Merge remote branch 'upstream/gallium-0.1' into nouveau-gallium-0.1

49 files changed, 599 insertions, 238 deletions

diff --git a/src/gallium/drivers/cell/common.h b/src/gallium/drivers/cell/common.h
index d59e4f7036e..b0928fefd2e 100644
--- a/src/gallium/drivers/cell/common.h
+++ b/src/gallium/drivers/cell/common.h
@@ -92,8 +92,9 @@
 #define CELL_CMD_STATE_BIND_VS       18
 #define CELL_CMD_STATE_BLEND         19
 #define CELL_CMD_STATE_ATTRIB_FETCH  20
-#define CELL_CMD_VS_EXECUTE          21
-#define CELL_CMD_FLUSH_BUFFER_RANGE  22
+#define CELL_CMD_STATE_LOGICOP       21
+#define CELL_CMD_VS_EXECUTE          22
+#define CELL_CMD_FLUSH_BUFFER_RANGE  23
 
 
 #define CELL_NUM_BUFFERS 4
@@ -124,6 +125,12 @@ struct cell_command_blend {
 };
 
 
+struct cell_command_logicop {
+   uint64_t base;               /**< Effective address of code start. */
+   unsigned size;               /**< Size in bytes of test code. */
+};
+
+
 /**
  * Tell SPUs about the framebuffer size, location
  */
diff --git a/src/gallium/drivers/cell/ppu/cell_context.c b/src/gallium/drivers/cell/ppu/cell_context.c
index ccbbd1d331c..12eb5aa2547 100644
--- a/src/gallium/drivers/cell/ppu/cell_context.c
+++ b/src/gallium/drivers/cell/ppu/cell_context.c
@@ -103,8 +103,8 @@ cell_create_context(struct pipe_screen *screen,
    cell->pipe.destroy = cell_destroy_context;
 
    /* state setters */
-   cell->pipe.set_vertex_buffer = cell_set_vertex_buffer;
-   cell->pipe.set_vertex_element = cell_set_vertex_element;
+   cell->pipe.set_vertex_buffers = cell_set_vertex_buffers;
+   cell->pipe.set_vertex_elements = cell_set_vertex_elements;
 
    cell->pipe.draw_arrays = cell_draw_arrays;
    cell->pipe.draw_elements = cell_draw_elements;
diff --git a/src/gallium/drivers/cell/ppu/cell_context.h b/src/gallium/drivers/cell/ppu/cell_context.h
index 9e79db0acef..7f656a97447 100644
--- a/src/gallium/drivers/cell/ppu/cell_context.h
+++ b/src/gallium/drivers/cell/ppu/cell_context.h
@@ -92,6 +92,8 @@ struct cell_context
    const struct cell_vertex_shader_state *vs;
    const struct cell_fragment_shader_state *fs;
 
+   struct spe_function logic_op;
+
    struct pipe_blend_color blend_color;
    struct pipe_clip_state clip;
    struct pipe_constant_buffer constants[2];
@@ -101,8 +103,8 @@ struct cell_context
    struct cell_texture *texture[PIPE_MAX_SAMPLERS];
    uint num_textures;
    struct pipe_viewport_state viewport;
-   struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX];
-   struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX];
+   struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+   struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
 
    ubyte *cbuf_map[PIPE_MAX_COLOR_BUFS];
    ubyte *zsbuf_map;
@@ -139,7 +141,7 @@ struct cell_context
 
 
    struct spe_function attrib_fetch;
-   unsigned attrib_fetch_offsets[PIPE_ATTRIB_MAX];
+   unsigned attrib_fetch_offsets[PIPE_MAX_ATTRIBS];
 };
 
 
diff --git a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
index c839fb4d12d..b896252f817 100644
--- a/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
+++ b/src/gallium/drivers/cell/ppu/cell_draw_arrays.c
@@ -123,7 +123,7 @@ cell_draw_elements(struct pipe_context *pipe,
    /*
     * Map vertex buffers
     */
-   for (i = 0; i < PIPE_ATTRIB_MAX; i++) {
+   for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
       if (sp->vertex_buffer[i].buffer) {
          void *buf = pipe->winsys->buffer_map(pipe->winsys,
                                               sp->vertex_buffer[i].buffer,
@@ -151,7 +151,7 @@ cell_draw_elements(struct pipe_context *pipe,
    /*
     * unmap vertex/index buffers - will cause draw module to flush
     */
-   for (i = 0; i < PIPE_ATTRIB_MAX; i++) {
+   for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
       if (sp->vertex_buffer[i].buffer) {
          draw_set_mapped_vertex_buffer(draw, i, NULL);
          pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer);
diff --git a/src/gallium/drivers/cell/ppu/cell_flush.c b/src/gallium/drivers/cell/ppu/cell_flush.c
index 66a5627d844..3aaf3de6684 100644
--- a/src/gallium/drivers/cell/ppu/cell_flush.c
+++ b/src/gallium/drivers/cell/ppu/cell_flush.c
@@ -35,12 +35,19 @@
 
 
 void
-cell_flush(struct pipe_context *pipe, unsigned flags)
+cell_flush(struct pipe_context *pipe, unsigned flags,
+           struct pipe_fence_handle **fence)
 {
    struct cell_context *cell = cell_context(pipe);
 
+   if (fence) {
+      *fence = NULL;
+      /* XXX: Implement real fencing */
+      flags |= CELL_FLUSH_WAIT;
+   }
+
    if (flags & PIPE_FLUSH_SWAPBUFFERS)
-      flags |= PIPE_FLUSH_WAIT;
+      flags |= CELL_FLUSH_WAIT;
 
    draw_flush( cell->draw );
    cell_flush_int(pipe, flags);
@@ -58,7 +65,7 @@ cell_flush_int(struct pipe_context *pipe, unsigned flags)
    ASSERT(!flushing);
    flushing = TRUE;
 
-   if (flags & PIPE_FLUSH_WAIT) {
+   if (flags & CELL_FLUSH_WAIT) {
       uint64_t *cmd = (uint64_t *) cell_batch_alloc(cell, sizeof(uint64_t));
       *cmd = CELL_CMD_FINISH;
    }
@@ -72,7 +79,7 @@ cell_flush_int(struct pipe_context *pipe, unsigned flags)
    }
 #endif
 
-   if (flags & PIPE_FLUSH_WAIT) {
+   if (flags & CELL_FLUSH_WAIT) {
       /* Wait for ack */
       for (i = 0; i < cell->num_spus; i++) {
          uint k = wait_mbox_message(cell_global.spe_contexts[i]);
diff --git a/src/gallium/drivers/cell/ppu/cell_flush.h b/src/gallium/drivers/cell/ppu/cell_flush.h
index 7f940ae76b6..8f0645c4293 100644
--- a/src/gallium/drivers/cell/ppu/cell_flush.h
+++ b/src/gallium/drivers/cell/ppu/cell_flush.h
@@ -29,8 +29,11 @@
 #ifndef CELL_FLUSH
 #define CELL_FLUSH
 
+#define CELL_FLUSH_WAIT 0x80000000
+
 extern void
-cell_flush(struct pipe_context *pipe, unsigned flags);
+cell_flush(struct pipe_context *pipe, unsigned flags,
+           struct pipe_fence_handle **fence);
 
 extern void
 cell_flush_int(struct pipe_context *pipe, unsigned flags);
diff --git a/src/gallium/drivers/cell/ppu/cell_state.h b/src/gallium/drivers/cell/ppu/cell_state.h
index 31ce505e211..82580ea35ab 100644
--- a/src/gallium/drivers/cell/ppu/cell_state.h
+++ b/src/gallium/drivers/cell/ppu/cell_state.h
@@ -48,13 +48,13 @@
 #define CELL_NEW_VERTEX_INFO   0x8000
 
 
-void cell_set_vertex_element(struct pipe_context *,
-                             unsigned index,
-                             const struct pipe_vertex_element *);
+void cell_set_vertex_elements(struct pipe_context *,
+                              unsigned count,
+                              const struct pipe_vertex_element *);
 
-void cell_set_vertex_buffer(struct pipe_context *,
-                            unsigned index,
-                            const struct pipe_vertex_buffer *);
+void cell_set_vertex_buffers(struct pipe_context *,
+                             unsigned count,
+                             const struct pipe_vertex_buffer *);
 
 void cell_update_derived( struct cell_context *softpipe );
 
diff --git a/src/gallium/drivers/cell/ppu/cell_state_emit.c b/src/gallium/drivers/cell/ppu/cell_state_emit.c
index 5709b48f129..4c75caa0251 100644
--- a/src/gallium/drivers/cell/ppu/cell_state_emit.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_emit.c
@@ -29,6 +29,7 @@
 #include "cell_context.h"
 #include "cell_state.h"
 #include "cell_state_emit.h"
+#include "cell_state_per_fragment.h"
 #include "cell_batch.h"
 #include "cell_texture.h"
 #include "draw/draw_context.h"
@@ -50,6 +51,23 @@ emit_state_cmd(struct cell_context *cell, uint cmd,
 void
 cell_emit_state(struct cell_context *cell)
 {
+   if (cell->dirty & (CELL_NEW_FRAMEBUFFER | CELL_NEW_BLEND)) {
+      struct cell_command_logicop logicop;
+
+      if (cell->logic_op.store != NULL) {
+	 spe_release_func(& cell->logic_op);
+      }
+
+      cell_generate_logic_op(& cell->logic_op,
+			     & cell->blend->base,
+			     cell->framebuffer.cbufs[0]);
+
+      logicop.base = (intptr_t) cell->logic_op.store;
+      logicop.size = 64 * 4;
+      emit_state_cmd(cell, CELL_CMD_STATE_LOGICOP, &logicop,
+		     sizeof(logicop));
+   }
+
    if (cell->dirty & CELL_NEW_FRAMEBUFFER) {
       struct pipe_surface *cbuf = cell->framebuffer.cbufs[0];
       struct pipe_surface *zbuf = cell->framebuffer.zsbuf;
diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c
index c750b1d89dc..f10025bd7c7 100644
--- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.c
@@ -977,7 +977,6 @@ cell_generate_alpha_blend(struct cell_blend_state *cb)
       spe_allocate_register(f, 13),
       spe_allocate_register(f, 14),
    };
-   const int mask = spe_allocate_register(f, 15);
    unsigned func[4];
    unsigned sF[4];
    unsigned dF[4];
@@ -1114,9 +1113,6 @@ cell_generate_alpha_blend(struct cell_blend_state *cb)
                                 func[i], sF[i], dF[i],
                                 frag[i], src_factor[i],
                                 pixel[i], dst_factor[i]);
-         spe_selb(f, frag[i], pixel[i], frag[i], mask);
-      } else {
-         spe_or(f, frag[i], pixel[i], pixel[i]);
       }
    }
 
