<?xml version="1.0" encoding="UTF-8"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Yestino - The Signal · CUDA FP32</title><link>https://yestino.com/entities/cuda-fp32-5db068</link><description>Every event involving CUDA FP32</description><language>en</language><atom:link href="https://yestino.com/entities/cuda-fp32-5db068/feed.xml" rel="self" type="application/rss+xml"/><item><title>trunk/312ef3ee1cd45fbee99f2e4e8f6035f05ceeebed: Add BF16x9 precision mode for CUDA FP32 matmul (#195301)</title><link>https://yestino.com/events/trunk-312ef3ee1cd45fbee99f2e4e8f6035f05ceeebed-add-bf16x9-pr-3f9dff</link><guid isPermaLink="true">https://yestino.com/events/trunk-312ef3ee1cd45fbee99f2e4e8f6035f05ceeebed-add-bf16x9-pr-3f9dff</guid><pubDate>Wed, 02 Sep 2026 05:59:39 GMT</pubDate><description>Human Note Cublas has a cool bf16x9 mode for fp32 gemms; in theory it should be more accurate than tf32; ill keep prodding to get some intersting data showing…
Sources: PyTorch Releases</description></item></channel></rss>