901211 - Don't use static page size on ia64, sparc and mips

Status: RESOLVED FIXED

(Core :: Memory Allocator, defect)

Description

[Mike Hommey [:glandium]]

ia64, sparc and mips have non 4k page sizes, and even a page size that varies depending on the kernel or maybe the cpu model.

Comment 1

[Mike Hommey [:glandium]]

Attachment: Don't use static page size on ia64, sparc and mips

Comment 2

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/ec6bcb45443c

Comment 3

[Mike Hommey [:glandium]]

Comment on attachment 785361 [details] [diff] [review]
Don't use static page size on ia64, sparc and mips

[Approval Request Comment]
Requesting for beta because i'd like this in ESR24 so that I don't have to carry the patch around for the ESR.
User impact if declined: Runtime failure because of page size size not matching what jemalloc expects on ia64, sparc and mips.
Testing completed (on m-c, etc.): Tested on a ia64, sparc and mips.
Risk to taking this patch (and alternatives if risky): NPOTB.
String or IDL/UUID changes made by this patch: None

Comment 4

[Ryan VanderMeulen (PTO, back July 27)]

https://hg.mozilla.org/releases/mozilla-beta/rev/fc0020792338

Comment 5

[Benjamin Herrenschmidt]

This is the case on powerpc as well. Please add it to the list.

Note that this is a terrible design decision to begin with. I sincerely hope
the code can be reworked to no longer make that assumption (or at least stick
to something like "maximum" page size if that can help).

Comment 6

[Justin Lebar (not reading bugmail)]

FWIW we saw a significant improvement on our whole-browser benchmarks when we switched to using a static page size.  Trading that for breakage on our tier-three platforms doesn't seem "terrible" to me.  I don't think we have any plans to get rid of this assumption for platforms where it's a correct assumption.

Comment 7

[u279076]

Assuming no verification needed here. Please add the verifyme keyword and remove the [qa-] whiteboard tag to request verification.