@@ -1146,3 +1142,260 @@ cell_generate_alpha_blend(struct cell_blend_state *cb)
    }
 #endif
 }
+
+
+int PC_OFFSET(const struct spe_function *f, const void *d)
+{
+   const intptr_t pc = (intptr_t) f->csr;
+   const intptr_t ea = ~0x0f & (intptr_t) d;
+
+   return (ea - pc) >> 2;
+}
+
+
+/**
+ * Generate code to perform color conversion and logic op
+ *
+ * \bug
+ * The code generated by this function should also perform dithering.
+ *
+ * \bug
+ * The code generated by this function should also perform color-write
+ * masking.
+ *
+ * \bug
+ * This routine is hard-coded to only work with ARGB8 data.
+ */
+void
+cell_generate_logic_op(struct spe_function *f, struct pipe_blend_state *blend,
+                       struct pipe_surface *surf)
+{
+   const unsigned logic_op = (blend->logicop_enable)
+       ? blend->logicop_func : PIPE_LOGICOP_COPY;
+
+   /* This code generates a maximum of 37 instructions.  An additional 32
+    * bytes (equiv. to 8 instructions) are needed for data storage.  Round up
+    * to 64 to make it a happy power-of-two.
+    */
+   spe_init_func(f, 4 * 64);
+
+
+   /* Pixel colors in framebuffer format in AoS layout.
+    */
+   const int pixel[4] = {
+      spe_allocate_register(f, 3),
+      spe_allocate_register(f, 4),
+      spe_allocate_register(f, 5),
+      spe_allocate_register(f, 6),
+   };
+
+   /* Fragment colors stored as floats in SoA layout.
+    */
+   const int frag[4] = {
+      spe_allocate_register(f, 7),
+      spe_allocate_register(f, 8),
+      spe_allocate_register(f, 9),
+      spe_allocate_register(f, 10),
+   };
+
+   const int mask = spe_allocate_register(f, 11);
+
+
+   /* Short-circuit the noop and invert cases.
+    */
+   if ((logic_op == PIPE_LOGICOP_NOOP) || (blend->colormask == 0)) {
+      spe_bi(f, 0, 0, 0);
+      return;
+   } else if (logic_op == PIPE_LOGICOP_INVERT) {
+      spe_nor(f, pixel[0], pixel[0], pixel[0]);
+      spe_nor(f, pixel[1], pixel[1], pixel[1]);
+      spe_nor(f, pixel[2], pixel[2], pixel[2]);
+      spe_nor(f, pixel[3], pixel[3], pixel[3]);
+      spe_bi(f, 0, 0, 0);
+      return;
+   }
+
+
+   const int tmp[4] = {
+      spe_allocate_available_register(f),
+      spe_allocate_available_register(f),
+      spe_allocate_available_register(f),
+      spe_allocate_available_register(f),
+   };
+
+   const int shuf_xpose_hi = spe_allocate_available_register(f);
+   const int shuf_xpose_lo = spe_allocate_available_register(f);
+   const int shuf_color = spe_allocate_available_register(f);
+
+
+   /* Pointer to the begining of the function's private data area.
+    */
+   uint32_t *const data = ((uint32_t *) f->store) + (64 - 8);
+
+
+   /* Convert fragment colors to framebuffer format in AoS layout.
+    */
+   data[0] = 0x00010203;
+   data[1] = 0x10111213;
+   data[2] = 0x04050607;
+   data[3] = 0x14151617;
+
+   data[4] = 0x0c000408;
+   data[5] = 0x80808080;
+   data[6] = 0x80808080;
+   data[7] = 0x80808080;
+
+   spe_ilh(f, tmp[0], 0x0808);
+   spe_lqr(f, shuf_xpose_hi, PC_OFFSET(f, data+0));
+   spe_lqr(f, shuf_color, PC_OFFSET(f, data+4));
+   spe_a(f, shuf_xpose_lo, shuf_xpose_hi, tmp[0]);
+
+   spe_shufb(f, tmp[0], frag[0], frag[2], shuf_xpose_hi);
+   spe_shufb(f, tmp[1], frag[0], frag[2], shuf_xpose_lo);
+   spe_shufb(f, tmp[2], frag[1], frag[3], shuf_xpose_hi);
+   spe_shufb(f, tmp[3], frag[1], frag[3], shuf_xpose_lo);
+
+   spe_shufb(f, frag[0], tmp[0], tmp[2], shuf_xpose_hi);
+   spe_shufb(f, frag[1], tmp[0], tmp[2], shuf_xpose_lo);
+   spe_shufb(f, frag[2], tmp[1], tmp[3], shuf_xpose_hi);
+   spe_shufb(f, frag[3], tmp[1], tmp[3], shuf_xpose_lo);
+
+   spe_cfltu(f, frag[0], frag[0], 32);
+   spe_cfltu(f, frag[1], frag[1], 32);
+   spe_cfltu(f, frag[2], frag[2], 32);
+   spe_cfltu(f, frag[3], frag[3], 32);
+
+   spe_shufb(f, frag[0], frag[0], pixel[0], shuf_color);
+   spe_shufb(f, frag[1], frag[1], pixel[1], shuf_color);
+   spe_shufb(f, frag[2], frag[2], pixel[2], shuf_color);
+   spe_shufb(f, frag[3], frag[3], pixel[3], shuf_color);
+
+
+   /* If logic op is enabled, perform the requested logical operation on the
+    * converted fragment colors and the pixel colors.
+    */
+   switch (logic_op) {
+   case PIPE_LOGICOP_CLEAR:
+      spe_il(f, frag[0], 0);
+      spe_il(f, frag[1], 0);
+      spe_il(f, frag[2], 0);
+      spe_il(f, frag[3], 0);
+      break;
+   case PIPE_LOGICOP_NOR:
+      spe_nor(f, frag[0], frag[0], pixel[0]);
+      spe_nor(f, frag[1], frag[1], pixel[1]);
+      spe_nor(f, frag[2], frag[2], pixel[2]);
+      spe_nor(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_AND_INVERTED:
+      spe_andc(f, frag[0], pixel[0], frag[0]);
+      spe_andc(f, frag[1], pixel[1], frag[1]);
+      spe_andc(f, frag[2], pixel[2], frag[2]);
+      spe_andc(f, frag[3], pixel[3], frag[3]);
+      break;
+   case PIPE_LOGICOP_COPY_INVERTED:
+      spe_nor(f, frag[0], frag[0], frag[0]);
+      spe_nor(f, frag[1], frag[1], frag[1]);
+      spe_nor(f, frag[2], frag[2], frag[2]);
+      spe_nor(f, frag[3], frag[3], frag[3]);
+      break;
+   case PIPE_LOGICOP_AND_REVERSE:
+      spe_andc(f, frag[0], frag[0], pixel[0]);
+      spe_andc(f, frag[1], frag[1], pixel[1]);
+      spe_andc(f, frag[2], frag[2], pixel[2]);
+      spe_andc(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_XOR:
+      spe_xor(f, frag[0], frag[0], pixel[0]);
+      spe_xor(f, frag[1], frag[1], pixel[1]);
+      spe_xor(f, frag[2], frag[2], pixel[2]);
+      spe_xor(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_NAND:
+      spe_nand(f, frag[0], frag[0], pixel[0]);
+      spe_nand(f, frag[1], frag[1], pixel[1]);
+      spe_nand(f, frag[2], frag[2], pixel[2]);
+      spe_nand(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_AND:
+      spe_and(f, frag[0], frag[0], pixel[0]);
+      spe_and(f, frag[1], frag[1], pixel[1]);
+      spe_and(f, frag[2], frag[2], pixel[2]);
+      spe_and(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_EQUIV:
+      spe_eqv(f, frag[0], frag[0], pixel[0]);
+      spe_eqv(f, frag[1], frag[1], pixel[1]);
+      spe_eqv(f, frag[2], frag[2], pixel[2]);
+      spe_eqv(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_OR_INVERTED:
+      spe_orc(f, frag[0], pixel[0], frag[0]);
+      spe_orc(f, frag[1], pixel[1], frag[1]);
+      spe_orc(f, frag[2], pixel[2], frag[2]);
+      spe_orc(f, frag[3], pixel[3], frag[3]);
+      break;
+   case PIPE_LOGICOP_COPY:
+      break;
+   case PIPE_LOGICOP_OR_REVERSE:
+      spe_orc(f, frag[0], frag[0], pixel[0]);
+      spe_orc(f, frag[1], frag[1], pixel[1]);
+      spe_orc(f, frag[2], frag[2], pixel[2]);
+      spe_orc(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_OR:
+      spe_or(f, frag[0], frag[0], pixel[0]);
+      spe_or(f, frag[1], frag[1], pixel[1]);
+      spe_or(f, frag[2], frag[2], pixel[2]);
+      spe_or(f, frag[3], frag[3], pixel[3]);
+      break;
+   case PIPE_LOGICOP_SET:
+      spe_il(f, frag[0], ~0);
+      spe_il(f, frag[1], ~0);
+      spe_il(f, frag[2], ~0);
+      spe_il(f, frag[3], ~0);
+      break;
+
+   /* These two cases are short-circuited above.
+    */
+   case PIPE_LOGICOP_INVERT:
+   case PIPE_LOGICOP_NOOP:
+   default:
+      assert(0);
+   }
+
+
+   /* Apply fragment mask.
+    */
+   spe_ilh(f, tmp[0], 0x0000);
+   spe_ilh(f, tmp[1], 0x0404);
+   spe_ilh(f, tmp[2], 0x0808);
+   spe_ilh(f, tmp[3], 0x0c0c);
+
+   spe_shufb(f, tmp[0], mask, mask, tmp[0]);
+   spe_shufb(f, tmp[1], mask, mask, tmp[1]);
+   spe_shufb(f, tmp[2], mask, mask, tmp[2]);
+   spe_shufb(f, tmp[3], mask, mask, tmp[3]);
+
+   spe_selb(f, pixel[0], pixel[0], frag[0], tmp[0]);
+   spe_selb(f, pixel[1], pixel[1], frag[1], tmp[1]);
+   spe_selb(f, pixel[2], pixel[2], frag[2], tmp[2]);
+   spe_selb(f, pixel[3], pixel[3], frag[3], tmp[3]);
+
+   spe_bi(f, 0, 0, 0);
+
+#if 0
+   {
+      const uint32_t *p = f->store;
+      unsigned i;
+
+      printf("# %u instructions\n", f->csr - f->store);
+
+      printf("\t.text\n");
+      for (i = 0; i < 64; i++) {
+         printf("\t.long\t0x%04x\n", p[i]);
+      }
+      fflush(stdout);
+   }
+#endif
+}
diff --git a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h
index f699247f9e4..ab4de96c69d 100644
--- a/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h
+++ b/src/gallium/drivers/cell/ppu/cell_state_per_fragment.h
@@ -31,4 +31,8 @@ cell_generate_depth_stencil_test(struct cell_depth_stencil_alpha_state *cdsa);
 extern void
 cell_generate_alpha_blend(struct cell_blend_state *cb);
 
+extern void
+cell_generate_logic_op(struct spe_function *f, struct pipe_blend_state *blend,
+    struct pipe_surface *surf);
+
 #endif /* CELL_STATE_PER_FRAGMENT_H */
diff --git a/src/gallium/drivers/cell/ppu/cell_state_vertex.c b/src/gallium/drivers/cell/ppu/cell_state_vertex.c
index 563831b62db..6c83b8dc722 100644
--- a/src/gallium/drivers/cell/ppu/cell_state_vertex.c
+++ b/src/gallium/drivers/cell/ppu/cell_state_vertex.c
@@ -36,28 +36,34 @@
 
 
 void
-cell_set_vertex_element(struct pipe_context *pipe,
-                            unsigned index,
-                            const struct pipe_vertex_element *attrib)
+cell_set_vertex_elements(struct pipe_context *pipe,
+                         unsigned count,
+                         const struct pipe_vertex_element *elements)
 {
    struct cell_context *cell = cell_context(pipe);
-   assert(index < PIPE_ATTRIB_MAX);
-   cell->vertex_element[index] = *attrib; /* struct copy */
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(cell->vertex_element, elements, count * sizeof(elements[0]));
+
    cell->dirty |= CELL_NEW_VERTEX;
 
-   draw_set_vertex_element(cell->draw, index, attrib);
+   draw_set_vertex_elements(cell->draw, count, elements);
 }
 
 
 void
-cell_set_vertex_buffer(struct pipe_context *pipe,
-                           unsigned index,
-                           const struct pipe_vertex_buffer *buffer)
+cell_set_vertex_buffers(struct pipe_context *pipe,
+                        unsigned count,
+                        const struct pipe_vertex_buffer *buffers)
 {
    struct cell_context *cell = cell_context(pipe);
-   assert(index < PIPE_ATTRIB_MAX);
-   cell->vertex_buffer[index] = *buffer; /* struct copy */
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(cell->vertex_buffer, buffers, count * sizeof(buffers[0]));
+
    cell->dirty |= CELL_NEW_VERTEX;
 
-   draw_set_vertex_buffer(cell->draw, index, buffer);
+   draw_set_vertex_buffers(cell->draw, count, buffers);
 }
diff --git a/src/gallium/drivers/cell/ppu/cell_vbuf.c b/src/gallium/drivers/cell/ppu/cell_vbuf.c
index cc727ff4edb..3a181b585c5 100644
--- a/src/gallium/drivers/cell/ppu/cell_vbuf.c
+++ b/src/gallium/drivers/cell/ppu/cell_vbuf.c
@@ -243,7 +243,7 @@ cell_vbuf_draw(struct vbuf_render *vbr,
 
 #if 0
    /* helpful for debug */
-   cell_flush_int(&cell->pipe, PIPE_FLUSH_WAIT);
+   cell_flush_int(&cell->pipe, CELL_FLUSH_WAIT);
 #endif
 }
 
diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c
index 4828a8023bd..49d5443cde7 100644
--- a/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c
+++ b/src/gallium/drivers/cell/ppu/cell_vertex_fetch.c
@@ -263,7 +263,7 @@ void cell_update_vertex_fetch(struct draw_context *draw)
    struct cell_context *const cell =
        (struct cell_context *) draw->driver_private;
    struct spe_function *p = &cell->attrib_fetch;
-   unsigned function_index[PIPE_ATTRIB_MAX];
+   unsigned function_index[PIPE_MAX_ATTRIBS];
    unsigned unique_attr_formats;
    int out_ptr;
    int in_ptr;
diff --git a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c
index f5c27852c14..f753960a0fb 100644
--- a/src/gallium/drivers/cell/ppu/cell_vertex_shader.c
+++ b/src/gallium/drivers/cell/ppu/cell_vertex_shader.c
@@ -69,7 +69,7 @@ cell_vertex_shader_queue_flush(struct draw_context *draw)
    batch = cell_batch_alloc(cell, sizeof(batch[0]) + sizeof(*cf));
    batch[0] = CELL_CMD_STATE_ATTRIB_FETCH;
    cf = (struct cell_attribute_fetch_code *) (&batch[1]);
-   cf->base = cell->attrib_fetch.store;
+   cf->base = (uint64_t) cell->attrib_fetch.store;
    cf->size = ROUNDUP16((unsigned)((void *) cell->attrib_fetch.csr 
 				   - (void *) cell->attrib_fetch.store));
 
@@ -133,7 +133,7 @@ cell_vertex_shader_queue_flush(struct draw_context *draw)
       vs->num_elts = n;
       send_mbox_message(cell_global.spe_contexts[0], CELL_CMD_VS_EXECUTE);
 
-      cell_flush_int(& cell->pipe, PIPE_FLUSH_WAIT);
+      cell_flush_int(& cell->pipe, CELL_FLUSH_WAIT);
    }
 
    draw->vs.post_nr = draw->vs.queue_nr;
diff --git a/src/gallium/drivers/cell/spu/spu_exec.c b/src/gallium/drivers/cell/spu/spu_exec.c
index 061fbebf618..48edc62f49b 100644
--- a/src/gallium/drivers/cell/spu/spu_exec.c
+++ b/src/gallium/drivers/cell/spu/spu_exec.c
@@ -1453,7 +1453,7 @@ exec_instruction(
       break;
 
    case TGSI_OPCODE_TXP:
-      /* Texture lookup with projection
+      /* Texture lookup with projection */
       /* src[0] = texcoord (src[0].w = projection) */
       /* src[1] = sampler unit */
       exec_tex(mach, inst, TRUE, TRUE);
diff --git a/src/gallium/drivers/cell/spu/spu_main.c b/src/gallium/drivers/cell/spu/spu_main.c
index 0a490ab277f..d7f46f80246 100644
--- a/src/gallium/drivers/cell/spu/spu_main.c
+++ b/src/gallium/drivers/cell/spu/spu_main.c
@@ -55,7 +55,7 @@ struct spu_global spu;
 
 struct spu_vs_context draw;
 
-static unsigned char attribute_fetch_code_buffer[136 * PIPE_ATTRIB_MAX]
+static unsigned char attribute_fetch_code_buffer[136 * PIPE_MAX_ATTRIBS]
     ALIGN16_ATTRIB;
 
 static unsigned char depth_stencil_code_buffer[4 * 64]
@@ -64,6 +64,9 @@ static unsigned char depth_stencil_code_buffer[4 * 64]
 static unsigned char fb_blend_code_buffer[4 * 64]
     ALIGN16_ATTRIB;
 
+static unsigned char logicop_code_buffer[4 * 64]
+    ALIGN16_ATTRIB;
+
 
 /**
  * Tell the PPU that this SPU has finished copying a buffer to
@@ -358,7 +361,7 @@ cmd_state_vs_array_info(const struct cell_array_info *vs_info)
 {
    const unsigned attr = vs_info->attr;
 
