Hey all, When putting my porter board under a bit of load with linux-next from earlier this month (20160405), I'm seeing occasional glitches on the HDMI output. These seems to be triggered by stressing the memory. A simple test-case running weston (with the pixman renderer), weston terminal in combination with running 2 instances of the stream memory benchmark in the background triggers the issue. Video here: http://people.collabora.co.uk/~sjoerd/20160419_001.mp4 The same test on a 3.10.31 based kernel from renesas-backport.git has no issues (image is completely stable on the HDMI output). Our theory thusfar is that there is a buffer/fifo underflow while streaming the pixels out to the encoder/display. Unfortunately I couldn't find anything thusfar in either the public manuals or the old renesas-backports tree related to e.g. setting up fifo depths, memory priorities etc. Hopefully someone on the list has a good suggestion of where to look or what could be the issue here. As extra information, i've been putting together some pieces to run the proprietary IMG driver against that same kernel. Which seems to work well enough, apart from triggering what appears to be the same issue but worse. A video of weston running glmark2 with the pvr driver on this kernel can be found here (or as one of my collegues calls it "light speed horse": https://people.collabora.com/~sjoerd/20160418_001.mp4 -- Sjoerd Simons Collabora Ltd.
