rlp, trie: faster trie node encoding (#24126)

This change speeds up trie hashing and all other activities that require
RLP encoding of trie nodes by approximately 20%. The speedup is achieved by
avoiding reflection overhead during node encoding.

The interface type trie.node now contains a method 'encode' that works with
rlp.EncoderBuffer. Management of EncoderBuffers is left to calling code.
trie.hasher, which is pooled to avoid allocations, now maintains an
EncoderBuffer. This means memory resources related to trie node encoding
are tied to the hasher pool.

Co-authored-by: Felix Lange <fjl@twurst.com>

trie/node.go

@@ -28,8 +28,9 @@ import (
 var indices = []string{"0", "1", "2", "3", "4", "5", "6", "7", "8", "9", "a", "b", "c", "d", "e", "f", "[17]"}
 
 type node interface {
-	fstring(string) string
 	cache() (hashNode, bool)
+	encode(w rlp.EncoderBuffer)
+	fstring(string) string
 }
 
 type (
@@ -52,16 +53,9 @@ var nilValueNode = valueNode(nil)
 
 // EncodeRLP encodes a full node into the consensus RLP format.
 func (n *fullNode) EncodeRLP(w io.Writer) error {
-	var nodes [17]node
-
-	for i, child := range &n.Children {
-		if child != nil {
-			nodes[i] = child
-		} else {
-			nodes[i] = nilValueNode
-		}
-	}
-	return rlp.Encode(w, nodes)
+	eb := rlp.NewEncoderBuffer(w)
+	n.encode(eb)
+	return eb.Flush()
 }
 
 func (n *fullNode) copy() *fullNode   { copy := *n; return &copy }