-   ASSERT(attr < PIPE_ATTRIB_MAX);
+   ASSERT(attr < PIPE_MAX_ATTRIBS);
    draw.vertex_fetch.src_ptr[attr] = vs_info->base;
    draw.vertex_fetch.pitch[attr] = vs_info->pitch;
    draw.vertex_fetch.size[attr] = vs_info->size;
@@ -513,6 +516,22 @@ cmd_batch(uint opcode)
          pos += (1 + ROUNDUP8(sizeof(struct cell_attribute_fetch_code)) / 8);
          break;
       }
+      case CELL_CMD_STATE_LOGICOP: {
+         struct cell_command_logicop *code =
+             (struct cell_command_logicop *) &buffer[pos+1];
+
+              mfc_get(logicop_code_buffer,
+                      (unsigned int) code->base,  /* src */
+                      code->size,
+                      TAG_BATCH_BUFFER,
+                      0, /* tid */
+                      0  /* rid */);
+         wait_on_mask(1 << TAG_BATCH_BUFFER);
+
+	 spu.logicop = (logicop_func) logicop_code_buffer;
+         pos += (1 + ROUNDUP8(sizeof(struct cell_command_logicop)) / 8);
+         break;
+      }
       case CELL_CMD_FLUSH_BUFFER_RANGE: {
 	 struct cell_buffer_range *br = (struct cell_buffer_range *)
 	     &buffer[pos+1];
diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h
index 49f5d99674a..c20452931a9 100644
--- a/src/gallium/drivers/cell/spu/spu_main.h
+++ b/src/gallium/drivers/cell/spu/spu_main.h
@@ -77,9 +77,14 @@ struct spu_blend_results {
 typedef struct spu_blend_results (*blend_func)(
     qword frag_r, qword frag_g, qword frag_b, qword frag_a,
     qword pixel_r, qword pixel_g, qword pixel_b, qword pixel_a,
-    qword const_r, qword const_g, qword const_b, qword const_a,
+    qword const_r, qword const_g, qword const_b, qword const_a);
+
+typedef struct spu_blend_results (*logicop_func)(
+    qword pixel_r, qword pixel_g, qword pixel_b, qword pixel_a,
+    qword frag_r, qword frag_g, qword frag_b, qword frag_a,
     qword frag_mask);
 
+
 struct spu_framebuffer {
    void *color_start;              /**< addr of color surface in main memory */
    void *depth_start;              /**< addr of depth surface in main memory */
@@ -111,6 +116,8 @@ struct spu_global
    blend_func blend;
    qword const_blend_color[4] ALIGN16_ATTRIB;
 
+   logicop_func logicop;
+
    struct pipe_sampler_state sampler[PIPE_MAX_SAMPLERS];
    struct cell_command_texture texture;
 
diff --git a/src/gallium/drivers/cell/spu/spu_tri.c b/src/gallium/drivers/cell/spu/spu_tri.c
index e6a1ce01dfd..95c629a8aae 100644
--- a/src/gallium/drivers/cell/spu/spu_tri.c
+++ b/src/gallium/drivers/cell/spu/spu_tri.c
@@ -305,7 +305,6 @@ emit_quad( int x, int y, mask_t mask )
    if (spu_extract(spu_orx(mask), 0)) {
       const int ix = x - setup.cliprect_minx;
       const int iy = y - setup.cliprect_miny;
-      const vector unsigned char shuffle = spu.color_shuffle;
       vector float colors[4];
 
       spu.cur_ctile_status = TILE_STATUS_DIRTY;
@@ -330,45 +329,53 @@ emit_quad( int x, int y, mask_t mask )
       }
 
 
+      /* Convert fragment data from AoS to SoA format.
+       */
+      qword soa_frag[4];
+      _transpose_matrix4x4((vec_float4 *) soa_frag, colors);
+
       /* Read the current framebuffer values.
-       *
-       * Ignore read_fb for now.  In the future we can use this to avoid
-       * reading the framebuffer if read_fb is false and the fragment mask is
-       * all 0xffffffff.  This is the common case, so it is probably worth
-       * the effort.  We'll have to profile to determine whether or not the
-       * extra conditional branches hurt overall performance.
        */
-      vec_float4 aos_pix[4] = {
-         spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+0]),
-         spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+1]),
-         spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+0]),
-         spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+1]),
+      const qword pix[4] = {
+         (qword) spu_splats(spu.ctile.ui[iy+0][ix+0]),
+         (qword) spu_splats(spu.ctile.ui[iy+0][ix+1]),
+         (qword) spu_splats(spu.ctile.ui[iy+1][ix+0]),
+         (qword) spu_splats(spu.ctile.ui[iy+1][ix+1]),
       };
 
       qword soa_pix[4];
-      qword soa_frag[4];
 
-      /* Convert pixel and fragment data from AoS to SoA format.
-       */
-      _transpose_matrix4x4((vec_float4 *) soa_pix, aos_pix);
-      _transpose_matrix4x4((vec_float4 *) soa_frag, colors);
+      if (spu.read_fb) {
+         /* Convert pixel data from AoS to SoA format.
+          */
+         vec_float4 aos_pix[4] = {
+            spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+0]),
+            spu_unpack_A8R8G8B8(spu.ctile.ui[iy+0][ix+1]),
+            spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+0]),
+            spu_unpack_A8R8G8B8(spu.ctile.ui[iy+1][ix+1]),
+         };
+
+         _transpose_matrix4x4((vec_float4 *) soa_pix, aos_pix);
+      }
 
-      const struct spu_blend_results result =
+
+      struct spu_blend_results result =
           (*spu.blend)(soa_frag[0], soa_frag[1], soa_frag[2], soa_frag[3],
                        soa_pix[0], soa_pix[1], soa_pix[2], soa_pix[3],
                        spu.const_blend_color[0], spu.const_blend_color[1],
-                       spu.const_blend_color[2], spu.const_blend_color[3],
-                       (qword) mask);
+                       spu.const_blend_color[2], spu.const_blend_color[3]);
 
 
       /* Convert final pixel data from SoA to AoS format.
        */
-      _transpose_matrix4x4(aos_pix, (const vec_float4 *) &result);
-
-      spu.ctile.ui[iy+0][ix+0] = spu_pack_color_shuffle(aos_pix[0], shuffle);
-      spu.ctile.ui[iy+0][ix+1] = spu_pack_color_shuffle(aos_pix[1], shuffle);
-      spu.ctile.ui[iy+1][ix+0] = spu_pack_color_shuffle(aos_pix[2], shuffle);
-      spu.ctile.ui[iy+1][ix+1] = spu_pack_color_shuffle(aos_pix[3], shuffle);
+      result = (*spu.logicop)(pix[0], pix[1], pix[2], pix[3],
+                              result.r, result.g, result.b, result.a,
+                              (qword) mask);
+
+      spu.ctile.ui[iy+0][ix+0] = spu_extract((vec_uint4) result.r, 0);
+      spu.ctile.ui[iy+0][ix+1] = spu_extract((vec_uint4) result.g, 0);
+      spu.ctile.ui[iy+1][ix+0] = spu_extract((vec_uint4) result.b, 0);
+      spu.ctile.ui[iy+1][ix+1] = spu_extract((vec_uint4) result.a, 0);
    }
 #endif
 }
diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.c b/src/gallium/drivers/cell/spu/spu_vertex_shader.c
index 8363efeeb6e..3119a78c060 100644
--- a/src/gallium/drivers/cell/spu/spu_vertex_shader.c
+++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.c
@@ -86,8 +86,8 @@ run_vertex_program(struct spu_vs_context *draw,
    struct spu_exec_machine *machine = &draw->machine;
    unsigned int j;
 
-   ALIGN16_DECL(struct spu_exec_vector, inputs, PIPE_ATTRIB_MAX);
-   ALIGN16_DECL(struct spu_exec_vector, outputs, PIPE_ATTRIB_MAX);
+   ALIGN16_DECL(struct spu_exec_vector, inputs, PIPE_MAX_ATTRIBS);
+   ALIGN16_DECL(struct spu_exec_vector, outputs, PIPE_MAX_ATTRIBS);
    const float *scale = draw->viewport.scale;
    const float *trans = draw->viewport.translate;
 
diff --git a/src/gallium/drivers/cell/spu/spu_vertex_shader.h b/src/gallium/drivers/cell/spu/spu_vertex_shader.h
index 54a4b8d9b9f..4c74f5e74d5 100644
--- a/src/gallium/drivers/cell/spu/spu_vertex_shader.h
+++ b/src/gallium/drivers/cell/spu/spu_vertex_shader.h
@@ -16,10 +16,10 @@ struct spu_vs_context {
    struct pipe_viewport_state viewport;
 
    struct {
-      uint64_t src_ptr[PIPE_ATTRIB_MAX];
-      unsigned pitch[PIPE_ATTRIB_MAX];
-      unsigned size[PIPE_ATTRIB_MAX];
-      unsigned code_offset[PIPE_ATTRIB_MAX];
+      uint64_t src_ptr[PIPE_MAX_ATTRIBS];
+      unsigned pitch[PIPE_MAX_ATTRIBS];
+      unsigned size[PIPE_MAX_ATTRIBS];
+      unsigned code_offset[PIPE_MAX_ATTRIBS];
       unsigned nr_attrs;
       boolean dirty;
 
diff --git a/src/gallium/drivers/failover/fo_context.c b/src/gallium/drivers/failover/fo_context.c
index afc0d7eb1ec..cb95ba516f9 100644
--- a/src/gallium/drivers/failover/fo_context.c
+++ b/src/gallium/drivers/failover/fo_context.c
@@ -69,7 +69,7 @@ static boolean failover_draw_elements( struct pipe_context *pipe,
 					start, 
 					count )) {
 
