somnial's submissions

1.		What happens when a GPU writes memory (doubleword.ai)
		5 points by somnial 6 days ago \| past \| discuss
2.		What happens when a GPU reads memory? (doubleword.ai)
		15 points by somnial 24 days ago \| past
3.		The case for disaggregated LLM serving (doubleword.ai)
		4 points by somnial 25 days ago \| past
4.		On-the-fly snapshot compression for elastic inference at scale (doubleword.ai)
		6 points by somnial 26 days ago \| past \| 1 comment
5.		NVLink, NVSwitch, and All That (doubleword.ai)
		5 points by somnial 46 days ago \| past \| 2 comments
6.		The Anatomy of an Instruction Pipeline Hazard (hiraditya.github.io)
		9 points by somnial 54 days ago \| past
7.		Width vs. Depth: Speculating on the Margin (doubleword.ai)
		17 points by somnial 66 days ago \| past \| 1 comment
8.		Pushing memory bound CUDA kernels past the speed of light with data compression (fergusfinn.com)
		2 points by somnial 3 months ago \| past
9.		Speculative KV coding: ~4× losslessly compressed KV cache using a small model (fergusfinn.com)
		2 points by somnial 3 months ago \| past
10.		70x faster cold(ish) starts for SGLang (fergusfinn.com)
		1 point by somnial 4 months ago \| past
11.		LLM powered data structures: A lock-free binary search tree (fergusfinn.com)
		1 point by somnial 7 months ago \| past
12.		Parallel Primitives for Multi-Agent Workflows (fergusfinn.com)
		1 point by somnial 8 months ago \| past
13.		Scheduling in LLM Inference (fergusfinn.com)
		1 point by somnial 9 months ago \| past
14.		How fast can an LLM go? (fergusfinn.com)
		2 points by somnial 10 months ago \| past