- 03 Aug, 2015 6 commits
-
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
For GoogleCloudPlatform/gcsfuse#99.
-
- 29 Jul, 2015 34 commits
-
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
On Linux this significantly increases gcsfuse sequential read throughput.
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
As best I can tell, this instructs the kernel not to synchronize on the completion of a read request before sending further requests. This definitely means further read requests, but I believe it means any sort of request as well. It appears to make a big difference to gcsfuse read support.
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
It's too confusing with a distinct set of IDs for logging.
-
Aaron Jacobs authored
-
Aaron Jacobs authored
This adds up to a significant performance gain for gcsfuse, in its sequential read benchmark. Also ReadDirOp.
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
In preparation for allocating them up-front for ReadFile and ReadDir.
-
Aaron Jacobs authored
The new API lends itself toward reading directly into the output buffer, saving a memmove and allowing for easy reuse of output buffers.
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-
Aaron Jacobs authored
-