Community
Participate
Working Groups
The idea is to create a visualizer, based on the Multicore Visualizer, but adapted to visualize an Epiphany chip ( http://www.adapteva.com/epiphany-multicore-intellectual-property/ ). ATM, the Epiphany chip is available as part of the parallella board (http://www.parallella.org/) One interesting aspect of the Epiphany chip is that its multiple eCores have to be displayed in a pre-defined configuration - that closely reflect the physical layout of the chip - since their position relative to the other eCores matter. So the Epiphany Visualizer had to take that factor into account and adapt its display algorithm. In contrast, the cores of an Intel or AMD CPU do not have this property.
Here is a preliminary version of the Epiphany Visualizer. It's not yet meant to be reviewed, as is. https://git.eclipse.org/r/#/c/24473/ The visualizer will need to be completed when the e-server (GDB server / stub for Epiphany) is completed / improved. In the meantime, some of the responses that would come from e-server are stubbed / faked. This means that the visualizer can be tried without Epiphany hardware, just by starting any GDB session. Here are a few pointers: - to switch between a simulated 16/64 eCores Epiphany chip, select one of the two versions of the constant "TARGET_TYPE" and comment the other one, near the top of EpiphanyVisualizer.java - the model is being filled-in with bogus information ATM. You can change the model info in EpiphanyVisualizer#getVisualizerModel(). Once e-server can provide this information, we will update so that the model is filled-in with "real" data. - The Epiphany visualizer shows just the eCores by Default. To also see the eMesh network, you need to use the context menu on the visualizer to enable load monitoring. Note: the load information is obtained from the e-server using a proprietary (for now) "info os load" command, that will not be understood by a vanilla GDB, and will probably cause an exception on non-Epiphany hardware or older Epiphany e-server versions. We plan to take come of the ideas from this code and integrate them to the current Visualizer framework. Stay tuned.
Re-posted the code here (I forgot to have a changeid in the commit message) https://git.eclipse.org/r/25133
I have updated the Epiphany visualizer code to use the new Graphic object container currently under review; see Bug 432854 However, when trying to push the new version, I have run into what seems to be this Gerrit issue: https://code.google.com/p/gerrit/issues/detail?id=2296 So instead I'll attach a patch to this bug. It's really not meant for review at this point, in any case, but just to give the occasion to try the new code. Apply the patch over this: https://git.eclipse.org/r/#/c/25133/3 cd [...]/org.eclipse.cdt git am < bug_432055.patch Then import the new plugin: Import -> git -> Projects from git -> Existing local repository -> org.eclipse.cdt -> dsf-gdb -> org.eclipse.cdt.dsf.gdb.multicorevisualizer.epiphany.ui
Created attachment 242160 [details] Version fo Epiphany visualizer that uses the new GraphicObjectContainer class
Oops - correction: Apply the patch over this: https://git.eclipse.org/r/#/c/25079/3
Hi, I was able to push a new version of the Epiphany visualizer, using the git command-line client instead of egit. It's rebased on the latest patch for Bug 432854. https://git.eclipse.org/r/#/c/25133/4 Regards, Marc