<html> <head> <meta content="text/html; charset=UTF-8" http-equiv="Content-Type"> </head> <body bgcolor="#FFFFFF" text="#000000"> Hi Russell, <div class="moz-cite-prefix">在 2015/3/31 19:57, Russell King - ARM Linux 写道: </div> <blockquote cite="mid:20150331115718.GS24899@n2100.arm.linux.org.uk" type="cite"> <pre wrap="">On Tue, Mar 31, 2015 at 05:02:20AM -0400, Yang Kuankuan wrote: </pre> <blockquote type="cite"> <pre wrap="">Besides, as you are going an dw_hdmi cleanups, I want to point another bugs that relate to the HDMI CTS test. There are somethings wrong with General Control Pack, as for now the encoder color depth is 8-bit packing mode, so the color depth only support 24 bits per pixel video, In this case the CD filed in GCP should set to "Color Depth Not indicated". </pre> </blockquote> <pre wrap=""> I'm not sure I follow. According to the iMX6 documentation, setting the CD field to either 0000 or 0100 indicates that the color depth is 24 bits per pixel, 8 bits per component, 8 bit packing mode - there's no documented difference between these. </pre> </blockquote> Yeah, that is the point. Though there's no designedware datasheet difference between "0000b" & 0100b" in vp_pr_cd, but I think the color_depth in vp_pr_cd register are mapping to CD0-CD3 filed in GCP packet, and acutally "0000b" & "0100b" are differnent in CD filed ("0000b Color Depth not indicated" & "0100b 24 bit per pixel"). If the CD filed is non-zero, that indicate we are support depth color mode(but I think the depth color mode need at least 36bpp). Besides as the HDMI Specification descripte, "If the Sink dose not receive a GCP with non-zero CD from more than 4 consecutive video fields, it should exit deep color mode and reverting to 24-bit color"(24-bit color is default). In the end I think if we only output 24-bit color, we should make the CD field to zero :) <blockquote cite="mid:20150331115718.GS24899@n2100.arm.linux.org.uk" type="cite"> <pre wrap=""> Are you saying that you wish to pass something other than 24 bpp video to your HDMI encoder? </pre> <blockquote type="cite"> <pre wrap="">In the end we should keep the *csc_color_depth(HDMI_CSC_SCALE)* & *color_depth(HDMI_VP_PR_CD)* to zero, code should modify like this GCP would test pass: </pre> </blockquote> <pre wrap=""> >From what you're describing, you want CD field = 0 and CSC_SCALE = 0. It looks like hdmi_video_packetize() will set the CD field to zero if hdmi_data.enc_color_depth = 0, but that would cause hdmi_video_sample() and hdmi_video_csc() to fail. Maybe those two functions should be fixed to accept a color depth of zero, and maybe you need to set enc_color_depth to 0? Interestingly, HDMI_CSC_SCALE_CSC_COLORDE_PTH_24BPP is defined to be zero, but again, in the iMX6 data, it could take a value of either 0x00 or 0x40. I think hdmi_video_csc() should set this to 0x40 if hdmi_data.enc_color_depth = 0, and 0x40 for hdmi_data.enc_color_depth = 8. </pre> </blockquote> Agree! If we only output 24-bit color, we should let the hdmi_data.enc_color_depth = 0, and hdmi_video_csc() & hdmi_video_packetize() should handle the "enc_color_depth = 0". Best regards. Yakir Yang <blockquote cite="mid:20150331115718.GS24899@n2100.arm.linux.org.uk" type="cite"> <pre wrap=""> </pre> </blockquote> </body> </html>