-	 failover->hw->flush( failover->hw, ~0 );
+	 failover->hw->flush( failover->hw, ~0, NULL );
 	 failover->mode = FO_SW;
       }
    }
@@ -92,7 +92,7 @@ static boolean failover_draw_elements( struct pipe_context *pipe,
        * intervening flush.  Unlikely to be much performance impact to
        * this:
        */
-      failover->sw->flush( failover->sw, ~0 );
+      failover->sw->flush( failover->sw, ~0, NULL );
    }
 
    return TRUE;
diff --git a/src/gallium/drivers/failover/fo_context.h b/src/gallium/drivers/failover/fo_context.h
index 8f3ad3ee798..c6409fe1e1c 100644
--- a/src/gallium/drivers/failover/fo_context.h
+++ b/src/gallium/drivers/failover/fo_context.h
@@ -84,15 +84,16 @@ struct failover_context {
    struct pipe_scissor_state scissor;
    struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
    struct pipe_viewport_state viewport;
-   struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX];
-   struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX];
+   struct pipe_vertex_buffer vertex_buffers[PIPE_MAX_ATTRIBS];
+   struct pipe_vertex_element vertex_elements[PIPE_MAX_ATTRIBS];
+
+   uint num_vertex_buffers;
+   uint num_vertex_elements;
 
    void *sw_sampler_state[PIPE_MAX_SAMPLERS];
    void *hw_sampler_state[PIPE_MAX_SAMPLERS];
 
    unsigned dirty;
-   unsigned dirty_vertex_buffer;
-   unsigned dirty_vertex_element;
 
    unsigned num_samplers;
    unsigned num_textures;
diff --git a/src/gallium/drivers/failover/fo_state.c b/src/gallium/drivers/failover/fo_state.c
index 11eec2714eb..6a797066322 100644
--- a/src/gallium/drivers/failover/fo_state.c
+++ b/src/gallium/drivers/failover/fo_state.c
@@ -402,32 +402,35 @@ failover_set_viewport_state( struct pipe_context *pipe,
 
 
 static void
-failover_set_vertex_buffer(struct pipe_context *pipe,
-			   unsigned unit,
-                           const struct pipe_vertex_buffer *vertex_buffer)
+failover_set_vertex_buffers(struct pipe_context *pipe,
+                            unsigned count,
+                            const struct pipe_vertex_buffer *vertex_buffers)
 {
    struct failover_context *failover = failover_context(pipe);
 
-   failover->vertex_buffer[unit] = *vertex_buffer;
+   memcpy(failover->vertex_buffers, vertex_buffers,
+          count * sizeof(vertex_buffers[0]));
    failover->dirty |= FO_NEW_VERTEX_BUFFER;
-   failover->dirty_vertex_buffer |= (1<<unit);
-   failover->sw->set_vertex_buffer( failover->sw, unit, vertex_buffer );
-   failover->hw->set_vertex_buffer( failover->hw, unit, vertex_buffer );
+   failover->num_vertex_buffers = count;
+   failover->sw->set_vertex_buffers( failover->sw, count, vertex_buffers );
+   failover->hw->set_vertex_buffers( failover->hw, count, vertex_buffers );
 }
 
 
 static void
-failover_set_vertex_element(struct pipe_context *pipe,
-			    unsigned unit,
-			    const struct pipe_vertex_element *vertex_element)
+failover_set_vertex_elements(struct pipe_context *pipe,
+                             unsigned count,
+                             const struct pipe_vertex_element *vertex_elements)
 {
    struct failover_context *failover = failover_context(pipe);
 
-   failover->vertex_element[unit] = *vertex_element;
+   memcpy(failover->vertex_elements, vertex_elements,
+          count * sizeof(vertex_elements[0]));
+
    failover->dirty |= FO_NEW_VERTEX_ELEMENT;
-   failover->dirty_vertex_element |= (1<<unit);
-   failover->sw->set_vertex_element( failover->sw, unit, vertex_element );
-   failover->hw->set_vertex_element( failover->hw, unit, vertex_element );
+   failover->num_vertex_elements = count;
+   failover->sw->set_vertex_elements( failover->sw, count, vertex_elements );
+   failover->hw->set_vertex_elements( failover->hw, count, vertex_elements );
 }
 
 void
@@ -474,7 +477,7 @@ failover_init_state_functions( struct failover_context *failover )
    failover->pipe.set_scissor_state = failover_set_scissor_state;
    failover->pipe.set_sampler_textures = failover_set_sampler_textures;
    failover->pipe.set_viewport_state = failover_set_viewport_state;
-   failover->pipe.set_vertex_buffer = failover_set_vertex_buffer;
-   failover->pipe.set_vertex_element = failover_set_vertex_element;
+   failover->pipe.set_vertex_buffers = failover_set_vertex_buffers;
+   failover->pipe.set_vertex_elements = failover_set_vertex_elements;
    failover->pipe.set_constant_buffer = failover_set_constant_buffer;
 }
diff --git a/src/gallium/drivers/failover/fo_state_emit.c b/src/gallium/drivers/failover/fo_state_emit.c
index 3de931e04eb..bd4fce9d209 100644
--- a/src/gallium/drivers/failover/fo_state_emit.c
+++ b/src/gallium/drivers/failover/fo_state_emit.c
@@ -53,8 +53,6 @@
 void
 failover_state_emit( struct failover_context *failover )
 {
-   unsigned i;
-
    if (failover->dirty & FO_NEW_BLEND)
       failover->sw->bind_blend_state( failover->sw,
                                       failover->blend->sw_state );
@@ -104,24 +102,16 @@ failover_state_emit( struct failover_context *failover )
    }
 
    if (failover->dirty & FO_NEW_VERTEX_BUFFER) {
-      for (i = 0; i < PIPE_ATTRIB_MAX; i++) {
-	 if (failover->dirty_vertex_buffer & (1<<i)) {
-	    failover->sw->set_vertex_buffer( failover->sw, i, 
-					     &failover->vertex_buffer[i] );
-	 }
-      }
+      failover->sw->set_vertex_buffers( failover->sw,
+                                        failover->num_vertex_buffers,
+                                        failover->vertex_buffers );
    }
 
    if (failover->dirty & FO_NEW_VERTEX_ELEMENT) {
-      for (i = 0; i < PIPE_ATTRIB_MAX; i++) {
-	 if (failover->dirty_vertex_element & (1<<i)) {
-	    failover->sw->set_vertex_element( failover->sw, i, 
-					      &failover->vertex_element[i] );
-	 }
-      }
+      failover->sw->set_vertex_elements( failover->sw,
+                                         failover->num_vertex_elements,
+                                         failover->vertex_elements );
    }
 
    failover->dirty = 0;
-   failover->dirty_vertex_element = 0;
-   failover->dirty_vertex_buffer = 0;
 }
diff --git a/src/gallium/drivers/i915simple/i915_batch.h b/src/gallium/drivers/i915simple/i915_batch.h
index fb88cd6db0d..4ea06ce02bf 100644
--- a/src/gallium/drivers/i915simple/i915_batch.h
+++ b/src/gallium/drivers/i915simple/i915_batch.h
@@ -44,9 +44,9 @@
 
 #define ADVANCE_BATCH()
 
-#define FLUSH_BATCH() do { 				\
+#define FLUSH_BATCH(fence) do { 			\
    if (0) i915_dump_batchbuffer( i915 );		\
-   i915->winsys->batch_flush( i915->winsys );		\
+   i915->winsys->batch_flush( i915->winsys, fence );	\
    i915->batch_start = NULL;				\
    i915->hardware_dirty = ~0;				\
 } while (0)
diff --git a/src/gallium/drivers/i915simple/i915_blit.c b/src/gallium/drivers/i915simple/i915_blit.c
index db4671ff553..24449e3fb33 100644
--- a/src/gallium/drivers/i915simple/i915_blit.c
+++ b/src/gallium/drivers/i915simple/i915_blit.c
@@ -70,7 +70,7 @@ i915_fill_blit(struct i915_context *i915,
 
 
    if (!BEGIN_BATCH(6, 1)) {
-      FLUSH_BATCH();
+      FLUSH_BATCH(NULL);
       assert(BEGIN_BATCH(6, 1));
    }
    OUT_BATCH(CMD);
@@ -145,7 +145,7 @@ i915_copy_blit( struct i915_context *i915,
 
 
    if (!BEGIN_BATCH(8, 2)) {
-      FLUSH_BATCH();
+      FLUSH_BATCH(NULL);
       assert(BEGIN_BATCH(8, 2));
    }
    OUT_BATCH(CMD);
diff --git a/src/gallium/drivers/i915simple/i915_context.c b/src/gallium/drivers/i915simple/i915_context.c
index 15ff2360b74..fee33d82de4 100644
--- a/src/gallium/drivers/i915simple/i915_context.c
+++ b/src/gallium/drivers/i915simple/i915_context.c
@@ -65,7 +65,7 @@ i915_draw_elements( struct pipe_context *pipe,
    /*
     * Map vertex buffers
     */
-   for (i = 0; i < PIPE_ATTRIB_MAX; i++) {
+   for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
       if (i915->vertex_buffer[i].buffer) {
          void *buf
             = pipe->winsys->buffer_map(pipe->winsys,
@@ -96,7 +96,7 @@ i915_draw_elements( struct pipe_context *pipe,
    /*
     * unmap vertex/index buffers
     */
-   for (i = 0; i < PIPE_ATTRIB_MAX; i++) {
+   for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
       if (i915->vertex_buffer[i].buffer) {
          pipe->winsys->buffer_unmap(pipe->winsys, i915->vertex_buffer[i].buffer);
          draw_set_mapped_vertex_buffer(draw, i, NULL);
diff --git a/src/gallium/drivers/i915simple/i915_context.h b/src/gallium/drivers/i915simple/i915_context.h
index 746f18ba38d..8e707ea574a 100644
--- a/src/gallium/drivers/i915simple/i915_context.h
+++ b/src/gallium/drivers/i915simple/i915_context.h
@@ -232,7 +232,7 @@ struct i915_context
    struct pipe_scissor_state scissor;

    struct i915_texture *texture[PIPE_MAX_SAMPLERS];

    struct pipe_viewport_state viewport;

-   struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX];

+   struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];

 

    unsigned dirty;

 

diff --git a/src/gallium/drivers/i915simple/i915_flush.c b/src/gallium/drivers/i915simple/i915_flush.c
index 96a54281f11..7d23e6b6b90 100644
--- a/src/gallium/drivers/i915simple/i915_flush.c
+++ b/src/gallium/drivers/i915simple/i915_flush.c
@@ -37,11 +37,9 @@
 #include "i915_batch.h"
 
