泰晓资讯·11月 / 第二期 / 2015

Google：我们最新的机器学习系统 TensorFlow，开源了速来围观

TensorFlow has extensive built-in support for deep learning, but is far more general than that – any computation that you can express as a computational flow graph, you can compute with TensorFlow

Any gradient-based machine learning algorithm will benefit from TensorFlow’s auto-differentiation and suite of first-rate optimizers.

And it’s easy to express your new ideas in TensorFlow via the flexible Python interface.

TensorFlow was built from the ground up to be fast, portable, and ready for production service. You can move your idea seamlessly from training on your desktop GPU to running on your mobile phone.

这个公司的首要财富，不是已经创造了什么，而是启发创造、开放与分享的“工程师教”。非执着于效率与算计的驱动方式所能达到。

微软亚洲研究院：我们也开源了分布式机器学习工具包 (DMTK)

DMTK由一个服务于分布式机器学习的框架和一组分布式机器学习算法构成，是一个将机器学习算法应用在大数据上的强大工具包。

DMTK分布式机器学习框架：它由参数服务器和客户端软件开发包（SDK）两部分构成。

机器学习，要火！

ARM 界 3U 齐来：ARM 发布 64 位超低功耗的 A35，高通公布骁龙 820 细节，三星发布 Exynos 8890

ARM 出品的 IP 由 32 位时代的 { A5|A7, A9, A15|A17 } 过渡到了 64 位时代的 { A35, A53, A57|A72 }。

四核骁龙 820 含两个 Kryo LP (Little) 和两个 Kryo HP (Big)，而八核 Exynos 8890 含四个 A53 (Little) 和四个 M1 (Big)。

Khronos: 来看我们在 SIGGRAPH 2015 Asia 上的幻灯（pdf）

OpenCL, SPIP-V, OpenGL, Vulakn, WebGL 2.0

Vulkan 与 Linux window system 的整合层面，借助 DRI3，Mir、Wayland 以及 X.org 将支持之。

AMD GPU 之开源驱动 “AMDGPU” 支持 PowerPlay 啦，大放异彩：5 倍性能提升、Tonga 与 Fiji 核心之性能飙升

闲时做一枚安静的 GPU，忙时超超频，也许是性能飙升的原因。另看到一说，没有 re-clocking 初始化时，GPU 以安全频率跑，由此性能也打折。

Phoronix: KDBUS 回炉重铸

Greg Kroah-Hartman commented on a kernel mailing list thread this weekend about the KDBUS refactoring, “based on the result of the code being in Rawhide, it is now being reworked / redesigned. The result will be posted for review ‘when it’s ready’.”

Apparently with the new design, KDBUS will hopefully be more universal and less D-Bus specific, based on the recent systemd 2015 conference.

That kernel mailing list thread has since boiled down into arguments over KDBUS developers not posting any design documents and doing much of their development discussion outside of the kernel mailing list, often via IRC or other forms of communication by systemd developers.（交易中请勿使用阿里旺旺以外的聊天工具沟通！^_^;）

总线的多播情形中，消息按序、可靠递送成了一个棘手问题，参见“KDBUS 实现为一个内核驱动，真地好吗？”