Merge commit 'origin/gallium-0.1' into gallium-tex-surfaces

Conflicts:

	src/mesa/state_tracker/st_atom_sampler.c
	src/mesa/state_tracker/st_cb_texture.c

15 files changed, 257 insertions, 248 deletions

diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c
index 5d5125f7cb..826b432f09 100644
--- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c
+++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.c
@@ -88,6 +88,10 @@
 #define TEMP_OUTPUT_C      TGSI_EXEC_TEMP_OUTPUT_C
 #define TEMP_PRIMITIVE_I   TGSI_EXEC_TEMP_PRIMITIVE_I
 #define TEMP_PRIMITIVE_C   TGSI_EXEC_TEMP_PRIMITIVE_C
+#define TEMP_3_I           TGSI_EXEC_TEMP_THREE_I
+#define TEMP_3_C           TGSI_EXEC_TEMP_THREE_C
+#define TEMP_HALF_I        TGSI_EXEC_TEMP_HALF_I
+#define TEMP_HALF_C        TGSI_EXEC_TEMP_HALF_C
 #define TEMP_R0            TGSI_EXEC_TEMP_R0
 
 #define FOR_EACH_CHANNEL(CHAN)\
@@ -262,6 +266,8 @@ tgsi_exec_machine_init(
       mach->Temps[TEMP_2_I].xyzw[TEMP_2_C].f[i] = 2.0f;
       mach->Temps[TEMP_128_I].xyzw[TEMP_128_C].f[i] = 128.0f;
       mach->Temps[TEMP_M128_I].xyzw[TEMP_M128_C].f[i] = -128.0f;
+      mach->Temps[TEMP_3_I].xyzw[TEMP_3_C].f[i] = 3.0f;
+      mach->Temps[TEMP_HALF_I].xyzw[TEMP_HALF_C].f[i] = 0.5f;
    }
 }
 
diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h
index 92e2e5e985..19bd78df3d 100644
--- a/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h
+++ b/src/gallium/auxiliary/tgsi/exec/tgsi_exec.h
@@ -133,9 +133,15 @@ struct tgsi_exec_labels
 #define TGSI_EXEC_TEMP_PRIMITIVE_I  34
 #define TGSI_EXEC_TEMP_PRIMITIVE_C  2
 
-#define TGSI_EXEC_TEMP_R0           35
+#define TGSI_EXEC_TEMP_THREE_I      34
+#define TGSI_EXEC_TEMP_THREE_C      3
 
-#define TGSI_EXEC_NUM_TEMPS   (32 + 4)
+#define TGSI_EXEC_TEMP_HALF_I       35
+#define TGSI_EXEC_TEMP_HALF_C       0
+
+#define TGSI_EXEC_TEMP_R0           36
+
+#define TGSI_EXEC_NUM_TEMPS   (32 + 5)
 #define TGSI_EXEC_NUM_ADDRS   1
 #define TGSI_EXEC_NUM_IMMEDIATES  256
 
diff --git a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c
index 2fd76a3072..8018bd7fa4 100755
--- a/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c
+++ b/src/gallium/auxiliary/tgsi/exec/tgsi_sse2.c
@@ -36,7 +36,11 @@
 
 #ifdef PIPE_ARCH_X86
 
-#define HIGH_PRECISION 1  /* for 1/sqrt() */
+/* for 1/sqrt()
+ *
+ * This costs about 100fps (close to 10%) in gears:
+ */
+#define HIGH_PRECISION 1
 
 
 #define FOR_EACH_CHANNEL( CHAN )\
@@ -757,20 +761,6 @@ emit_rcp (
       make_xmm( xmm_src ) );
 }
 