 
-/**
- * In future we may want a fence-like interface instead of finish.
- */
 static void i915_flush( struct pipe_context *pipe,
-			unsigned flags )
+                        unsigned flags,
+                        struct pipe_fence_handle **fence )
 {
    struct i915_context *i915 = i915_context(pipe);
 
@@ -60,7 +58,7 @@ static void i915_flush( struct pipe_context *pipe,
 	 flush |= FLUSH_MAP_CACHE;
 
       if (!BEGIN_BATCH(1, 0)) {
-	 FLUSH_BATCH();
+	 FLUSH_BATCH(NULL);
 	 assert(BEGIN_BATCH(1, 0));
       }
       OUT_BATCH( flush );
@@ -69,11 +67,7 @@ static void i915_flush( struct pipe_context *pipe,
 
    /* If there are no flags, just flush pending commands to hardware:
     */
-   FLUSH_BATCH();
-
-   if (flags & PIPE_FLUSH_WAIT) {
-      i915->winsys->batch_finish(i915->winsys);
-   }
+   FLUSH_BATCH(fence);
 }
 
 
diff --git a/src/gallium/drivers/i915simple/i915_prim_emit.c b/src/gallium/drivers/i915simple/i915_prim_emit.c
index d8de5178f60..b6fb0a6d88f 100644
--- a/src/gallium/drivers/i915simple/i915_prim_emit.c
+++ b/src/gallium/drivers/i915simple/i915_prim_emit.c
@@ -140,7 +140,7 @@ emit_prim( struct draw_stage *stage,
    assert(vertex_size >= 12); /* never smaller than 12 bytes */
 
    if (!BEGIN_BATCH( 1 + nr * vertex_size / 4, 0 )) {
-      FLUSH_BATCH();
+      FLUSH_BATCH(NULL);
 
       /* Make sure state is re-emitted after a flush: 
        */
diff --git a/src/gallium/drivers/i915simple/i915_prim_vbuf.c b/src/gallium/drivers/i915simple/i915_prim_vbuf.c
index eb64f51943b..7fb2adbb53b 100644
--- a/src/gallium/drivers/i915simple/i915_prim_vbuf.c
+++ b/src/gallium/drivers/i915simple/i915_prim_vbuf.c
@@ -161,7 +161,7 @@ i915_vbuf_render_draw( struct vbuf_render *render,
       i915_emit_hardware_state( i915 );
 
    if (!BEGIN_BATCH( 1 + (nr_indices + 1)/2, 1 )) {
-      FLUSH_BATCH();
+      FLUSH_BATCH(NULL);
 
       /* Make sure state is re-emitted after a flush: 
        */
diff --git a/src/gallium/drivers/i915simple/i915_state.c b/src/gallium/drivers/i915simple/i915_state.c
index 1cec36e206d..4404bc45901 100644
--- a/src/gallium/drivers/i915simple/i915_state.c
+++ b/src/gallium/drivers/i915simple/i915_state.c
@@ -687,23 +687,24 @@ static void i915_delete_rasterizer_state(struct pipe_context *pipe,
    FREE(raster);
 }
 
-static void i915_set_vertex_buffer( struct pipe_context *pipe,
-                                    unsigned index,
-                                    const struct pipe_vertex_buffer *buffer )
+static void i915_set_vertex_buffers(struct pipe_context *pipe,
+                                    unsigned count,
+                                    const struct pipe_vertex_buffer *buffers)
 {
    struct i915_context *i915 = i915_context(pipe);
-   i915->vertex_buffer[index] = *buffer;
+
+   memcpy(i915->vertex_buffer, buffers, count * sizeof(buffers[0]));
    /* pass-through to draw module */
-   draw_set_vertex_buffer(i915->draw, index, buffer);
+   draw_set_vertex_buffers(i915->draw, count, buffers);
 }
 
-static void i915_set_vertex_element( struct pipe_context *pipe,
-                                     unsigned index,
-                                     const struct pipe_vertex_element *element)
+static void i915_set_vertex_elements(struct pipe_context *pipe,
+                                     unsigned count,
+                                     const struct pipe_vertex_element *elements)
 {
    struct i915_context *i915 = i915_context(pipe);
    /* pass-through to draw module */
-   draw_set_vertex_element(i915->draw, index, element);
+   draw_set_vertex_elements(i915->draw, count, elements);
 }
 
 
@@ -742,6 +743,6 @@ i915_init_state_functions( struct i915_context *i915 )
    i915->pipe.set_scissor_state = i915_set_scissor_state;
    i915->pipe.set_sampler_textures = i915_set_sampler_textures;
    i915->pipe.set_viewport_state = i915_set_viewport_state;
-   i915->pipe.set_vertex_buffer = i915_set_vertex_buffer;
-   i915->pipe.set_vertex_element = i915_set_vertex_element;
+   i915->pipe.set_vertex_buffers = i915_set_vertex_buffers;
+   i915->pipe.set_vertex_elements = i915_set_vertex_elements;
 }
diff --git a/src/gallium/drivers/i915simple/i915_state_emit.c b/src/gallium/drivers/i915simple/i915_state_emit.c
index a7498d22b7e..6f947d43468 100644
--- a/src/gallium/drivers/i915simple/i915_state_emit.c
+++ b/src/gallium/drivers/i915simple/i915_state_emit.c
@@ -115,7 +115,7 @@ i915_emit_hardware_state(struct i915_context *i915 )
 #endif
    
    if(!BEGIN_BATCH(dwords, relocs)) {
-      FLUSH_BATCH();
+      FLUSH_BATCH(NULL);
       assert(BEGIN_BATCH(dwords, relocs));
    }
 
diff --git a/src/gallium/drivers/i915simple/i915_winsys.h b/src/gallium/drivers/i915simple/i915_winsys.h
index aea30032818..5e16543f4ec 100644
--- a/src/gallium/drivers/i915simple/i915_winsys.h
+++ b/src/gallium/drivers/i915simple/i915_winsys.h
@@ -56,6 +56,7 @@ extern "C" {
  */
 
 struct pipe_buffer;
+struct pipe_fence_handle;
 struct pipe_winsys;
 struct pipe_screen;
 
@@ -103,8 +104,8 @@ struct i915_winsys {
 			unsigned access_flags,
 			unsigned delta );
    
-   void (*batch_flush)( struct i915_winsys *sws );
-   void (*batch_finish)( struct i915_winsys *sws );
+   void (*batch_flush)( struct i915_winsys *sws,
+                        struct pipe_fence_handle **fence );
 };
 
 #define I915_BUFFER_ACCESS_WRITE   0x1 
diff --git a/src/gallium/drivers/i965simple/brw_clip.h b/src/gallium/drivers/i965simple/brw_clip.h
index a89d08b7910..d70fc094ff5 100644
--- a/src/gallium/drivers/i965simple/brw_clip.h
+++ b/src/gallium/drivers/i965simple/brw_clip.h
@@ -116,7 +116,7 @@ struct brw_clip_compile {
    unsigned last_mrf;
 
    unsigned header_position_offset;
-   unsigned offset[PIPE_ATTRIB_MAX];
+   unsigned offset[PIPE_MAX_ATTRIBS];
 };
 
 #define ATTR_SIZE  (4*4)
diff --git a/src/gallium/drivers/i965simple/brw_context.h b/src/gallium/drivers/i965simple/brw_context.h
index b83a13c3b6b..0c96ba17327 100644
--- a/src/gallium/drivers/i965simple/brw_context.h
+++ b/src/gallium/drivers/i965simple/brw_context.h
@@ -433,17 +433,17 @@ struct brw_cached_batch_item {
 

 

 

-/* Protect against a future where PIPE_ATTRIB_MAX > 32.  Wouldn't life

+/* Protect against a future where PIPE_MAX_ATTRIBS > 32.  Wouldn't life

  * be easier if C allowed arrays of packed elements?

  */

-#define ATTRIB_BIT_DWORDS  ((PIPE_ATTRIB_MAX+31)/32)

+#define ATTRIB_BIT_DWORDS  ((PIPE_MAX_ATTRIBS+31)/32)

 

 

 

 

 struct brw_vertex_info {

-   unsigned varying;  /* varying:1[PIPE_ATTRIB_MAX] */

-   unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_ATTRIB_MAX] */

+   unsigned varying;  /* varying:1[PIPE_MAX_ATTRIBS] */

+   unsigned sizes[ATTRIB_BIT_DWORDS * 2]; /* sizes:2[PIPE_MAX_ATTRIBS] */

 };

 

 

@@ -496,9 +496,9 @@ struct brw_context
       /* Arrays with buffer objects to copy non-bufferobj arrays into

        * for upload:

        */

-      const struct pipe_vertex_buffer *vbo_array[PIPE_ATTRIB_MAX];

+      const struct pipe_vertex_buffer *vbo_array[PIPE_MAX_ATTRIBS];

 

-      struct brw_vertex_element_state inputs[PIPE_ATTRIB_MAX];

+      struct brw_vertex_element_state inputs[PIPE_MAX_ATTRIBS];

 

 #define BRW_NR_UPLOAD_BUFS 17

 #define BRW_UPLOAD_INIT_SIZE (128*1024)

diff --git a/src/gallium/drivers/i965simple/brw_flush.c b/src/gallium/drivers/i965simple/brw_flush.c
index 5216c680cf6..e6001c30d94 100644
--- a/src/gallium/drivers/i965simple/brw_flush.c
+++ b/src/gallium/drivers/i965simple/brw_flush.c
@@ -36,14 +36,11 @@
 #include "brw_batch.h"
 
