Re: measuring memory usage of Arrow structures

2022-10-28 Thread Yaron Gvili
Yes, byte_size.h is what I was looking for, thanks. Yaron. From: David Li Sent: Friday, October 28, 2022 9:52 AM To: dev@arrow.apache.org Subject: Re: measuring memory usage of Arrow structures This would also be a good cookbook example, so I filed [1]. (Sorry

Re: measuring memory usage of Arrow structures

2022-10-28 Thread David Li
This would also be a good cookbook example, so I filed [1]. (Sorry for the double-email.) [1]: https://github.com/apache/arrow-cookbook/issues/273 On Fri, Oct 28, 2022, at 09:51, David Li wrote: > I think Yaron is looking more for the APIs in arrow/util/byte_size.h > [1] which has utilities for

Re: measuring memory usage of Arrow structures

2022-10-28 Thread David Li
I think Yaron is looking more for the APIs in arrow/util/byte_size.h [1] which has utilities for estimating the size of the buffers in an array or record batch. (I'm assuming you're not concerned with the size of the C++ structures themselves.) [1]: https://github.com/apache/arrow/blob/master/c

Re: measuring memory usage of Arrow structures

2022-10-28 Thread Rok Mihevc
Hey Yaron, If you're using jemalloc you can use jemalloc_get_stat [1] to monitor total memory allocation. Another option would be LoggingMemoryPool, see tests on possible usage [2]. Rok [1] https://github.com/apache/arrow/blob/master/cpp/src/arrow/memory_pool_jemalloc.cc#L157 [2] https://github.