-#if HIGH_PRECISION
-static void XSTDCALL
-rsqrt4f(
-   float *store )
-{
-   const unsigned X = 0;
-
-   store[X + 0] = 1.0F / sqrtf( store[X + 0] );
-   store[X + 1] = 1.0F / sqrtf( store[X + 1] );
-   store[X + 2] = 1.0F / sqrtf( store[X + 2] );
-   store[X + 3] = 1.0F / sqrtf( store[X + 3] );
-}
-#endif
-
 static void
 emit_rsqrt(
    struct x86_function *func,
@@ -778,13 +768,6 @@ emit_rsqrt(
    unsigned xmm_src )
 {
 #if HIGH_PRECISION
-#if 1
-   emit_func_call_dst_src(
-      func,
-      xmm_dst,
-      xmm_src,
-      rsqrt4f );
-#else
    /* Although rsqrtps() and rcpps() are low precision on some/all SSE
     * implementations, it is possible to improve its precision at
     * fairly low cost, using a newton/raphson step, as below:
@@ -794,21 +777,25 @@ emit_rsqrt(
     *
     * See: http://softwarecommunity.intel.com/articles/eng/1818.htm
     */
-   /* This is some code that woudl do the above for a scalar 'a'.  We
-    * obviously are interested in a vector version:
-    *
-    * movss   xmm3, a;
-    * movss   xmm1, half;
-    * movss   xmm2, three;
-    * rsqrtss xmm0, xmm3;
-    * mulss   xmm3, xmm0;
-    * mulss   xmm1, xmm0;
-    * mulss   xmm3, xmm0;
-    * subss   xmm2, xmm3;
-    * mulss   xmm1, xmm2;
-    * movss   x,    xmm1;
-    */
-#endif
+   {
+      struct x86_reg dst = make_xmm( xmm_dst );
+      struct x86_reg src = make_xmm( xmm_src );
+      struct x86_reg tmp0 = make_xmm( 2 );
+      struct x86_reg tmp1 = make_xmm( 3 );
+
+      assert( xmm_dst != xmm_src );
+      assert( xmm_dst != 2 && xmm_dst != 3 );
+      assert( xmm_src != 2 && xmm_src != 3 );
+
+      sse_movaps(  func, dst,  get_temp( TGSI_EXEC_TEMP_HALF_I, TGSI_EXEC_TEMP_HALF_C ) );
+      sse_movaps(  func, tmp0, get_temp( TGSI_EXEC_TEMP_THREE_I, TGSI_EXEC_TEMP_THREE_C ) );
+      sse_rsqrtps( func, tmp1, src  );
+      sse_mulps(   func, src,  tmp1 );
+      sse_mulps(   func, dst,  tmp1 );
+      sse_mulps(   func, src,  tmp1 );
+      sse_subps(   func, tmp0, src  );
+      sse_mulps(   func, dst,  tmp0 );
+   }
 #else
    /* On Intel CPUs at least, this is only accurate to 12 bits -- not
     * good enough.
@@ -1295,9 +1282,9 @@ emit_instruction(
    case TGSI_OPCODE_RSQ:
    /* TGSI_OPCODE_RECIPSQRT */
       FETCH( func, *inst, 0, 0, CHAN_X );
-      emit_rsqrt( func, 0, 0 );
+      emit_rsqrt( func, 1, 0 );
       FOR_EACH_DST0_ENABLED_CHANNEL( *inst, chan_index ) {
-         STORE( func, *inst, 0, 0, chan_index );
+         STORE( func, *inst, 1, 0, chan_index );
       }
       break;
 
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c
index 056ae829ae..7d71aefda9 100644
--- a/src/gallium/auxiliary/util/u_gen_mipmap.c
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.c
@@ -783,23 +783,23 @@ set_vertex_data(struct gen_mipmap_state *ctx, float width, float height)
 {
    void *buf;
 
-   ctx->vertices[0][0][0] = -0.5f; /*x*/
-   ctx->vertices[0][0][1] = -0.5f; /*y*/
+   ctx->vertices[0][0][0] = 0.0f; /*x*/
+   ctx->vertices[0][0][1] = 0.0f; /*y*/
    ctx->vertices[0][1][0] = 0.0f; /*s*/
    ctx->vertices[0][1][1] = 0.0f; /*t*/
 
