Originally posted by uid313
View Post
Announcement
Collapse
No announcement yet.
Linux 3.10 May Have New Multi-Platform Support
Collapse
X
-
-
Originally posted by schmidtbag View PostActually that reminds me of another point that bothers me - for some reason that I'm unaware of, it seems as though most distros use the 3.0.6* (the same version as the Android kernel - coincidence?) or 3.2 kernel. Not only does that force them to maintain a device-specific kernel but it hinders compatibility and performance. The 3.5 kernel has pretty huge performance improvements over 3.2 and older. I feel like if they spent the time to work on the newer kernels, there could be a 1-disk-image-fits-all with better performance.
I don't mean to sound ungrateful for the legitimately hard work linux devs put in for ARM, and like I said there might just be something big I don't know about, but they're making things harder for everyone, including themselves.
Fedora is usually ahead in the game and use recent kernels too.
Gentoo and Arch probably use pretty recent kernels too.
Leave a comment:
-
Originally posted by uid313 View PostKernel 3.7 is quite new and few distributions have shipped since then, you have to wait until next distribution release, many distributions use a 6-month release cycle.
Also in 3.7 it was just the initial commit, it was still immature code, it was the start, the code in 3.10 is the continuation of the work from 3.7.
The code in 3.7 was just a start, it wasn't really ready, it was experimental.
So maybe there will be generic ARM distributions around 6 months after kernel 3.10.
I don't mean to sound ungrateful for the legitimately hard work linux devs put in for ARM, and like I said there might just be something big I don't know about, but they're making things harder for everyone, including themselves.
Leave a comment:
-
Originally posted by schmidtbag View PostFINALLY
I cannot get over how annoying it is to work on ARM dev platforms due to the variety of kernels. The day when I can use the same disk image on 2 different boards (such as OMAP and Exynos) will be a very happy day for me.
What I don't understand though is if this has been around since 3.7, how come I haven't seen any "generic" ARM distros? Most ARM images work the same way, involving a FAT boot partition with uboot (and sometimes the kernel), and a rootfs partition with everything else. There doesn't seem to be nearly as many drivers to worry about with ARM as there are with x86, so I doubt disk images will get too bloated if someone were to create a generic disk image. I think the platform-specific images are nice since they will be optimized and get rid of the extra clutter, but I've come to notice many of them lack drivers that most users would take for granted.
Also in 3.7 it was just the initial commit, it was still immature code, it was the start, the code in 3.10 is the continuation of the work from 3.7.
The code in 3.7 was just a start, it wasn't really ready, it was experimental.
So maybe there will be generic ARM distributions around 6 months after kernel 3.10.
Leave a comment:
-
FINALLY
I cannot get over how annoying it is to work on ARM dev platforms due to the variety of kernels. The day when I can use the same disk image on 2 different boards (such as OMAP and Exynos) will be a very happy day for me.
What I don't understand though is if this has been around since 3.7, how come I haven't seen any "generic" ARM distros? Most ARM images work the same way, involving a FAT boot partition with uboot (and sometimes the kernel), and a rootfs partition with everything else. There doesn't seem to be nearly as many drivers to worry about with ARM as there are with x86, so I doubt disk images will get too bloated if someone were to create a generic disk image. I think the platform-specific images are nice since they will be optimized and get rid of the extra clutter, but I've come to notice many of them lack drivers that most users would take for granted.
Leave a comment:
-
Originally posted by Sonadow View PostWhat about drivers? Each ARM SoC contains a ton of vastly different components like GPU cores, WiFi radios, GSM radios, modems, etc etc...
Even within one SoC family (eg: Exynos) different versions have different components.
Leave a comment:
-
Originally posted by uid313 View PostA x86 image has always run on all x86 machines.
ARM has been a mess, and a image compiled for ARM only worked one one board, not other ARM boards.
So with this, you compile one ARM image that runs on many ARM boards.
Even within one SoC family (eg: Exynos) different versions have different components.
Leave a comment:
-
Originally posted by kaprikawn View PostCan someone a bit more knowledgeable than me clarify what this means please.
Does this mean that I can compile a kernel image from the same source code for both say a desktop x86 machine and one of these SoCs? Or does it mean that the same kernel image will work on both a x86 machine and on one of these SoCs after it's been compiled?
Or have I completely missed the point?
ARM has been a mess, and a image compiled for ARM only worked one one board, not other ARM boards.
So with this, you compile one ARM image that runs on many ARM boards.
Leave a comment:
-
Can someone a bit more knowledgeable than me clarify what this means please.
Does this mean that I can compile a kernel image from the same source code for both say a desktop x86 machine and one of these SoCs? Or does it mean that the same kernel image will work on both a x86 machine and on one of these SoCs after it's been compiled?
Or have I completely missed the point?
Leave a comment:
-
Just a correction in the article, the Nexus 7 has a Tegra 3. I think you meant to say the Nexus 10, which has the Exynos 5250.
Leave a comment:
Leave a comment: