zeroshade commented on a change in pull request #11206: URL: https://github.com/apache/arrow/pull/11206#discussion_r718817052
########## File path: go/arrow/memory/cgo_allocator.go ########## @@ -0,0 +1,108 @@ +// Licensed to the Apache Software Foundation (ASF) under one +// or more contributor license agreements. See the NOTICE file +// distributed with this work for additional information +// regarding copyright ownership. The ASF licenses this file +// to you under the Apache License, Version 2.0 (the +// "License"); you may not use this file except in compliance +// with the License. You may obtain a copy of the License at +// +// http://www.apache.org/licenses/LICENSE-2.0 +// +// Unless required by applicable law or agreed to in writing, software +// distributed under the License is distributed on an "AS IS" BASIS, +// WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. +// See the License for the specific language governing permissions and +// limitations under the License. + +// +build cgo +// +build ccalloc + +package memory + +import ( + "runtime" + + cga "github.com/apache/arrow/go/arrow/memory/internal/cgoalloc" +) + +// CgoArrowAllocator is an allocator which exposes the C++ memory pool class +// from the Arrow C++ Library as an allocator for memory buffers to use in Go. +// The build tag 'ccalloc' must be used in order to include it as it requires +// linking against the arrow library. +// +// The primary reason to use this would be as an allocator when dealing with +// exporting data across the cdata interface in order to ensure that the memory +// is allocated safely on the C side so it can be held on the CGO side beyond +// the context of a single function call. If the memory in use isn't allocated +// on the C side, then it is not safe for any pointers to data to be held outside +// of Go beyond the context of a single Cgo function call as it will be invisible +// to the Go garbage collector and could potentially get moved without being updated. Review comment: Yea, I figured it was simplest to use the memory pool that already exists in the libarrow library (as I use the libarrow library in my future changes to connect to the compute API), but I could potentially use something like https://github.com/spinlock/jemalloc-go to just include jemalloc as a go-getable dependency. Which i'll play around with and see how it does. I liked the libarrrow memory pool because of the consistency it provided and feature handling such as the logging proxy and the tracking of how many bytes had been allocated in it, but it is likely pretty easy to do similar functionality and use jemalloc directly. As long as consumers are using Reserve and otherwise pre-allocating memory, it would cut down the number of cgo calls to mitigate performance hits. Alternately I could pursue looking into a slab style allocator that manages much larger chunks of memory that it hands out as a way to amortize the number of calls, but that's a separate thing to look into later. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: github-unsubscr...@arrow.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
