{"blog_url":"https://vengineer.hatenablog.com/","provider_name":"Hatena Blog","html":"<iframe src=\"https://hatenablog-parts.com/embed?url=https%3A%2F%2Fvengineer.hatenablog.com%2Fentry%2F2020%2F07%2F28%2F060000\" title=\"Google TPU v2/v3 \u306e\u8a73\u3057\u3044\u8a18\u4e8b - Vengineer\u306e\u5984\u60f3\" class=\"embed-card embed-blogcard\" scrolling=\"no\" frameborder=\"0\" style=\"display: block; width: 100%; height: 190px; max-width: 500px; margin: 10px 0px;\"></iframe>","title":"Google TPU v2/v3 \u306e\u8a73\u3057\u3044\u8a18\u4e8b","author_url":"https://blog.hatena.ne.jp/Vengineer/","height":"190","provider_url":"https://hatena.blog","type":"rich","author_name":"Vengineer","published":"2020-07-28 06:00:00","version":"1.0","width":"100%","blog_title":"Vengineer\u306e\u5984\u60f3","description":"@Vengineer\u306e\u622f\u8a00 : TwitterSystemVerilog\u306e\u4e16\u754c\u3078\u3088\u3046\u3053\u305d\u3001\u3059\u3079\u3066\u306f\u3001SystemC v0.9\u516c\u958b\u304b\u3089\u59cb\u307e\u3063\u305f \u5148\u9031\u306e\u9023\u4f11\u306e\u6642\u306b\u898b\u3064\u3051\u305f COMMUNICATIONS of th ACM \u63b2\u8f09\u306e\u8a18\u4e8b \u300cA Domain-Specific Supercomputer for Training Deep Neural Networks\u300d Google TPU v2 \u306e\u8a73\u3057\u3044\u8a18\u4e8b\u3067\u3059\u3002PDF \u7248\u3082\u3042\u308a\u307e\u3059\u3002 Google TPU v2 Pod \u306f\u30012D-torus topology \u3067\u300116x16 2D torus\u306e\u3088\u3046\u3067\u3059\u3002 \u4e0b\u56f3\u306f\u3001\u5f15\u7528\u306e\u305f\u3081\u306b\u7d44\u307f\u8fbc\u307fURL\u3067\u8868\u793a\u3055\u2026","categories":[],"url":"https://vengineer.hatenablog.com/entry/2020/07/28/060000","image_url":"https://dl.acm.org/cms/attachment/7bfa1b68-8187-4512-b7bf-6214c6d698bd/f1.jpg"}