<rss version="2.0" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:atom="http://www.w3.org/2005/Atom"><channel><title>Hacker News: hahajahen</title><link>https://news.ycombinator.com/user?id=hahajahen</link><description>Hacker News RSS</description><docs>https://hnrss.org/</docs><generator>hnrss v2.1.1</generator><lastBuildDate>Sun, 12 Apr 2026 06:24:47 +0000</lastBuildDate><atom:link href="https://hnrss.org/user?id=hahajahen" rel="self" type="application/rss+xml"></atom:link><item><title><![CDATA[New comment by hahajahen in "Bend: a high-level language that runs on GPUs (via HVM2)"]]></title><description><![CDATA[
<p>OP is claiming amazing results, people are poking obvious holes that good single core implementations completely rip the scalability claims to shreds. Near linear scalability is not impressive if—even at the highest throughput—the computation pales by comparison to Rust on a single core.<p>I do not see how the comparison to the iPhone here stands.</p>
]]></description><pubDate>Sat, 18 May 2024 18:13:04 +0000</pubDate><link>https://news.ycombinator.com/item?id=40400838</link><dc:creator>hahajahen</dc:creator><comments>https://news.ycombinator.com/item?id=40400838</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40400838</guid></item><item><title><![CDATA[New comment by hahajahen in "Bend: a high-level language that runs on GPUs (via HVM2)"]]></title><description><![CDATA[
<p>The insult against anyone who pushes back a little bit is not a good sign, I agree. From all we can see now, the massive speedups being claimed have <i>zero optimal baselines</i>. I badly would like to identify these.</p>
]]></description><pubDate>Sat, 18 May 2024 00:07:08 +0000</pubDate><link>https://news.ycombinator.com/item?id=40395371</link><dc:creator>hahajahen</dc:creator><comments>https://news.ycombinator.com/item?id=40395371</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40395371</guid></item><item><title><![CDATA[New comment by hahajahen in "Bend: a high-level language that runs on GPUs (via HVM2)"]]></title><description><![CDATA[
<p>The fact that you don’t know the answer to this question, and don’t even seem to think it is relevant, is chilling.<p>People want to be able to ground your work—which you are claiming is the “parallel future of computation”—in <i>something</i> familiar. Insulting them and telling them their concerns are irrelevant just isn’t going to work.<p>I would urge you to think about what a standard comparison versus Haskell would look like. Presumably it would be something that dealt with a large state space, but also top down computation (something you couldn’t easily do with matrices). Big examples might include simply taking a giant Haskell benchmark (given the setting of inets it seems like a natural fit) that is implemented in a fairly optimal way—-both algorithmically and also wrt performance—-and compare directly on large inputs.<p>Sorry to trash on you here, not trying to come across as insulting, but I agree that “reductions per second” is meaningless without a nuanced understanding of the potentially massive encoding blowup that compilation introduces.<p>We want to believe, but the claims here are big</p>
]]></description><pubDate>Sat, 18 May 2024 00:04:59 +0000</pubDate><link>https://news.ycombinator.com/item?id=40395363</link><dc:creator>hahajahen</dc:creator><comments>https://news.ycombinator.com/item?id=40395363</comments><guid isPermaLink="false">https://news.ycombinator.com/item?id=40395363</guid></item></channel></rss>