Announcement

Collapse
No announcement yet.

Hawaii On Wayland's Weston Shows Low RAM Usage

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

  • Hawaii On Wayland's Weston Shows Low RAM Usage

    Phoronix: Hawaii On Wayland's Weston Shows Low RAM Usage

    At the beginning of the day I wrote about the Hawaii desktop environment having a new Weston Shell release. This Qt5-based desktop is not only nice for showing off the next-generation Linux display stack, but with it also comes a dramatic reduction in system memory usage...

    Phoronix, Linux Hardware Reviews, Linux hardware benchmarks, Linux server benchmarks, Linux benchmarking, Desktop Linux, Linux performance, Open Source graphics, Linux How To, Ubuntu benchmarks, Ubuntu hardware, Phoronix Test Suite

  • #2
    Even still, it's quite impressive. I was wondering if we'd ever return to those days, but at this point it looks like these compositors could run on something like the Raspberry Pi with ease. When we're optimizing our stacks for underpowered environments, the high performance computers we've found to be commonplace will be able to take on a lot more work.

    I'm definitely exciting to see direct comparisons of GTK and Qt themselves, sans environment, so we can see just how much more efficient Wayland allows traditional applications to be. Seeing real improvements in rendering performance in an app like Blender or GIMP, for example, could draw a more compelling case than pure RAM usage.

    Either way, it looks like Wayland's approach is already a big improvement, and this is only the beginning. I'm very optimistic.

    Comment


    • #3
      Originally posted by scionicspectre View Post
      Even still, it's quite impressive. I was wondering if we'd ever return to those days, but at this point it looks like these compositors could run on something like the Raspberry Pi with ease. When we're optimizing our stacks for underpowered environments, the high performance computers we've found to be commonplace will be able to take on a lot more work.
      Lower ram usage doesn't always mean lower cpu / gpu usage. Also, I wonder how much is loaded in VRAM for each case.

      Comment


      • #4
        Originally posted by doom_Oo7 View Post
        Lower ram usage doesn't always mean lower cpu / gpu usage. Also, I wonder how much is loaded in VRAM for each case.
        That is the question. I'm simply guessing that it's much easier to improve on our current technology with a more efficient display architecture which is ready to be iterated on extensively. Even if you discount the technical merits entirely and assume Wayland and X11 are exactly on par with each other, the ease of development itself will make Wayland more likely for improvements.

        Of course, we have yet to find out. I really want to see some good, scientific benchmarks when more comparable environments have arrived.

        Comment


        • #5
          I've contacted the Maui Project about the lack of 32 bit builds for their LiveCD, and they say as long as they have to use QT5 git, offering a 32 bit version would be rather labor intensive. I'm hoping that several months down the line perhaps, if they can use a regular Qt5 release, maybe we could get a 32 bit version to take advantage of the decreased memory footprint on older machines..

          Comment


          • #6
            Originally posted by doom_Oo7 View Post
            Lower ram usage doesn't always mean lower cpu / gpu usage. Also, I wonder how much is loaded in VRAM for each case.
            Yes, good point. I don't know the details, but I think Xorg still stores all its pixmaps in normal RAM (correct me if I'm wrong?). With Weston and the DRM backend,
            those drawing buffers would all exist directly in VRAM, which would inflate the visible numbers quite a bit. Too bad there's no standard way to measure VRAM usage..
            (aside from some vague GL extensions maybe)

            Comment


            • #7
              If you read the process memory usage correctly, X uses less memory for itself than a single wayland client and weston in particular (just by virtue of those each loading LLVM.so and that by itself using more private memory in each instance than X). Not quite the victory the author was intending to portray.

              Comment


              • #8
                The RAM usage is hardly relevant for most people now, what's far more interesting is how efficiently it's using the RAM it allocates for itself.

                Comment


                • #9
                  Originally posted by scottishduck View Post
                  The RAM usage is hardly relevant for most people now, what's far more interesting is how efficiently it's using the RAM it allocates for itself.
                  You don't use a beagleboard or a raspberry pi, do you ? =p

                  Comment


                  • #10
                    Originally posted by doom_Oo7 View Post
                    Lower ram usage doesn't always mean lower cpu / gpu usage. Also, I wonder how much is loaded in VRAM for each case.
                    Remember you have to use TOP to see that
                    "But watching top suggests that both Xorg and Compiz are using less memory and fewer CPU cycles under Mir than they were with X handling the hardware directly"

                    Comment

                    Working...
                    X