TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

Distance-Based ISA for Efficient Register Management

14 点作者 abhi9u2 天前

4 条评论

thechao2 天前
When I read these papers I always wonder if the authors did <i>any</i> research into priors before doing their work. I usually see sus-af &quot;redos&quot; in low tier AI (the fad of the decade); so, CPU arch is original? Another commenter pointed out Mill is probably the closest one I can think of off the top of my head. I think the old rolling window Power stuff IBM published in the ... 90s? might be exactly this? I can&#x27;t remember if that arch used an autowindow (STRAIGHT) or if it was more manual.<p>EDIT: Derp! RIP-relative x64? I&#x27;ve not coded x64 in 15 years, so I don&#x27;t remember how pervasive the addressing mode is.<p>I guess the author&#x27;s are trying to explore a ground-up implementation? Maybe the uarch is really exotic?
ahartmetz2 天前
They say that &quot;the distance-based representation is novel&quot;, but it seems at least similar to the Mill CPU&#x27;s belt idea.
stefanha2 天前
The article mentions getting SPEC CPU running but doesn&#x27;t share performance results or scalability results (now the CPU can decode twice as many instructions, etc). Can someone who has been following the research in this area share some results?
hyperhello2 天前
A question, is it distance as in from the PC (absolute addressing from a datum) or as in the number of instructions executed (meaning a 4 byte conditional branch can shift the whole history)?
评论 #44182017 未加载