-   ctx->vertices[1][0][0] = width - 0.5f; /*x*/
-   ctx->vertices[1][0][1] = -0.5f;  /*y*/
-   ctx->vertices[1][1][0] = 1.0f; /*s*/
-   ctx->vertices[1][1][1] = 0.0f; /*t*/
+   ctx->vertices[1][0][0] = width;
+   ctx->vertices[1][0][1] = 0.0f;
+   ctx->vertices[1][1][0] = 1.0f;
+   ctx->vertices[1][1][1] = 0.0f;
 
-   ctx->vertices[2][0][0] = width - 0.5f;
-   ctx->vertices[2][0][1] = height - 0.5f;
+   ctx->vertices[2][0][0] = width;
+   ctx->vertices[2][0][1] = height;
    ctx->vertices[2][1][0] = 1.0f;
    ctx->vertices[2][1][1] = 1.0f;
 
-   ctx->vertices[3][0][0] = -0.5f;
-   ctx->vertices[3][0][1] = height - 0.5f;
+   ctx->vertices[3][0][0] = 0.0f;
+   ctx->vertices[3][0][1] = height;
    ctx->vertices[3][1][0] = 0.0f;
    ctx->vertices[3][1][1] = 1.0f;
 
diff --git a/src/gallium/drivers/softpipe/sp_tex_sample.c b/src/gallium/drivers/softpipe/sp_tex_sample.c
index 5b63f97997..be0b57d9fa 100644
--- a/src/gallium/drivers/softpipe/sp_tex_sample.c
+++ b/src/gallium/drivers/softpipe/sp_tex_sample.c
@@ -1051,5 +1051,19 @@ sp_get_samples(struct tgsi_sampler *sampler,
    default:
       assert(0);
    }
+
+#if 0 /* DEBUG */
+   {
+      int i;
+      printf("Sampled at %f, %f, %f:\n", s[0], t[0], p[0]);
+      for (i = 0; i < 4; i++) {
+         printf("Frag %d: %f %f %f %f\n", i,
+                rgba[0][i],
+                rgba[1][i],
+                rgba[2][i],
+                rgba[3][i]);
+      }
+   }
+#endif
 }
 
diff --git a/src/gallium/drivers/softpipe/sp_tile_cache.c b/src/gallium/drivers/softpipe/sp_tile_cache.c
index 142faf5074..28c29da87c 100644
--- a/src/gallium/drivers/softpipe/sp_tile_cache.c
+++ b/src/gallium/drivers/softpipe/sp_tile_cache.c
@@ -133,7 +133,7 @@ sp_destroy_tile_cache(struct softpipe_tile_cache *tc)
    uint pos;
 
    for (pos = 0; pos < NUM_ENTRIES; pos++) {
-      //assert(tc->entries[pos].x < 0);
+      /*assert(tc->entries[pos].x < 0);*/
    }
    if (tc->surface) {
       pipe_surface_reference(&tc->surface, NULL);
@@ -338,8 +338,8 @@ sp_tile_cache_flush_clear(struct pipe_context *pipe,
       for (x = 0; x < w; x += TILE_SIZE) {
          if (is_clear_flag_set(tc->clear_flags, x, y)) {
             pipe_put_tile_raw(pipe, ps,
-                           x, y, TILE_SIZE, TILE_SIZE,
-                           tc->tile.data.color32, 0/*STRIDE*/);
+                              x, y, TILE_SIZE, TILE_SIZE,
+                              tc->tile.data.color32, 0/*STRIDE*/);
 
