The previous code would emit a full set of state during the first EmitState on

a new cmdbuf, to ensure that state wasn't lost across UNLOCK/LOCK pairs (in the
case of context switching).  This was rather inefficient.  Instead, after
flushing a cmdbuf, mark the state as needing to be saved on unlock.  Then, at
the beginning of flushing a cmdbuf, if we actually have lost the context, go
back and emit a new cmdbuf with the full set of state, before continuing with
the cmdbuf flush.  Provides a 10-15% improvement in ipers performance in my
tests, along with other apps.

Tested with:	ipers, glxgears, quake3

1 files changed, 7 insertions, 3 deletions

diff --git a/src/mesa/drivers/dri/radeon/radeon_ioctl.h b/src/mesa/drivers/dri/radeon/radeon_ioctl.h
index 695eb5770b..7ad4f3a932 100644
--- a/src/mesa/drivers/dri/radeon/radeon_ioctl.h
+++ b/src/mesa/drivers/dri/radeon/radeon_ioctl.h
@@ -104,6 +104,9 @@ extern void radeonWaitForVBlank( radeonContextPtr rmesa );
 extern void radeonInitIoctlFuncs( GLcontext *ctx );
 extern void radeonGetAllParams( radeonContextPtr rmesa );
 
+extern void radeonSaveHwState( radeonContextPtr rmesa );
+extern void radeonSetUpAtomList( radeonContextPtr rmesa );
+
 /* radeon_compat.c:
  */
 extern void radeonCompatEmitPrimitive( radeonContextPtr rmesa,
@@ -111,7 +114,6 @@ extern void radeonCompatEmitPrimitive( radeonContextPtr rmesa,
 				       GLuint hw_primitive,
 				       GLuint nrverts );
 
-
 /* ================================================================
  * Helper macros:
  */
@@ -130,7 +132,8 @@ do {						\
 #define RADEON_STATECHANGE( rmesa, ATOM )			\
 do {								\
    RADEON_NEWPRIM( rmesa );					\
-   move_to_head( &(rmesa->hw.dirty), &(rmesa->hw.ATOM));	\
+   rmesa->hw.ATOM.dirty = GL_TRUE;				\
+   rmesa->hw.is_dirty = GL_TRUE;				\
 } while (0)
 
 #define RADEON_DB_STATE( ATOM )			        \
@@ -144,7 +147,8 @@ static __inline int RADEON_DB_STATECHANGE(
    if (memcmp(atom->cmd, atom->lastcmd, atom->cmd_size*4)) {
       int *tmp;
       RADEON_NEWPRIM( rmesa );
-      move_to_head( &(rmesa->hw.dirty), atom );
+      atom->dirty = GL_TRUE;
+      rmesa->hw.is_dirty = GL_TRUE;
       tmp = atom->cmd; 
       atom->cmd = atom->lastcmd;
       atom->lastcmd = tmp;