Announcement

Collapse
No announcement yet.

Building The Default x86_64 Linux Kernel In Just 16 Seconds

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • discordian
    replied
    Originally posted by Michael View Post
    It's not a matter of "not heard of", but rather trying to be realistic - how many people actually build in tmpfs?
    You can build outside tmpfs?

    All buildslaves I setup build in tmpfs, debian packages can be built in tmpfs (and their wiki documents how: https://wiki.debian.org/sbuild#sbuild_overlays_in_tmpfs). With eclipse I add a link to /tmp/x.y.z/build (/tmp - as you might have gueesed - is a tmpfs), eclipse will automatically create the folder is missing.
    The only downside would be missing RAM, if the linker step needs alot of it...
    Last edited by discordian; 14 August 2019, 07:48 AM.

    Leave a comment:


  • Michael
    replied
    Originally posted by caligula View Post

    Can't believe Michael hasn't heard about building applications in RAM. It didn't make sense with older machines, but now even most desktops have sufficient RAM for building large projects such as the kernel in RAM. The sources require around 1 GB, and the resulting binaries also need 1-2 GB. While compiling, each gcc thread probably uses few hundred megabytes at most. So, 32 GB might be enough for building on a 64 core system.
    It's not a matter of "not heard of", but rather trying to be realistic - how many people actually build in tmpfs?

    Leave a comment:


  • caligula
    replied
    Originally posted by birdie View Post

    What about building in tmpfs? ;-)
    Can't believe Michael hasn't heard about building applications in RAM. It didn't make sense with older machines, but now even most desktops have sufficient RAM for building large projects such as the kernel in RAM. The sources require around 1 GB, and the resulting binaries also need 1-2 GB. While compiling, each gcc thread probably uses few hundred megabytes at most. So, 32 GB might be enough for building on a 64 core system.

    Leave a comment:


  • programmerjake
    replied
    Originally posted by discordian View Post
    Now you can checkout the Linux master branch, compile the kernel and kexec into it from initrd. All while still booting faster than many systems.
    Rolling release was yesterday, now it's run-from-source!
    tcc was able to boot from the linux source about 15 years ago: https://bellard.org/tcc/tccboot.html

    Leave a comment:


  • rbanffy
    replied
    Has anyone tried that with a last-generation Xeon Phi? 72 relatively slow cores but 4 threads per core and 16 GB of HBM should give some interesting results.

    Leave a comment:


  • tchiwam
    replied
    And redirect all outputs to /dev/null or a file on tmpfs

    Leave a comment:


  • darkbasic
    replied
    Optane vs tmpfs would be REALLY interesting to see!

    Leave a comment:


  • discordian
    replied
    Originally posted by AsuMagic View Post
    I'm almost certain there is like a category of languages which do something like that... What's the name again?
    Binaries! You edit their source in a Hexeditor.

    Leave a comment:


  • AsuMagic
    replied
    Originally posted by discordian View Post
    Now you can checkout the Linux master branch, compile the kernel and kexec into it from initrd. All while still booting faster than many systems.
    Rolling release was yesterday, now it's run-from-source!
    I'm almost certain there is like a category of languages which do something like that... What's the name again?

    Leave a comment:


  • discordian
    replied
    Now you can checkout the Linux master branch, compile the kernel and kexec into it from initrd. All while still booting faster than many systems.
    Rolling release was yesterday, now it's run-from-source!

    Leave a comment:

Working...
X