Incredible work and great patches as always 💪
中文: 一如既往的出色工作和精彩补丁💪
it was an eventful weekend, and solana now broadcast nearly twice as many shreds on XDP than on the legacy slower code and we retransmit nearly half on XDP BIG IBRL week https://twitter.com/alessandrod/status/2063977291370361059/photo/1
中文: 这是一个充满兴的周末,而SOLAna现在在XDP上播放的碎片几乎是旧旧版较慢代码的两倍 我们在XDP上重新传输近一半 大IBRL周
working on making transaction execution never pagefault in agave 4.2 in this branch oracle updates are basically free, fastest in this profiled block was 9us but it aborted. fastest successful 17us (humidifi) on avg all the top propAMMs are below 30us we're getting pretty good… https://twitter.com/alessandrod/status/2063514386812317790/photo/1
中文: 在 agave 4.2 中完成交易执行绝不会错 在此分支中,神赘更新基本免费,此配置区块中速度最快为9us,但成功失败。 所有顶级 pro-AMM 均低于 30 岁 我们越来越好了......
oh how the tables have turned
中文: 哦,桌子是怎么转动的
193 nodes on testnet have XDP on now. Retransmit time plummeting. Once we're done, retransmit will settle at ~400 microseconds. We finish testnet, then we do mainnet, then we finally do 100M CUs. https://twitter.com/alessandrod/status/2056771674293580001/photo/1
中文: 测试网上的193个节点现在已具有XDP。 重新传输时间急剧下降。完成后,再传输将稳定在约400微秒。 我们完成测试,然后做主网,最后再做100M CU。
see if you can tell when coinbase switched to agave 4.0 and got 4x faster https://twitter.com/alessandrod/status/2055473484420252151/photo/1
中文: 看看你是否可以判断Coinbase何时切换到 agave 4.0,并且速度更快 4 次
interesting, turns out some of you might actually be scientists
中文: 有趣的是,有些人可能真的是科学家
do it, I dare you
中文: 做,我敢
wait a minute
中文: 等一下
before the americans come and ruin summer
中文: 美国人来来之前,毁掉了夏天
I can't explain it but I've been laughing at this for the past 10 minutes https://twitter.com/alessandrod/status/2047392695413625219/photo/1
中文: 我无法解释,但过去10分钟我一直在笑这个
one of the largest validators turbine retransmit from 600ms to ~0.8ms agave 4.0 w/ XDP is going to be the biggest solana upgrade to date https://twitter.com/alessandrod/status/2046822618406522925/photo/1
中文: 最大的验证器之一 涡轮机从600米重新传输到约0.8米 agave 4.0 与 XDP 将成为迄今为止最大的 sonana 升级版
there's a 2x there coming in 4.1
中文: 有2倍的4.1
if in 4.0 you don't turn XDP on I pivot to AI
中文: 如果在4.0中你无法将XDP转向人工智能
there are many things to say about the dwarkesh jensen interview, but "as much as needed, as little as possible" perfectly summarizes how platforms should be built (including solana)
中文: 关于德克什·肯森的采访有很多话要说,但 尽可能少 完美地总结平台的构建方式(包括 salana)
Once again, with his tests @ax_1000x helped me find a bug before it reached mnb. This happens multiple times each release cycle. If you want solana to be safe and fast, consider staking with him, this is exactly the kind of operator we need to be rewarding. (Also give him a…
中文: 再次,通过测试 @ax_1000x ,我在发现错误时发现了漏洞。每个发布周期都会发生多次。 如果你想让 solana 安全且快速,可以考虑和他合作,这正是我们需要奖励的操作人员。 还给他一个......
bro so locked in forgets to tweet he's going to speak at rustconf @zbr0wn
中文: 兄弟在推特上非常困惑,他将在 rustconf @zbr0wn 上发话
this is cool but they'll never know what it feels like fixing turbine
中文: 这很酷,但他们永远不知道修理涡轮机是什么感觉
been doing more work on my turbine integration in slowlana and I've already found a major latency bug imma fix tomorrow I also realised that it's a great way to visualize how block building is continuous and blocks are just logical entities 🧬 https://twitter.com/alessandrod/status/2041207838245286117/photo/1
中文: 我在慢点拉纳的涡轮机集成方面一直在做更多的工作,而且我已经发现明天会修复一个重大的延迟漏洞 我也意识到,这是直观地直观地将块构建视为连续的、且只是逻辑实体的绝佳方式🧬
everything is eBPF
中文: 一切都是eBPF
cooked shred tracing in slowlana so I can accurately correlate replay stalls to turbine stalls look at this slot stalling for ~250ms because of a single shred that gets lost and repaired late lmao next couple of weeks we finally fix turbine and get 200ms slots https://twitter.com/alessandrod/status/2040480668120195258/photo/1
中文: 慢速回火的熟碎痕迹,以便我能准确地将回放摊位与涡轮压块关联起来 看看这个档位因一块碎片而拖延约250米,这种碎片会丢失并修复晚期的马 接下来的几周,我们终于修复了涡轮机,并获得了200ms的插槽
swqos next then connection churning
中文: 下一个 然后连接搅动
looking at agave 1.18 and thinking I can fix you you galactic pos
中文: 看着 agave 1.18 ,想着我能给你解决。