DeepSeek开源华为昇腾适配项目，TileLang实现跨芯片编译
DeepSeek open-sourced six Huawei Ascend projects whose matrix kernels hit 99.8%…
DeepSeek将自研TileLang内核开源适配昇腾，直接打通国产算力部署路径，对关注国产替代的工程团队极具参考价值。
DeepSeek open-sourced six Huawei Ascend projects whose matrix kernels hit 99.8% of the chip's peak speed.
DeepSeek 开源了六个华为昇腾项目，其矩阵内核达到了芯片峰值速度的 99.8%。
this will help reduce reliance on Nvidia further for them.
这将有助于进一步减少他们对英伟达的依赖。
The headline project is TileLang, the programming language DeepSeek writes its own GPU kernels in.
头条项目是 TileLang，这是 DeepSeek 编写自身 GPU 内核所使用的编程语言。
Most of the operators that trained its V4 models are written in TileLang, and each now has an Ascend version. That means the same source code can compile for Huawei chips instead of Nvidia ones.
训练其 V4 模型的大多数算子都是用 TileLang 编写的，且每个算子现在都有昇腾版本。这意味着相同的源代码可以编译为适配华为芯片而非英伟达芯片的版本。
a 2nd repo, DeepGEMM-Ascend, does matrix multiplication and reports 431 BF16 TFLOPS against a 432 ceiling. DeepEP-Ascend, which moves data between chips, shipped with no license file while the rest are MIT, so nobody can legally reuse it yet.
第二个仓库 DeepGEMM-Ascend 执行矩阵乘法，报告值为 431 BF16 TFLOPS，而理论上限为 432。 DeepEP-Ascend 负责在芯片间移动数据，发布时未附带许可证文件，而其他项目均为 MIT 许可证，因此目前无人能合法复用该代码。
Liang Wenfeng told investors Huawei could start delivering training chips this quarter.
梁文锋告诉投资者，华为可能在本季度开始交付训练芯片。
更进一步：量化金融体系
看懂新闻只是起点——沿量化金融路径，把它变成能交付的工程能力