 
-/**
- * In future we may want a fence-like interface instead of finish.
- */
 static void brw_flush( struct pipe_context *pipe,
-			unsigned flags )
+                       unsigned flags,
+                       struct pipe_fence_handle **fence )
 {
    struct brw_context *brw = brw_context(pipe);
-   struct pipe_fence_handle *fence;
 
    /* Do we need to emit an MI_FLUSH command to flush the hardware
     * caches?
@@ -65,11 +62,7 @@ static void brw_flush( struct pipe_context *pipe,
 
    /* If there are no flags, just flush pending commands to hardware:
     */
-   FLUSH_BATCH( &fence );
-
-   if (flags & PIPE_FLUSH_WAIT) {
-//      brw->winsys->wait_fence(brw->winsys, fence);
-   }
+   FLUSH_BATCH( fence );
 }
 
 
diff --git a/src/gallium/drivers/i965simple/brw_state.c b/src/gallium/drivers/i965simple/brw_state.c
index f5efe9fc06d..376f1487b29 100644
--- a/src/gallium/drivers/i965simple/brw_state.c
+++ b/src/gallium/drivers/i965simple/brw_state.c
@@ -277,45 +277,49 @@ static void brw_set_viewport_state( struct pipe_context *pipe,
 }

 

 

-static void brw_set_vertex_buffer( struct pipe_context *pipe,

-				   unsigned index,

-				   const struct pipe_vertex_buffer *buffer )

+static void brw_set_vertex_buffers(struct pipe_context *pipe,

+				   unsigned count,

+				   const struct pipe_vertex_buffer *buffers)

 {

    struct brw_context *brw = brw_context(pipe);

-   brw->vb.vbo_array[index] = buffer;

+   memcpy(brw->vb.vbo_array, buffers, count * sizeof(buffers[0]));

 }

 

-static void brw_set_vertex_element(struct pipe_context *pipe,

-                                   unsigned index,

-                                   const struct pipe_vertex_element *element)

+static void brw_set_vertex_elements(struct pipe_context *pipe,

+                                    unsigned count,

+                                    const struct pipe_vertex_element *elements)

 {

    /* flush ? */

    struct brw_context *brw = brw_context(pipe);

+   uint i;

 

-   assert(index < PIPE_ATTRIB_MAX);

-   struct brw_vertex_element_state el;

-   memset(&el, 0, sizeof(el));

+   assert(count <= PIPE_MAX_ATTRIBS);

 

-   el.ve0.src_offset = element->src_offset;

-   el.ve0.src_format = brw_translate_surface_format(element->src_format);

-   el.ve0.valid = 1;

-   el.ve0.vertex_buffer_index = element->vertex_buffer_index;

+   for (i = 0; i < count; i++) {

+      struct brw_vertex_element_state el;

+      memset(&el, 0, sizeof(el));

 

-   el.ve1.dst_offset   = index * 4;

+      el.ve0.src_offset = elements[i].src_offset;

+      el.ve0.src_format = brw_translate_surface_format(elements[i].src_format);

+      el.ve0.valid = 1;

+      el.ve0.vertex_buffer_index = elements[i].vertex_buffer_index;

 

-   el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC;

-   el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC;

-   el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC;

-   el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC;

+      el.ve1.dst_offset   = i * 4;

 

-   switch (element->nr_components) {

-   case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0;

-   case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0;

-   case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT;

-      break;

-   }

+      el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_SRC;

+      el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_SRC;

+      el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_SRC;

+      el.ve1.vfcomponent0 = BRW_VFCOMPONENT_STORE_SRC;

 

-   brw->vb.inputs[index] = el;

+      switch (elements[i].nr_components) {

+      case 1: el.ve1.vfcomponent1 = BRW_VFCOMPONENT_STORE_0;

+      case 2: el.ve1.vfcomponent2 = BRW_VFCOMPONENT_STORE_0;

+      case 3: el.ve1.vfcomponent3 = BRW_VFCOMPONENT_STORE_1_FLT;

+         break;

+      }

+

+      brw->vb.inputs[i] = el;

+   }

 }

 

 

@@ -457,6 +461,6 @@ brw_init_state_functions( struct brw_context *brw )
    brw->pipe.set_scissor_state = brw_set_scissor_state;

    brw->pipe.set_sampler_textures = brw_set_sampler_textures;

    brw->pipe.set_viewport_state = brw_set_viewport_state;

-   brw->pipe.set_vertex_buffer = brw_set_vertex_buffer;

-   brw->pipe.set_vertex_element = brw_set_vertex_element;

+   brw->pipe.set_vertex_buffers = brw_set_vertex_buffers;

+   brw->pipe.set_vertex_elements = brw_set_vertex_elements;

 }

diff --git a/src/gallium/drivers/i965simple/brw_wm.c b/src/gallium/drivers/i965simple/brw_wm.c
index 1c4b5b5ede2..7fc5f59a98c 100644
--- a/src/gallium/drivers/i965simple/brw_wm.c
+++ b/src/gallium/drivers/i965simple/brw_wm.c
@@ -161,8 +161,7 @@ static void brw_wm_populate_key( struct brw_context *brw,
 
       if (unit) {
 
-	 if (unit->compare &&
-             unit->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+	 if (unit->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
 	    key->shadowtex_mask |= 1<<i;
 	 }
 	 if (t->Image[0][t->BaseLevel]->InternalFormat == GL_YCBCR_MESA)
diff --git a/src/gallium/drivers/i965simple/brw_wm.h b/src/gallium/drivers/i965simple/brw_wm.h
index a1ac0f504a6..b29c4393f01 100644
--- a/src/gallium/drivers/i965simple/brw_wm.h
+++ b/src/gallium/drivers/i965simple/brw_wm.h
@@ -76,7 +76,7 @@ struct brw_wm_prog_key {
 
 #define PROGRAM_INTERNAL_PARAM
 #define MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS 1024 /* 72 for GL_ARB_f_p */
-#define BRW_WM_MAX_INSN  (MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS*3 + PIPE_ATTRIB_MAX + 3)
+#define BRW_WM_MAX_INSN  (MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS*3 + PIPE_MAX_ATTRIBS + 3)
 #define BRW_WM_MAX_GRF   128		/* hardware limit */
 #define BRW_WM_MAX_VREG  (BRW_WM_MAX_INSN * 4)
 #define BRW_WM_MAX_REF   (BRW_WM_MAX_INSN * 12)
@@ -84,7 +84,7 @@ struct brw_wm_prog_key {
 #define BRW_WM_MAX_CONST 256
 #define BRW_WM_MAX_KILLS MAX_NV_FRAGMENT_PROGRAM_INSTRUCTIONS
 
-#define PAYLOAD_DEPTH     (PIPE_ATTRIB_MAX)
+#define PAYLOAD_DEPTH     (PIPE_MAX_ATTRIBS)
 
 #define MAX_IFSN 32
 #define MAX_LOOP_DEPTH 32
diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
index 16fb06f176f..e298ed37c36 100644
--- a/src/gallium/drivers/softpipe/sp_context.c
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -174,8 +174,8 @@ softpipe_create( struct pipe_screen *screen,
    softpipe->pipe.set_sampler_textures = softpipe_set_sampler_textures;
    softpipe->pipe.set_viewport_state = softpipe_set_viewport_state;
 
-   softpipe->pipe.set_vertex_buffer = softpipe_set_vertex_buffer;
-   softpipe->pipe.set_vertex_element = softpipe_set_vertex_element;
+   softpipe->pipe.set_vertex_buffers = softpipe_set_vertex_buffers;
+   softpipe->pipe.set_vertex_elements = softpipe_set_vertex_elements;
 
    softpipe->pipe.draw_arrays = softpipe_draw_arrays;
    softpipe->pipe.draw_elements = softpipe_draw_elements;
diff --git a/src/gallium/drivers/softpipe/sp_context.h b/src/gallium/drivers/softpipe/sp_context.h
index 19e6cfaf02e..dc9d0e6d5d6 100644
--- a/src/gallium/drivers/softpipe/sp_context.h
+++ b/src/gallium/drivers/softpipe/sp_context.h
@@ -77,8 +77,8 @@ struct softpipe_context {
    struct pipe_scissor_state scissor;
    struct pipe_texture *texture[PIPE_MAX_SAMPLERS];
    struct pipe_viewport_state viewport;
-   struct pipe_vertex_buffer vertex_buffer[PIPE_ATTRIB_MAX];
-   struct pipe_vertex_element vertex_element[PIPE_ATTRIB_MAX];
+   struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
+   struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
    unsigned dirty;
 
    unsigned num_samplers;
@@ -92,7 +92,7 @@ struct softpipe_context {
    /*
     * Mapped vertex buffers
     */
-   ubyte *mapped_vbuffer[PIPE_ATTRIB_MAX];
+   ubyte *mapped_vbuffer[PIPE_MAX_ATTRIBS];
    
    /** Mapped constant buffers */
    void *mapped_constants[PIPE_SHADER_TYPES];
diff --git a/src/gallium/drivers/softpipe/sp_draw_arrays.c b/src/gallium/drivers/softpipe/sp_draw_arrays.c
index 5b5a0fe5735..ab54050d3fe 100644
--- a/src/gallium/drivers/softpipe/sp_draw_arrays.c
+++ b/src/gallium/drivers/softpipe/sp_draw_arrays.c
@@ -125,7 +125,7 @@ softpipe_draw_elements(struct pipe_context *pipe,
    /*
     * Map vertex buffers
     */
-   for (i = 0; i < PIPE_ATTRIB_MAX; i++) {
+   for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
       if (sp->vertex_buffer[i].buffer) {
          void *buf
             = pipe->winsys->buffer_map(pipe->winsys,
@@ -153,7 +153,7 @@ softpipe_draw_elements(struct pipe_context *pipe,
    /*
     * unmap vertex/index buffers - will cause draw module to flush
     */
-   for (i = 0; i < PIPE_ATTRIB_MAX; i++) {
+   for (i = 0; i < PIPE_MAX_ATTRIBS; i++) {
       if (sp->vertex_buffer[i].buffer) {
          draw_set_mapped_vertex_buffer(draw, i, NULL);
          pipe->winsys->buffer_unmap(pipe->winsys, sp->vertex_buffer[i].buffer);
diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c
index 2cbd0d7cabe..0625b69099b 100644
--- a/src/gallium/drivers/softpipe/sp_flush.c
+++ b/src/gallium/drivers/softpipe/sp_flush.c
@@ -40,13 +40,10 @@
 #include "sp_winsys.h"
 
