10:52 <DPA> Regarding the glamor + linear issue (#3721), I wasn't able to figure out how to trigger the copy from the shadow bo yet. I don't really understand how it all works yet either.

10:52 <DPA> But even if I could figure that part out, could that become any good performance and memory bandwidth wise?

10:52 <DPA> I tried calling "pctx->set_framebuffer_state(pctx, &ctx->framebuffer_s);" in etna_flush, but it doesn't end up doing anything.

10:52 <DPA> I wonder, wouldn't it make more sense to move the whole shadow bo to linear copy stuff into the kernel, and do the copying just when the linear bo/next frame is needed by mxsfb?

10:59 <lynxeye> DPA: The kernel can not do a efficient GPU copy. The kernel in fact doesn't even know the layout information of the BO.

11:00 <lynxeye> What you want to do is exactly the other way around: remember cbuf resources set in set_framebuffer if they aren't marked as explicit flush and then call the flush_resource on a ctx flush on those remembered resources

11:36 <DPA> Ok, thanks, I'll try that.