We have a DM3730 connected with an LCD Display. Unfortunately, the display is very dark and my experiments suggest that it still needs to be gamma corrected. Reading the TI documentation I found that the Display Subsystem has a Palette/gamma lookup-table. If I understand correctly, I have to create a lookup-table in main memory, pass its address into the DISPC_GFX_TABLE_BA (0x4805 04B8) Register (translate virtual address to physical address?) and then the lookup-table is loaded into the internal DSS memory.
We run the Linux Kernel 2.6.32 and I found the definition "#define DISPC_GFX_TABLE_BA 0x00B8" in the Linux kernel source tree under drivers/video/omap/dispc.c. But it seems like that the register is not actually used anywhere. Are there any examples of using the lookup-table and is this even the best approach to correct the color of our display?
Thank you
(Reposted here, because I accidentally posted this in the DM3xx Forum)
EDIT for clarification: I am trying to transfer a gamma table from user space into the Palette/gamma lookup-table of the Display Subsystem. I found a related post (http://e2e.ti.com/support/dsp/omap_applications_processors/f/447/t/59105.aspx) and there, it is suggested that I set the lookup-table from the application layer.
Quote (emphasis mine): "Next, declare an array of 256 unsigned integers and populate it with the RGB values you want for each color index entry. Set DISPC_GFX_TABLE_BA to the physical address of the previously configured array (Note 1, this needs the physical address and not the virtual address so you will likely need to map it back. Note 2, this needs to be a 32 bit aligned address so depending on how the array is created it may be necessary to create an array of char slightly larger and populate the array such that the start is on a 32 bit boundary)"
But how can I map a virtual address from user space back to a physical address? He suggests to look at mem_util (https://gstreamer.ti.com/gf/project/am_sysinfo/scmsvn/?action=browse&path=%2Ftrunk%2Fmem_util%2Fmem_util.c&revision=2&sortby=log&view=markup), but this maps physical addresses to virtual addresses, as far as I can see, and not the other way around.