             /* do this? */
             clear_clear_flag(tc->clear_flags, x, y);
@@ -373,8 +373,8 @@ sp_flush_tile_cache(struct softpipe_context *softpipe,
          if (tile->x >= 0) {
             if (tc->depth_stencil) {
                pipe_put_tile_raw(pipe, ps,
-                              tile->x, tile->y, TILE_SIZE, TILE_SIZE,
-                              tile->data.depth32, 0/*STRIDE*/);
+                                 tile->x, tile->y, TILE_SIZE, TILE_SIZE,
+                                 tile->data.depth32, 0/*STRIDE*/);
             }
             else {
                pipe_put_tile_rgba(pipe, ps,
@@ -391,7 +391,7 @@ sp_flush_tile_cache(struct softpipe_context *softpipe,
 #endif
    }
    else if (tc->texture) {
-      /* caching a texture, mark all entries as embpy */
+      /* caching a texture, mark all entries as empty */
       for (pos = 0; pos < NUM_ENTRIES; pos++) {
          tc->entries[pos].x = -1;
       }
diff --git a/src/gallium/include/pipe/p_defines.h b/src/gallium/include/pipe/p_defines.h
index 2e77fb42a8..83330ef22f 100644
--- a/src/gallium/include/pipe/p_defines.h
+++ b/src/gallium/include/pipe/p_defines.h
@@ -274,7 +274,7 @@ enum pipe_texture_target {
 #define PIPE_CAP_MAX_POINT_WIDTH_AA      17
 #define PIPE_CAP_MAX_TEXTURE_ANISOTROPY  18
 #define PIPE_CAP_MAX_TEXTURE_LOD_BIAS    19
-#define PIPE_CAP_BITMAP_TEXCOORD_BIAS    20
+
 
 #ifdef __cplusplus
 }
diff --git a/src/gallium/winsys/xlib/xm_winsys.c b/src/gallium/winsys/xlib/xm_winsys.c
index 8ca90ef4c6..a70752428a 100644
--- a/src/gallium/winsys/xlib/xm_winsys.c
+++ b/src/gallium/winsys/xlib/xm_winsys.c
@@ -58,6 +58,7 @@
 
 
 /**
+ * Subclass of pipe_buffer for Xlib winsys.
  * Low-level OS/window system memory buffer
  */
 struct xm_buffer
@@ -74,12 +75,10 @@ struct xm_buffer
 #endif
 };
 
-#if defined(USE_XSHM) && !defined(XFree86Server)
-# define XSHM_ENABLED(b) ((b)->shm)
-#else
-# define XSHM_ENABLED(b) 0
-#endif
 
+/**
+ * Subclass of pipe_surface for Xlib winsys
+ */
 struct xmesa_surface
 {
    struct pipe_surface surface;
@@ -89,6 +88,9 @@ struct xmesa_surface
 };
 
 
+/**
+ * Subclass of pipe_winsys for Xlib winsys
+ */
 struct xmesa_pipe_winsys
 {
    struct pipe_winsys base;
@@ -97,23 +99,16 @@ struct xmesa_pipe_winsys
 };
 
 
-static void alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb,
-    unsigned width, unsigned height);
 
 /** Cast wrapper */
 static INLINE struct xmesa_surface *
 xmesa_surface(struct pipe_surface *ps)
 {
-//   assert(0);
    return (struct xmesa_surface *) ps;
 }
 
 
-
-/**
- * Turn the softpipe opaque buffer pointer into a dri_bufmgr opaque
- * buffer pointer...
- */
+/** Cast wrapper */
 static INLINE struct xm_buffer *
 xm_buffer( struct pipe_buffer *buf )
 {
@@ -121,6 +116,130 @@ xm_buffer( struct pipe_buffer *buf )
 }
 
