IPC in Android X-86

50 views
Skip to first unread message

Arghyadip Paul

unread,
Apr 25, 2012, 1:59:35 PM4/25/12
to andro...@googlegroups.com
Hi All,

I am new to this group but I feel awesome after getting my first hands on with the android x 86.

Currently I am thinking how the different IPC Mechanisms and Shared Memory implementation exists in Android X-86. From the http://elinux.org/Android_Kernel_Features#List_of_kernel_features_unique_to_Android its evident that mainstream android deprecated the standard POSIX SHM and implemented their own ashmem library. Their main motivation behind doing this is to support an ARM based environment which Linux 2.6 kernel was not aimed for.

But as Android X86's main motivation is to support android on X 86 based environment , I wish to know if we can again port back POSIX SHM and do a performance evaluation for the same to see if there is any improvement.

Best,
Arghyadip
Graduate Student,
UCSB, Goleta, CA

rvdb

unread,
Apr 26, 2012, 4:12:29 AM4/26/12
to Android-x86
Well, I believe it is a bit more complex.
The main reason google doesn't want to use ipc/shm ) I think, is that
android is build
to "kill"any task if it likes to do so.
IPC and shm memory will not be freeed, so memory can become a mess.

If an application is written "nice" it can reuse this memory, but if
not............
I do not believe performance would be the main reason to use posix
memory, but
a lot of x86 applications/drivers use shm, so porting is much easier
if we could use it.

For the crystalhd driver (which I believe is written nice in memory
management :-) ) I
used shm. (Just a few stub routines, since the kernel will handle all
this).
It is working oke, but do not tell google.

Greetings Rene



On Apr 25, 7:59 pm, Arghyadip Paul <arghya...@gmail.com> wrote:
> Hi All,
>
> I am new to this group but I feel awesome after getting my first hands on
> with the android x 86.
>
> Currently I am thinking how the different IPC Mechanisms and Shared Memory
> implementation exists in Android X-86. From thehttp://elinux.org/Android_Kernel_Features#List_of_kernel_features_uni...its

Arghyadip Paul

unread,
Apr 30, 2012, 4:27:54 AM4/30/12
to andro...@googlegroups.com
Thanks for your reply.


On Thursday, April 26, 2012 1:12:29 AM UTC-7, rvdb wrote:
Well, I believe it is a bit more complex.
The main reason google doesn't want to use ipc/shm ) I think, is that
android is build
to "kill"any task if it likes to do so.
IPC and shm memory will not be freeed, so memory can become a mess.

If an application is written "nice" it can reuse this memory, but if
not............
I do not believe performance would be the main reason to use posix
memory, but
a lot of x86 applications/drivers use shm, so porting is much easier
if we could use it.

For the crystalhd driver (which I believe is written nice in memory
management :-) ) I
used shm. (Just a few stub routines, since the kernel will handle all
this).
It is working oke, but do not tell google.
Can you tell me how can I get this version of the code where stubs are written for crystalHD driver which uses Posix SHM methods? does the mainstream Android X86 branch contains this? 

Greetings Rene

Best,
Arghyadip 
Reply all
Reply to author
Forward
0 new messages