Add a pooled recv buffer allocator #2362
Conversation
Side note: I don't think this is entirely true. It's only true if a |
|
@swift-server-bot test this please |
3 similar comments
|
@swift-server-bot test this please |
|
@swift-server-bot test this please |
|
@swift-server-bot test this please |
|
@swift-server-bot test this please |
glbrntt
force-pushed
the
gb-read-buffers
branch
4 times, most recently
from
abeb277 to
ab1cece
Compare
|
API breakage is the usual false positive of adding to a protocol with a default implementation: |
| var index = startIndex | ||
|
|
||
| // Sweep from after the lasted used index through to the end. | ||
| while index != self.buffers.endIndex { |
There was a problem hiding this comment.
This is definitely better, but these methods are still large and repetitive. This pair of while loops, for example, can certainly be refactored into an extension method on array which will allow us to reduce the number of lines of code in this method.
| if let result = try self.buffer?.modifyIfUniquelyOwned(minimumCapacity: nextBufferSize, body) { | ||
| // `result` can only be non-nil if `buffer` is non-nil. | ||
| return (self.buffer!, result) | ||
| } else if !self.buffers.isEmpty { |
There was a problem hiding this comment.
I don't think we need this conditional: the code written below has to work correctly regardless so I think we can just remove it.
Motivation: Channels can read `ChannelOptions.maxMessagesPerRead` times from a socket in each read cycle. They typically re-use the same buffer for each read and rely on it CoWing if necessary. If we read more than once in a cycle then we may CoW the buffer. Instead of reusing one buffer we can reuse a pool of buffers limited by `maxMessagesPerRead` and cycle through each, reducing the chance of CoWing the buffers. Modifications: - Extend `RecvByteBufferAllocator` to provide the size of the next buffer with a default implementation returning `nil`. - Add an recv buffer pool which lazily grows up to a fixed size and attempts to reuse buffers where possible if doing so avoids CoWing. Results: Fewer allocations
glbrntt
force-pushed
the
gb-read-buffers
branch
from
15b5e80 to
3fd3245
Compare
Motivation:
Channels can read
ChannelOptions.maxMessagesPerReadtimes from a socket in each read cycle. They typically re-use the same buffer for each read and rely on it CoWing if necessary. If we read more than once in a cycle then we will always CoW the buffer. Instead of reusing one buffer we can reuse a pool of buffers limited bymaxMessagesPerReadand cycle through each, reducing the chance of CoWing the buffers.Modifications:
NIOPoolableRecvByteBufferAllocatorwhich extendsRecvByteBufferAllocatorto provide the size of the next buffer.NIOPoolableRecvByteBufferAllocatorResults:
Fewer allocations