Commit 91adca2d authored by Kirill Smelkov's avatar Kirill Smelkov

.

parent de3ef2c0
...@@ -3,6 +3,13 @@ ...@@ -3,6 +3,13 @@
- scheduler won't use free P if a G is taking time and other Gs are queued after it on the same P: - scheduler won't use free P if a G is taking time and other Gs are queued after it on the same P:
https://groups.google.com/forum/#!topic/golang-nuts/8KYER1ALelg https://groups.google.com/forum/#!topic/golang-nuts/8KYER1ALelg
( not quite so exactly. The scheduler is stealing one, but if GC wants to
stop the world, and one goroutine persists looping, others won't be run until
the looping one parks and GC performs.
From this point of view the above issues is actually about
https://github.com/golang/go/issues/10958 )
e870f06c (runtime: yield time slice to most recently readied G) e870f06c (runtime: yield time slice to most recently readied G)
ea0386f8 (runtime: improve randomized stealing logic) ea0386f8 (runtime: improve randomized stealing logic)
4bb491b1 (runtime: improve scheduler fairness) 4bb491b1 (runtime: improve scheduler fairness)
...@@ -46,6 +53,11 @@ runtime/HACKING.md ...@@ -46,6 +53,11 @@ runtime/HACKING.md
calls convI2I all the time (no caching) which is not negliable calls convI2I all the time (no caching) which is not negliable
( recheck it is still true after https://github.com/golang/go/commit/3d1699ea:
-> it is still the case; convI2I cost is ~ 8ns - i.e. it is not big for
non-hot codepaths - uncontended mutex lock/unlock is 15-25ns. )
- x/net/trace to trace requests and connection logging (and packets ?) - x/net/trace to trace requests and connection logging (and packets ?)
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment