungssicherheit bei geringen Kosten ermöglichen.

https://www.solarserver.de/2026/04/13/start-up-hyca-heat-klimafreundliche-prozesswaerme-fuer-die-industrie/

#PtX #Wasserstoff

Grafik: Forschungszentrum Jülich / Adobe Stock

solarserver Mar 16

Analyse: wo sich Power-to-X in Deutschland und Europa durchsetzt:
Noch sind Projekte, um aus grünem Strom Wasserstoff und synthetische Kraftstoffe zu machen, selten. Eine Studie von EY und VDMA zeigt, wo sich die Projekte anfangen durchzusetzen: in Häfen, Chemieparks und Industriezentren.

https://www.solarserver.de/2026/03/16/analyse-wo-sich-power-to-x-in-deutschland-und-europa-durchsetzt/

#Photovoltaik #Windenergie #PtX #Wasserstoff

Grafik: Thomas / stock.adobe.com

Szymon Bęczkowski Mar 5

Today Wolfspeed introduced their 10kV power MOSFET for general market.

https://investor.wolfspeed.com/news/news-details/2026/Wolfspeed-Introduces-Industrys-First-Commercially-Available-10000-V-Silicon-Carbide-Power-MOSFET/default.aspx

We have been working with these chips for 13 years now so for all your medium voltage needs please talk do to me.

We want to partner up with companies and research institutions that would like to use these chips and need a custom power module or a whole power converter.

#energy #energyTransition #PtX #electronics #academia

Wolfspeed Introduces Industry's First Commercially Available 10,000 V Silicon Carbide Power MOSFET

Revolutionary 10 kV SiC Technology Delivers Game-Changing Solution for Grid Modernization, Industrial Electrification and AI Data Center Infrastructure Wolfspeed, Inc., a global leader in silicon carbide (SiC) technology, today announced the industry's first commercially available 10 kV SiC power MOSFET. This milestone marks a defining moment in the industry — unlocking architectural freedom, delivering unprecedented system durability, and advancing access to reliable and sustainable power for the world's most demanding applications. This groundbreaking advancement challenges conventions in power conversion technology, delivering a transformative solution to modernize the grid and critical power infrastructure, to accelerate industrial electrification, and to unleash the potential for AI data center growth. Wolfspeed's 10 kV Bare Die MOSFET “This represents a historic leap in power electronics that will reshape how the world generates, distributes, and uses energy, unlocking the

HGPU group Jan 11

Equivalence Checking of ML GPU Kernels

#CUDA #PTX #LLM

https://hgpu.org/?p=30500

Equivalence Checking of ML GPU Kernels

With the rapid progress of deep learning and large language models (LLMs), companies now spend enormous sums executing GPU kernels. These kernels have, therefore, become prime targets for aggressiv…

hgpu.org

HGPU group Jan 4

KernelEvolve: Scaling Agentic Kernel Coding for Heterogeneous AI Accelerators at Meta

#CUDA #Triton #PTX #AI #Meta #LLM

https://hgpu.org/?p=30493

KernelEvolve: Scaling Agentic Kernel Coding for Heterogeneous AI Accelerators at Meta

Making deep learning recommendation model (DLRM) training and inference fast and efficient is important. However, this presents three key system challenges – model architecture diversity, ker…

hgpu.org

FCLC Dec 29

“Having explicit control over GPU SIMT engines implicit mask/predication register is something I actively wish I had on a monthly basis at the IR/PTX level”

#GPU #PTX #LLVMIR #LLVM

Yay!

50%

Nay!

WTF are you on about felix?

50%

Poll ended at Dec 30 at 2:05pm.

HGPU group Dec 29

Tilus: A Tile-Level GPGPU Programming Language for Low-Precision Computation

#CUDA #PTX #Triton #ProgrammingLanguages #Package

https://hgpu.org/?p=30481

Tilus: A Tile-Level GPGPU Programming Language for Low-Precision Computation

Serving Large Language Models (LLMs) is critical for AI-powered applications, yet it demands substantial computational resources, particularly in memory bandwidth and computational throughput. Low-…

hgpu.org

Show thread