 
-/* There will be actual work to do here.  In future we may want a
- * fence-like interface instead of finish, and perhaps flush will take
- * flags to indicate what type of flush is required.
- */
 void
 softpipe_flush( struct pipe_context *pipe,
-		unsigned flags )
+		unsigned flags,
+                struct pipe_fence_handle **fence )
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
    uint i;
@@ -72,5 +69,8 @@ softpipe_flush( struct pipe_context *pipe,
     * to unmap surfaces when flushing.
     */
    softpipe_unmap_surfaces(softpipe);
+
+   if (fence)
+      *fence = NULL;
 }
 
diff --git a/src/gallium/drivers/softpipe/sp_flush.h b/src/gallium/drivers/softpipe/sp_flush.h
index 34ec617866b..68d9b5fa835 100644
--- a/src/gallium/drivers/softpipe/sp_flush.h
+++ b/src/gallium/drivers/softpipe/sp_flush.h
@@ -29,7 +29,9 @@
 #define SP_FLUSH_H
 
 struct pipe_context;
+struct pipe_fence_handle;
 
-void softpipe_flush(struct pipe_context *pipe, unsigned flags );
+void softpipe_flush(struct pipe_context *pipe, unsigned flags,
+                    struct pipe_fence_handle **fence);
 
 #endif
diff --git a/src/gallium/drivers/softpipe/sp_prim_setup.c b/src/gallium/drivers/softpipe/sp_prim_setup.c
index 6a81e4d8cc2..c7eb12b3bb7 100644
--- a/src/gallium/drivers/softpipe/sp_prim_setup.c
+++ b/src/gallium/drivers/softpipe/sp_prim_setup.c
@@ -44,6 +44,7 @@
 #include "pipe/p_shader_tokens.h"
 
 #define DEBUG_VERTS 0
+#define DEBUG_FRAGS 0
 
 /**
  * Triangle edge info
@@ -92,6 +93,11 @@ struct setup_stage {
       unsigned y_flags;
       unsigned mask;     /**< mask of MASK_BOTTOM/TOP_LEFT/RIGHT bits */
    } span;
+
+#if DEBUG_FRAGS
+   uint numFragsEmitted;  /**< per primitive */
+   uint numFragsWritten;  /**< per primitive */
+#endif
 };
 
 
@@ -160,7 +166,20 @@ emit_quad( struct setup_stage *setup, int x, int y, unsigned mask )
    setup->quad.x0 = x;
    setup->quad.y0 = y;
    setup->quad.mask = mask;
+#if DEBUG_FRAGS
+   if (mask & 1) setup->numFragsEmitted++;
+   if (mask & 2) setup->numFragsEmitted++;
+   if (mask & 4) setup->numFragsEmitted++;
+   if (mask & 8) setup->numFragsEmitted++;
+#endif
    sp->quad.first->run(sp->quad.first, &setup->quad);
+#if DEBUG_FRAGS
+   mask = setup->quad.mask;
+   if (mask & 1) setup->numFragsWritten++;
+   if (mask & 2) setup->numFragsWritten++;
+   if (mask & 4) setup->numFragsWritten++;
+   if (mask & 8) setup->numFragsWritten++;
+#endif
 }
 
 
@@ -674,6 +693,11 @@ static void setup_tri( struct draw_stage *stage,
    debug_printf("%s\n", __FUNCTION__ );
    */
 
+#if DEBUG_FRAGS
+   setup->numFragsEmitted = 0;
+   setup->numFragsWritten = 0;
+#endif
+
    setup_sort_vertices( setup, prim );
    setup_tri_coefficients( setup );
    setup_tri_edges( setup );
@@ -702,6 +726,12 @@ static void setup_tri( struct draw_stage *stage,
    }
 
    flush_spans( setup );
+
+#if DEBUG_FRAGS
+   printf("Tri: %u frags emitted, %u written\n",
+          setup->numFragsEmitted,
+          setup->numFragsWritten);
+#endif
 }
 
 
diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c
index 861285101fd..a73df31383f 100644
--- a/src/gallium/drivers/softpipe/sp_quad_fs.c
+++ b/src/gallium/drivers/softpipe/sp_quad_fs.c
@@ -113,15 +113,18 @@ shade_quad(
       }
    }
    else {
-      /* copy input Z (which was interpolated by the executor) to output Z */
-      uint i;
-      for (i = 0; i < 4; i++) {
-         quad->outputs.depth[i] = machine->Inputs[0].xyzw[2].f[i];
-         /* XXX not sure the above line is always correct.  The following
-          * might be better:
-         quad->outputs.depth[i] = machine->QuadPos.xyzw[2].f[i];
-          */
-      }
+      /* compute Z values now, as in the quad earlyz stage */
+      /* XXX we should really only do this if the earlyz stage is not used */
+      const float fx = (float) quad->x0;
+      const float fy = (float) quad->y0;
+      const float dzdx = quad->posCoef->dadx[2];
+      const float dzdy = quad->posCoef->dady[2];
+      const float z0 = quad->posCoef->a0[2] + dzdx * fx + dzdy * fy;
+
+      quad->outputs.depth[0] = z0;
+      quad->outputs.depth[1] = z0 + dzdx;
+      quad->outputs.depth[2] = z0 + dzdy;
+      quad->outputs.depth[3] = z0 + dzdx + dzdy;
    }
 
    /* shader may cull fragments */
@@ -185,8 +188,8 @@ struct quad_stage *sp_quad_shade_stage( struct softpipe_context *softpipe )
    uint i;
 
    /* allocate storage for program inputs/outputs, aligned to 16 bytes */
-   qss->inputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->inputs) + 16);
-   qss->outputs = MALLOC(PIPE_ATTRIB_MAX * sizeof(*qss->outputs) + 16);
+   qss->inputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->inputs) + 16);
+   qss->outputs = MALLOC(PIPE_MAX_ATTRIBS * sizeof(*qss->outputs) + 16);
    qss->machine.Inputs = align16(qss->inputs);
    qss->machine.Outputs = align16(qss->outputs);
 
diff --git a/src/gallium/drivers/softpipe/sp_state.h b/src/gallium/drivers/softpipe/sp_state.h
index 0bb1095aece..6e6501f5bc4 100644
--- a/src/gallium/drivers/softpipe/sp_state.h
+++ b/src/gallium/drivers/softpipe/sp_state.h
@@ -151,13 +151,13 @@ void softpipe_set_sampler_textures( struct pipe_context *,
 void softpipe_set_viewport_state( struct pipe_context *,
                                   const struct pipe_viewport_state * );
 
-void softpipe_set_vertex_element(struct pipe_context *,
-                                 unsigned index,
-                                 const struct pipe_vertex_element *);
+void softpipe_set_vertex_elements(struct pipe_context *,
+                                  unsigned count,
+                                  const struct pipe_vertex_element *);
 
-void softpipe_set_vertex_buffer(struct pipe_context *,
-                                unsigned index,
-                                const struct pipe_vertex_buffer *);
+void softpipe_set_vertex_buffers(struct pipe_context *,
+                                 unsigned count,
+                                 const struct pipe_vertex_buffer *);
 
 
 void softpipe_update_derived( struct softpipe_context *softpipe );
diff --git a/src/gallium/drivers/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c
index f01a10de3b4..e0230e16a4e 100644
--- a/src/gallium/drivers/softpipe/sp_state_vertex.c
+++ b/src/gallium/drivers/softpipe/sp_state_vertex.c
@@ -37,28 +37,35 @@
 
 
 void
-softpipe_set_vertex_element(struct pipe_context *pipe,
-                            unsigned index,
-                            const struct pipe_vertex_element *attrib)
+softpipe_set_vertex_elements(struct pipe_context *pipe,
+                             unsigned count,
+                             const struct pipe_vertex_element *attribs)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
-   assert(index < PIPE_ATTRIB_MAX);
-   softpipe->vertex_element[index] = *attrib; /* struct copy */
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(softpipe->vertex_element, attribs,
+          count * sizeof(struct pipe_vertex_element));
+
    softpipe->dirty |= SP_NEW_VERTEX;
 
-   draw_set_vertex_element(softpipe->draw, index, attrib);
+   draw_set_vertex_elements(softpipe->draw, count, attribs);
 }
 
 
 void
-softpipe_set_vertex_buffer(struct pipe_context *pipe,
-                           unsigned index,
-                           const struct pipe_vertex_buffer *buffer)
+softpipe_set_vertex_buffers(struct pipe_context *pipe,
+                            unsigned count,
+                            const struct pipe_vertex_buffer *buffers)
 {
    struct softpipe_context *softpipe = softpipe_context(pipe);
-   assert(index < PIPE_ATTRIB_MAX);
-   softpipe->vertex_buffer[index] = *buffer; /* struct copy */
+
+   assert(count <= PIPE_MAX_ATTRIBS);
+
+   memcpy(softpipe->vertex_buffer, buffers, count * sizeof(buffers[0]));
+
    softpipe->dirty |= SP_NEW_VERTEX;
 
-   draw_set_vertex_buffer(softpipe->draw, index, buffer);
+   draw_set_vertex_buffers(softpipe->draw, count, buffers);
 }