Implementing Highly Efficient Transpose Kernel for Hopper Architecture with Mojo
By
timmyd
Sesame, salt, and substance. A flagship bake.
Summary
The article explains how to implement a highly efficient transpose kernel for the Hopper architecture using Mojo, achieving a bandwidth of 2775.49 GB/s. It compares the performance with pure CUDA on the same hardware, showing similar results.
Key quotes
· 3 pulledThe best kernel archives a bandwidth of 2775.49 GB/s, i.e. 84.1056%.
Mojo can achieve CUDA-like performance on the same task.
You may comp
You might also wanna read
Four practical steps to control Azure Foundry token costs for agentic AI workloads
This article provides practical guidance on controlling token costs in Microsoft Azure Foundry, particularly for agentic AI workloads where
purplefrogsystems.com·42m ago
MerLean-Prover: A Recursive Agent Harness for Lean 4 Theorem Proving Outperforms Baselines
MerLean-Prover is an end-to-end Lean4 theorem prover that replaces 'sorry' declarations with kernel-checkable proofs using three agent types
arxiv.org·49m ago
Why small pull request policies can backfire on software quality
The article critiques a common software engineering policy that limits pull requests (PRs) to small sizes (e.g., 500 lines, few files). Whil
apenwarr.ca·2h ago
How Anthropic contains Claude's expanding access across its products
Anthropic describes how it has evolved its approach to granting Claude, its AI assistant, increasingly broad access to internal systems over
anthropic.com·4h ago
Testing Cursor's Jira integration: How ticket quality affects AI agent performance
Cursor launched a Jira integration that lets developers assign tickets directly to an AI agent, eliminating context switching. The author te
bit.ly·4h ago
Netflix engineer's open-source tool cuts AI token usage by up to 90%
Netflix senior engineer Tejas Chopra created software called "Project Headroom" that prunes redundant tokens from AI agent instructions befo
theregister.com·4h ago