I've been trying to understand how this integer pool works. It is a lot of bit fiddling stuff I can't wrap my head around. I'm assuming there is a concept I'm missing with the m2id array and how it is or'ed with index 'n' that I don't know and would clear up a lot of my confusion. Are there are any general concepts/CS-theory that explain this seemingly-looking-simple code. I've put comments in the code to try and state my current understanding and where I am totally confused.
// Copyright 2009 The Go9p Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
//Original source: https://github.com/rminnich/go9p/blob/master/clnt_pool.go
package go9p
import "sync"
var m2id = [...]uint8{ // I think this is where the magic is.
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 4,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 5,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 4,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 6,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 4,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 5,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 4,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 7,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 4,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 5,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 4,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 6,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 4,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 5,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 4,
0, 1, 0, 2, 0, 1, 0, 3,
0, 1, 0, 2, 0, 1, 0, 0,
}
type pool struct {
sync.Mutex
need int
nchan chan uint32
maxid uint32
imap []byte
}
func newPool(maxid uint32) *pool {
p := new(pool)
p.maxid = maxid
p.nchan = make(chan uint32)
return p
}
func (p *pool) getId() uint32 {
var n uint32 = 0
var ret uint32
p.Lock()
for n = 0; n < uint32(len(p.imap)); n++ {
// it looks like every 0...n position of imap will be incremented to 255.
if p.imap[n] != 0xFF {
break
}
}
if int(n) >= len(p.imap) {
// This seems to be just growing the imap slice as needed.
// I don't quite understand the constant of '8' here.
m := uint32(len(p.imap) + 32)
if uint32(m*8) > p.maxid {
m = p.maxid/8 + 1
}
b := make([]byte, m)
copy(b, p.imap)
p.imap = b
}
if n >= uint32(len(p.imap)) {
// If you get here the I'm assuming all the ID's are used up and putId will return you the next released ID.
p.need++
p.Unlock()
ret = <-p.nchan
} else {
// This part I'm having a hard time grasping.
// It seems that each index of imap is incremented
// from 0 to 255 and magically or'd with ret to increment to the next number?
ret = uint32(m2id[p.imap[n]])
p.imap[n] |= 1 << ret
ret += n * 8
p.Unlock()
}
return ret
}
func (p *pool) putId(id uint32) {
p.Lock()
if p.need > 0 {
p.nchan <- id
p.need--
p.Unlock()
return
}
// This doesn't play well with what I thought was going on. I though that.
// I was thinking that imap[0] would always somehow magically return all the
// values from 0 to 255 and imap[1] would return 256 += 255 and so on.
// How does this work?
p.imap[id/8] &= ^(1 << (id % 8))
p.Unlock()
}
Optimization often leads to obscurity. Start with the basic concept. The pool of available Ids is represented by the underlying bit array of a slice of bytes. Id 19 is represented by left-to-right byte 2 (19 / 8) and right-to-left bit 3 (19 % 8).
Here's a simple implementation, ignoring details like locking and growing the bit array.
Output:
We can optimize this loop
by replacing it with a table (
m2id
) lookup for the bit shift value.The
m2idInit()
function shows how them2id
table bit shift values are calculated.For example,
Output:
There is no magic.
References:
Bit manipulation
The Go Programming Language Specification, Arithmetic operators