Turns out that `pool.Put(buf)` had to *allocate* because we needed to turn
`[]byte` into `interface{}`. Apparently, we've never done this correctly we just
never noticed because we never really used buffer pools extensively.
However, since migrating yamux to a buffer-pool backed buffer, this started
showing up in allocation profiles.
License: MIT
Signed-off-by: Steven Allen <steven@stebalien.com>
* go-datastore and friends: GetSize
* badger: new release, fewer allocations
* go-mplex: send fewer packets
* go-bitswap: pack multiple blocks in a single message, fewer allocations
* go-buffer-pool: replace the buffer pool from go-msgio
* yamux: fixed data race and uses go-buffer-pool for stream read-buffers to
reduce memory and allocations.
* go-libp2p-secio: get rid of a hot-spot allocation
* go-libp2p-peerstore: reduced allocations (at the cost of some memory)
More?
License: MIT
Signed-off-by: Steven Allen <steven@stebalien.com>
Also:
* Switch to gogo for filestore for consistency.
* Use the "faster" codegen for fewer allocations.
License: MIT
Signed-off-by: Steven Allen <steven@stebalien.com>
* Most of our datastores barf on non []byte values.
* We have to have a bunch of "is this a []byte" checks.
* Saves some allocations.
License: MIT
Signed-off-by: Steven Allen <steven@stebalien.com>
Also:
* Update the blockstore/blockservice methods to match.
* Construct a new temporary offline dag instead of having a
GetOfflineLinkService method.
License: MIT
Signed-off-by: Steven Allen <steven@stebalien.com>
Making this deterministic keeps us from creating an exponential amount
of objects as the number of pins in the set increases.
License: MIT
Signed-off-by: Jeromy <jeromyj@gmail.com>
Also change existing 'Node' type to 'ProtoNode' and use that most
everywhere for now. As we move forward with the integration we will try
and use the Node interface in more places that we're currently using
ProtoNode.
License: MIT
Signed-off-by: Jeromy <why@ipfs.io>
This change adds the /ipfs/bitswap/1.1.0 protocol. The new protocol
adds a 'payload' field to the protobuf message and deprecates the
existing 'blocks' field. The 'payload' field is an array of pairs of cid
prefixes and block data. The cid prefixes are used to ensure the correct
codecs and hash functions are used to handle the block on the receiving
end.
License: MIT
Signed-off-by: Jeromy <why@ipfs.io>
Switched from using a map to an array since the bounds are
small and fixed. This should save us some significant time and on
accesses
License: MIT
Signed-off-by: Jeromy <why@ipfs.io>
