Implement an argb cursor using the alpha layer.

  The code is <hash>ifdef'ed out by <hash>if SMI_CURSOR_ALPHA_PLANE
because the smi 502 hardware is buggy, but it appears to have worked
for some older hardware revisions (by looking at code available at
ftp.siliconmotion.com.tw).
  Keeping it at least for now, as it serves as a simple way to reproduce
the problems described in smi.h.

1 files changed, 36 insertions, 0 deletions

diff --git a/src/smi.h b/src/smi.h
index eb88b48..66d28df 100644
--- a/src/smi.h
+++ b/src/smi.h
@@ -72,6 +72,42 @@ authorization from the XFree86 Project and Silicon Motion.
 #define SMI_USE_VIDEO		1
 #define SMI_USE_CAPTURE		1
 
+/*
+ *   Leaving attempt implementation of an argb cursor using alpha plane
+ * for the smi 501/502 under this ifdef for now. Maybe it will be fixed
+ * in a subsequent hardware revision.
+ *   The problem is that the alpha plane will only work (that is, become
+ * visible) if alpha_plane_tl is set to top:=0 and left:=0.
+ *   Also, if alpha_plane_br does not match panel dimensions, the alpha
+ * plane will be displayed tilled in the "first" row, with corruption on
+ * on all odd columns.
+ *   Since setting the alpha fb_address works, to implement an argb cursor
+ * using the alpha plane, with the current hardware bugs, it would be
+ * required to:
+ *	o allocate an offscreen area of pSmi->lcdWidth * pSmi->lcdHeight * 2
+ *	o set statically tl/tr to 0,0,pSmi->lcdWidth-1,pSmi->lcdHeight-1
+ *	o use alpha format 3 (argb 4:4:4:4), or alternatively format 1
+ *	  (rgb 5:6:5), and in the last case, a global 50% alpha is the
+ *	  best bet, and for the argb cursors being used at the time of this
+ *	  writing, they look correct, while 100% opaque looks wrong.
+ *	o when positioning the pointer, first erase it from the offscreen
+ *	  area, then repaint it at the proper offset in the alpha offscreen
+ *	  area.
+ *  .... argb software cursor works way better
+ *   (There are some other alternatives, like using 8 bits indexed mode,
+ * but when using a global alpha value, instead of per pixel, most argb
+ * cursors will not look correctly, regardless of the alpha value, that
+ * should be from 50 to 100% transparency).
+ *   But still there would be the problem of memory requiring a 128 bit
+ * alignment, what would require either moving the image in the memory,
+ * and/or some trick with the vsync pixel panning.
+ *
+ *   Until the alpha layer is corrected in some newer revision (or removed?),
+ * it could be used as something like an alternate crt, that happens to be
+ * on top of the panel (and has 16 transparency levels).
+ */
+#define SMI_CURSOR_ALPHA_PLANE	0
+
 /******************************************************************************/
 /*			S T R U C T U R E S				      */
 /******************************************************************************/