[Linux-fbdev-devel] [PATCH] radeonfb: 16bpp accel broken. a work-around

[David E. <eg...@ha...>]

Dear Andrew, 

I've tracked Jurriaan's radeonfb problem down to copyarea() malfunctioning 
in 16bpp mode.  Here's a hack to just do a software copyarea() in 16bpp mode.
No more screen garbage, but then it's slow :-/
for instance, cat' ing the MAINTAINERS file:
   8bpp -  ~1.5 sec
  16bpp - ~19 sec  (with yucky fix)
  32bpp -  ~6 sec

-dte


radeonfb: 16bpp work-around - copyarea() doesn't work

Signed-off-by: David Eger <eg...@ha...>

diff -Nru a/drivers/video/aty/radeon_accel.c b/drivers/video/aty/radeon_accel.c
--- a/drivers/video/aty/radeon_accel.c	2004-06-25 09:15:24 +02:00
+++ b/drivers/video/aty/radeon_accel.c	2004-06-25 09:15:24 +02:00
@@ -99,7 +99,9 @@
   
 	if (info->state != FBINFO_STATE_RUNNING)
 		return;
-	if (radeon_accel_disabled()) {
+	if (radeon_accel_disabled() 
+		|| info->var.bits_per_pixel == 16) { 
+		/* hack alert: why is 16bpp broken? dunno */
 		cfb_copyarea(info, area);
 		return;
 	}
diff -Nru a/drivers/video/aty/radeon_base.c b/drivers/video/aty/radeon_base.c
--- a/drivers/video/aty/radeon_base.c	2004-06-25 09:15:24 +02:00
+++ b/drivers/video/aty/radeon_base.c	2004-06-25 09:15:24 +02:00
@@ -1573,6 +1573,12 @@
 	}
 #endif
 
+	/* for some reason, copyarea is broken at 16bpp... :-/ */
+	if (mode->bits_per_pixel == 16)
+		info->flags &= ~FBINFO_HWACCEL_COPYAREA;
+	else
+		info->flags |= FBINFO_HWACCEL_COPYAREA;
+	
 	/* Clear surface registers */
 	for (i=0; i<8; i++) {
 		newmode.surf_lower_bound[i] = 0;