make normalize ticksPerWheel more stable and high-performance #6070
Conversation
| n |= n >>> 4; | ||
| n |= n >>> 8; | ||
| n |= n >>> 16; | ||
| return (n < 0) ? 1 : (n >= 1073741824) ? 1073741824 : n + 1; |
There was a problem hiding this comment.
Can you make 1073741824 an static constant with a descriptive name ?
There was a problem hiding this comment.
Also share it with other places in the code.
There was a problem hiding this comment.
And constant better written in the form "1 << 30".
There was a problem hiding this comment.
@mmaxiaolei could you please address my comment ?
|
@mmaxiaolei also please sign our icla and adjust commit message to match our template: http://netty.io/s/ilcla |
@mmaxiaolei Could you please share some numbers? And add tests? |
|
@mmaxiaolei @slandelle also I wonder if it really makes a lot of difference as this one is only called in the constructor anyway. |
|
I agree with @slandelle and @normanmaurer ... please demonstrate the value this provides via JMH benchmark. Thanks. |
|
IMO benchmark not necessary. This optimization used also in JDK, referring to book "Hacker's Delight" sec 3-2 "Rounding Up/Down to the Next Power of 2". |
|
I am closing this now as it seems like @mmaxiaolei is not interested in this anymore. Please reopen if still interested and address the comments |
make normalize ticksPerWheel more stable and high-performance