[SERVER-18417] Provide packages with AVX2 architecture Created: 11/May/15  Updated: 27/Oct/23  Resolved: 27/Oct/23

Status: Closed
Project: Core Server
Component/s: Build, Packaging, Performance
Affects Version/s: None
Fix Version/s: None

Type: New Feature Priority: Major - P3
Reporter: Chad Kreimendahl Assignee: [DO NOT ASSIGN] Backlog - Server Development Platform Team (SDP) (Inactive)
Resolution: Won't Do Votes: 1
Labels: build-later
Remaining Estimate: Not Specified
Time Spent: Not Specified
Original Estimate: Not Specified

Assigned Teams:
Server Development Platform
Participants:

 Description   

The Haswell architecture has a new AVX2 instruction set that's supported by more current versions of gcc and visual-studio compilers. In-memory databases are known to benefit substantially from the incorporation of these instruction sets.

It would be great to have linux and windows packages with these already compiled. Notably aptitude and windows packages.

https://software.intel.com/en-us/articles/how-intel-avx2-improves-performance-on-server-applications



 Comments   
Comment by Andrew Morrow (Inactive) [ 23/Jun/20 ]

corporate.piyush@gmail.com - This ticket is currently slotted into a project that is approved, but not yet moving forward due to higher priority work. We do hope to get to it in the near future. While it is certainly possible to build the source code with flags that would enable this, I cannot recommend doing so as it will be difficult to ensure correctness without extensive regression testing.

Comment by Piyush Katariya [ 20/Jun/20 ]

Looks like this issue is out of attention. Please re-consider this issue.

 

It will help tremendously for querying larger dataset through aggregation pipeline or otherwise.   

 

Or Is there a way to compile the mongodb source code to include this feature and let compiler do the deep autovectorization of the code ?

Generated at Thu Feb 08 03:47:37 UTC 2024 using Jira 9.7.1#970001-sha1:2222b88b221c4928ef0de3161136cc90c8356a66.