Update FarmHashFingerprint64.java #7549
Conversation
Remove the data dependency between loads and both the rotates. We can rewrite the code in such a way that rotates can start computing as soon as the required loads are present instead of waiting for both the loads to retire.
|
Thanks for your pull request! It looks like this may be your first contribution to a Google open source project. Before we can look at your pull request, you'll need to sign a Contributor License Agreement (CLA). View this failed invocation of the CLA check for more information. For the most up to date status, view the checks section at the bottom of the pull request. |
|
This might be a good improvement, but our existing (internal) benchmarks are kind of questionable, and our fleet-wide profiling shows a negligible amount of CPU in these functions for us, so we are unlikely to improve the benchmarks enough to see if this would help. If you want to try to put something together with JMH, you could, but I'm not even sure I'd guarantee that we'd end up merging this, given that it makes the implementation code different from the C++ implementation. (Now, if you could get the C++ implementation updated to match this because this change showed benefits there, then we might follow.) |
|
Thank you Chris for the suggestions. One meta question on the comment -- is there a list somewhere, of functions in Guava, which consume the highest amount of CPU ?. |
Remove the data dependency between loads and both the rotates.
We can rewrite the code in such a way that rotates can start computing as soon as the required loads are present instead of waiting for both the loads to retire.