Skip to content

Commit

Permalink
Merge pull request #100 from zhusimaji/translation/performance/xla/ji…
Browse files Browse the repository at this point in the history
…t.md

performance/xla/jit.md 翻译
  • Loading branch information
leviding committed Oct 10, 2018
2 parents a72a6d8 + deca5f0 commit 92dc2d6
Showing 1 changed file with 1 addition and 1 deletion.
2 changes: 1 addition & 1 deletion performance/xla/jit.md
Expand Up @@ -10,7 +10,7 @@ TensorFlow / XLA 即时编译器通过 XLA 编译和运行 TensorFlow 图的各

有两种方式通过 XLA 运行 TensorFlow 计算图:一是用 CPU 或 GPU 设备上的即时编译操作,二是把操作放到 `XLA_CPU``XLA_GPU` TensorFlow 设备上。将操作直接放到一个 TensorFlow XLA 设备上强制执行,因此这种方法主要用于测试。

> 注意:The XLA CPU backend supports intra-op parallelism (i.e. it can shard a single operation across multiple cores) but it does not support inter-op parallelism (i.e. it cannot execute independent operations concurrently across multiple cores). XLA GPU 后端与标准的 TensorFlow 后端充分竞争,运行速度时快时慢。
> 注意:XLA CPU 后端支持外部运算并行化(比如它支持单个运算符并行计算),但是它不支持内部运算并行化(比如它无法执行多个独立运算符的并行计算)。XLA GPU 后端与标准的 TensorFlow 后端充分竞争,运行速度时快时慢。
### 开启即时编译

Expand Down

0 comments on commit 92dc2d6

Please sign in to comment.