Tim Armstrong has posted comments on this change. Change subject: IMPALA-3201: headers and reservation logic for new buffer pool ......................................................................
Patch Set 22: Right, I think a buffer internally should just be a pointer and a length. It would be easy to expose an allocate/free interface for raw buffers to clients, but I don't think it's desirable. A lot of functionality for debuggability is tied to the page object internally. E.g. enumerating all of the existing pages (see DebugString()). If we don't have any internal tracking of buffers, then we lose that ability to see which client allocated which memory. I think there are a lot of interesting things we could do with that info. We also talked about having additional functionality exposed in the buffer interface beyond what you could do with a raw buffer (reference counting, etc), so I think there's actually a difference between the "raw buffers" at the lowest layer of the system and the "buffer handles" you'd expose to clients. The exposed buffer interface would be more parallel to the page interface, rather than a lower layer. -- To view, visit http://gerrit.cloudera.org:8080/2569 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I35cc89e863efb4cc506657bfdaaaf633a10bbab6 Gerrit-PatchSet: 22 Gerrit-Project: Impala Gerrit-Branch: cdh5-trunk Gerrit-Owner: Tim Armstrong <[email protected]> Gerrit-Reviewer: Dan Hecht <[email protected]> Gerrit-Reviewer: Marcel Kornacker <[email protected]> Gerrit-Reviewer: Michael Ho Gerrit-Reviewer: Mostafa Mokhtar <[email protected]> Gerrit-Reviewer: Tim Armstrong <[email protected]> Gerrit-HasComments: No
