Re: [Intel-gfx] Workaround for flicker with panning on the i830 - found a way for tiled displays
Hi Daniel, Gosh, should have read the code more closely. We have a totally botched wm setup on i830M - the watermark code for the 2nd pipe is just not there! Well, nice try, but no cigar. (-: That's actually much worse than before. The display is now unstable on *both* the internal and external display, and inspecting the FW_BLC register, it is completely off. The current code leaves it at 0x1050101, but it should be at least 0x1060106, that is, the watermark needs to be higher, not lower. I still don't understand why a higher latency causes a lower watermark, but maybe things work different than in my mental model. Greetings, Thomas ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] Workaround for flicker with panning on the i830 - found a way for tiled displays
Am 15.11.2013 16:41, schrieb Daniel Vetter: Gosh, should have read the code more closely. We have a totally botched wm setup on i830M - the watermark code for the 2nd pipe is just not there! (-: Guess that explains something. Just disregard my earlier patch, simply superfluous. In the meantime, I recompiled the code and *decreased* the latency from 5000 to 3000, then getting a stable image, even the boot console is then stable (has never been before). Its still off by half a screen, but no longer flickering left and right. However, what I do not understand about the watermark computation is that a *lower* latency results in a *higher* number of entries in the FIFO. Shouldn't this quite the reverse? In specific, I do not understand the subtraction in intel_calculate_wm, line 1058. Anyhow, I let you proceed with the patch and I'm ready and happy to test it. Greetings, Thomas ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
Re: [Intel-gfx] Workaround for flicker with panning on the i830 - found a way for tiled displays
On Fri, Nov 15, 2013 at 02:16:11PM +0100, Thomas Richter wrote: > Hi Daniel, hi others, > > did even more experiments. I guess I understand now better. Indeed, > the trouble seems to be the watermark levels. I played more > with all that, and the real culprit seems to be the FW_BLC register > controlling the watermarks. > > On the i830 with the current settings, it is defined to be 0x1080304 > which sets the watermark a bit too low. If I set it to > 0x1080306 instead, I get a stable display in all panning positions > (hurray!). > > I would like to fix this, but I guess I would need to understand the > logic a little bit better. At the time being, you probably better > put the linear frame buffer workaround on hold, it looks I really > got something here. Gosh, should have read the code more closely. We have a totally botched wm setup on i830M - the watermark code for the 2nd pipe is just not there! I'll try to wip up a patch to fix this. -Daniel -- Daniel Vetter Software Engineer, Intel Corporation +41 (0) 79 365 57 48 - http://blog.ffwll.ch ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
[Intel-gfx] Workaround for flicker with panning on the i830 - found a way for tiled displays
Hi Daniel, hi others, did even more experiments. I guess I understand now better. Indeed, the trouble seems to be the watermark levels. I played more with all that, and the real culprit seems to be the FW_BLC register controlling the watermarks. On the i830 with the current settings, it is defined to be 0x1080304 which sets the watermark a bit too low. If I set it to 0x1080306 instead, I get a stable display in all panning positions (hurray!). I would like to fix this, but I guess I would need to understand the logic a little bit better. At the time being, you probably better put the linear frame buffer workaround on hold, it looks I really got something here. Greetings, Thomas ___ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx
