intel: restore old vertex submit paths for i8xx hardware.

Intel docs state that only 830/845 have VBOs, 855/865 don't. So
lets just not use them on i8xx at all.

This restores the old pre-vbo code and uses it on all 8xx hw.

6 files changed, 124 insertions, 6 deletions

diff --git a/src/mesa/drivers/dri/i915/intel_render.c b/src/mesa/drivers/dri/i915/intel_render.c
index 467abe4d7c..410052b3c2 100644
--- a/src/mesa/drivers/dri/i915/intel_render.c
+++ b/src/mesa/drivers/dri/i915/intel_render.c
@@ -117,6 +117,26 @@ intelDmaPrimitive(struct intel_context *intel, GLenum prim)
    intel_set_prim(intel, hw_prim[prim]);
 }
 
+static inline GLuint intel_get_vb_max(struct intel_context *intel)
+{
+   GLuint ret;
+
+   if (intel->intelScreen->no_vbo)
+      ret = intel->batch->size - 1500;
+   else
+      ret = INTEL_VB_SIZE;
+   ret /= (intel->vertex_size * 4);
+   return ret;
+}
+
+static inline GLuint intel_get_current_max(struct intel_context *intel)
+{
+
+   if (intel->intelScreen->no_vbo)
+      return intel_get_vb_max(intel);
+   else
+      return (INTEL_VB_SIZE - intel->prim.current_offset) / (intel->vertex_size * 4);
+}
 
 #define LOCAL_VARS struct intel_context *intel = intel_context(ctx)
 #define INIT( prim ) 				\
@@ -126,9 +146,8 @@ do {						\
 
 #define FLUSH() INTEL_FIREVERTICES(intel)
 
-#define GET_SUBSEQUENT_VB_MAX_VERTS() (INTEL_VB_SIZE / (intel->vertex_size * 4))
-#define GET_CURRENT_VB_MAX_VERTS() \
-   ((INTEL_VB_SIZE - intel->prim.current_offset) / (intel->vertex_size * 4))
+#define GET_SUBSEQUENT_VB_MAX_VERTS() intel_get_vb_max(intel)
+#define GET_CURRENT_VB_MAX_VERTS() intel_get_current_max(intel)
 
 #define ALLOC_VERTS(nr) intel_get_prim_space(intel, nr)
 
diff --git a/src/mesa/drivers/dri/i915/intel_tris.c b/src/mesa/drivers/dri/i915/intel_tris.c
index 797d6c5858..c4708dc7ab 100644
--- a/src/mesa/drivers/dri/i915/intel_tris.c
+++ b/src/mesa/drivers/dri/i915/intel_tris.c
@@ -61,9 +61,101 @@ static void intelRenderPrimitive(GLcontext * ctx, GLenum prim);
 static void intelRasterPrimitive(GLcontext * ctx, GLenum rprim,
                                  GLuint hwprim);
 
+static void
+intel_flush_inline_primitive(struct intel_context *intel)
+{
+   GLuint used = intel->batch->ptr - intel->prim.start_ptr;
+
+   assert(intel->prim.primitive != ~0);
+
+/*    _mesa_printf("/\n"); */
+
+   if (used < 8)
+      goto do_discard;
+
+   *(int *) intel->prim.start_ptr = (_3DPRIMITIVE |
+                                     intel->prim.primitive | (used / 4 - 2));
+
+   goto finished;
+
+ do_discard:
+   intel->batch->ptr -= used;
+
+ finished:
+   intel->prim.primitive = ~0;
+   intel->prim.start_ptr = 0;
+   intel->prim.flush = 0;
+}
+
+static void intel_start_inline(struct intel_context *intel, uint32_t prim)
+{
+   BATCH_LOCALS;
+   uint32_t batch_flags = LOOP_CLIPRECTS;
+
+   intel_wait_flips(intel);
+   intel->vtbl.emit_state(intel);
+
+   intel->no_batch_wrap = GL_TRUE;
+
+   /*_mesa_printf("%s *", __progname);*/
+
+   /* Emit a slot which will be filled with the inline primitive
+    * command later.
+    */
+   BEGIN_BATCH(2, batch_flags);
+   OUT_BATCH(0);
+
+   assert((intel->batch->dirty_state & (1<<1)) == 0);
+
+   intel->prim.start_ptr = intel->batch->ptr;
+   intel->prim.primitive = prim;
+   intel->prim.flush = intel_flush_inline_primitive;
+
+   OUT_BATCH(0);
+   ADVANCE_BATCH();
+
+   intel->no_batch_wrap = GL_FALSE;
+/*    _mesa_printf(">"); */
+}
+
+static void intel_wrap_inline(struct intel_context *intel)
+{
+   GLuint prim = intel->prim.primitive;
+
+   intel_flush_inline_primitive(intel);
+   intel_batchbuffer_flush(intel->batch);
+   intel_start_inline(intel, prim);  /* ??? */
+}
+
+static GLuint *intel_extend_inline(struct intel_context *intel, GLuint dwords)
+{
+   GLuint sz = dwords * sizeof(GLuint);
+   GLuint *ptr;
+
+   assert(intel->prim.flush == intel_flush_inline_primitive);
+
+   if (intel_batchbuffer_space(intel->batch) < sz)
+      intel_wrap_inline(intel);
+
+/*    _mesa_printf("."); */
+
+   intel->vtbl.assert_not_dirty(intel);
+
+   ptr = (GLuint *) intel->batch->ptr;
+   intel->batch->ptr += sz;
+
+   return ptr;
+}
+
 /** Sets the primitive type for a primitive sequence, flushing as needed. */
 void intel_set_prim(struct intel_context *intel, uint32_t prim)
 {
+   /* if we have no VBOs */
+
+   if (intel->intelScreen->no_vbo) {
+      intel_start_inline(intel, prim);
+      return;
+   }
    if (prim != intel->prim.primitive) {
       INTEL_FIREVERTICES(intel);
       intel->prim.primitive = prim;
@@ -75,6 +167,10 @@ uint32_t *intel_get_prim_space(struct intel_context *intel, unsigned int count)
 {
    uint32_t *addr;
 
+   if (intel->intelScreen->no_vbo) {
+      return intel_extend_inline(intel, count * intel->vertex_size);
+   }
+
    /* Check for space in the existing VB */
    if (intel->prim.vb_bo == NULL ||
        (intel->prim.current_offset +
@@ -155,7 +251,7 @@ void intel_flush_prim(struct intel_context *intel)
 
 #if 0
    printf("emitting %d..%d=%d vertices size %d\n", offset,
-	  intel->prim.current_offset, intel->prim.count,
+	  intel->prim.current_offset, count,
 	  intel->vertex_size * 4);
 #endif
 
diff --git a/src/mesa/drivers/dri/intel/intel_context.h b/src/mesa/drivers/dri/intel/intel_context.h
index 60028f090a..ee43ed7e83 100644
--- a/src/mesa/drivers/dri/intel/intel_context.h
+++ b/src/mesa/drivers/dri/intel/intel_context.h
@@ -184,6 +184,7 @@ struct intel_context
       GLuint id;
       uint32_t primitive;	/**< Current hardware primitive type */
       void (*flush) (struct intel_context *);
+      GLubyte *start_ptr; /**< for i8xx */
       dri_bo *vb_bo;
       uint8_t *vb;
       unsigned int start_offset; /**< Byte offset of primitive sequence */
diff --git a/src/mesa/drivers/dri/intel/intel_reg.h b/src/mesa/drivers/dri/intel/intel_reg.h
index 68d8a05b83..57ac8f0cc1 100644
--- a/src/mesa/drivers/dri/intel/intel_reg.h
+++ b/src/mesa/drivers/dri/intel/intel_reg.h
@@ -71,14 +71,14 @@
 /** @{
  * 915 definitions
  */
-#define S0_VB_OFFSET_MASK		0xffffffc
+#define S0_VB_OFFSET_MASK		0xffffffc0
 #define S0_AUTO_CACHE_INV_DISABLE	(1<<0)
 /** @} */
 
 /** @{
  * 830 definitions
  */
-#define S0_VB_OFFSET_MASK_830		0xffffff8
+#define S0_VB_OFFSET_MASK_830		0xffffff80
 #define S0_VB_PITCH_SHIFT_830		1
 #define S0_VB_ENABLE_830		(1<<0)
 /** @} */
diff --git a/src/mesa/drivers/dri/intel/intel_screen.c b/src/mesa/drivers/dri/intel/intel_screen.c
index cf09fad745..61b55b97b5 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.c
+++ b/src/mesa/drivers/dri/intel/intel_screen.c
@@ -461,6 +461,7 @@ intelCreateContext(const __GLcontextModes * mesaVis,
 				  sharedContextPrivate);
       }
    } else {
+      intelScreen->no_vbo = GL_TRUE;
       return i830CreateContext(mesaVis, driContextPriv, sharedContextPrivate);
    }
 #else
diff --git a/src/mesa/drivers/dri/intel/intel_screen.h b/src/mesa/drivers/dri/intel/intel_screen.h
index fc913da5bf..91f0d6d1ae 100644
--- a/src/mesa/drivers/dri/intel/intel_screen.h
+++ b/src/mesa/drivers/dri/intel/intel_screen.h
@@ -77,6 +77,7 @@ typedef struct
 
    GLboolean no_hw;
 
+   GLboolean no_vbo;
    int ttm;
    dri_bufmgr *bufmgr;