 
+/**
+ * X Shared Memory Image extension code
+ */
+#if defined(USE_XSHM) && !defined(XFree86Server)
+
+#define XSHM_ENABLED(b) ((b)->shm)
+
+static volatile int mesaXErrorFlag = 0;
+
+/**
+ * Catches potential Xlib errors.
+ */
+static int
+mesaHandleXError(XMesaDisplay *dpy, XErrorEvent *event)
+{
+   (void) dpy;
+   (void) event;
+   mesaXErrorFlag = 1;
+   return 0;
+}
+
+
+static GLboolean alloc_shm(struct xm_buffer *buf, unsigned size)
+{
+   XShmSegmentInfo *const shminfo = & buf->shminfo;
+
+   shminfo->shmid = shmget(IPC_PRIVATE, size, IPC_CREAT|0777);
+   if (shminfo->shmid < 0) {
+      return GL_FALSE;
+   }
+
+   shminfo->shmaddr = (char *) shmat(shminfo->shmid, 0, 0);
+   if (shminfo->shmaddr == (char *) -1) {
+      shmctl(shminfo->shmid, IPC_RMID, 0);
+      return GL_FALSE;
+   }
+
+   shminfo->readOnly = False;
+   return GL_TRUE;
+}
+
+
+/**
+ * Allocate a shared memory XImage back buffer for the given XMesaBuffer.
+ */
+static void
+alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb,
+                 unsigned width, unsigned height)
+{
+   /*
+    * We have to do a _lot_ of error checking here to be sure we can
+    * really use the XSHM extension.  It seems different servers trigger
+    * errors at different points if the extension won't work.  Therefore
+    * we have to be very careful...
+    */
+#if 0
+   GC gc;
+#endif
+   int (*old_handler)(XMesaDisplay *, XErrorEvent *);
+
+   b->tempImage = XShmCreateImage(xmb->xm_visual->display,
+                                  xmb->xm_visual->visinfo->visual,
+                                  xmb->xm_visual->visinfo->depth,
+                                  ZPixmap,
+                                  NULL,
+                                  &b->shminfo,
+                                  width, height);
+   if (b->tempImage == NULL) {
+      b->shm = 0;
+      return;
+   }
+
+
+   mesaXErrorFlag = 0;
+   old_handler = XSetErrorHandler(mesaHandleXError);
+   /* This may trigger the X protocol error we're ready to catch: */
+   XShmAttach(xmb->xm_visual->display, &b->shminfo);
+   XSync(xmb->xm_visual->display, False);
+
+   if (mesaXErrorFlag) {
+      /* we are on a remote display, this error is normal, don't print it */
+      XFlush(xmb->xm_visual->display);
+      mesaXErrorFlag = 0;
+      XDestroyImage(b->tempImage);
+      b->tempImage = NULL;
+      b->shm = 0;
+      (void) XSetErrorHandler(old_handler);
+      return;
+   }
+
+
+   /* Finally, try an XShmPutImage to be really sure the extension works */
+#if 0
+   gc = XCreateGC(xmb->xm_visual->display, xmb->drawable, 0, NULL);
+   XShmPutImage(xmb->xm_visual->display, xmb->drawable, gc,
+                b->tempImage, 0, 0, 0, 0, 1, 1 /*one pixel*/, False);
+   XSync(xmb->xm_visual->display, False);
+   XFreeGC(xmb->xm_visual->display, gc);
+   (void) XSetErrorHandler(old_handler);
+   if (mesaXErrorFlag) {
+      XFlush(xmb->xm_visual->display);
+      mesaXErrorFlag = 0;
+      XDestroyImage(b->tempImage);
+      b->tempImage = NULL;
+      b->shm = 0;
+      return;
+   }
+#endif
+}
+
+#else
+
+#define XSHM_ENABLED(b) 0
+
+static void
+alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb,
+                 unsigned width, unsigned height)
+{
+   b->shm = 0;
+}
+#endif /* USE_XSHM */
+
+
+
 
 /* Most callbacks map direcly onto dri_bufmgr operations:
  */
@@ -297,119 +416,6 @@ xm_get_name(struct pipe_winsys *pws)
 }
 
 
-#if defined(USE_XSHM) && !defined(XFree86Server)
-static volatile int mesaXErrorFlag = 0;
-
-/**
- * Catches potential Xlib errors.
- */
-static int
-mesaHandleXError(XMesaDisplay *dpy, XErrorEvent *event)
-{
-   (void) dpy;
-   (void) event;
-   mesaXErrorFlag = 1;
-   return 0;
-}
-
-
-static GLboolean alloc_shm(struct xm_buffer *buf, unsigned size)
-{
-   XShmSegmentInfo *const shminfo = & buf->shminfo;
-
-   shminfo->shmid = shmget(IPC_PRIVATE, size, IPC_CREAT|0777);
-   if (shminfo->shmid < 0) {
-      return GL_FALSE;
-   }
-
-   shminfo->shmaddr = (char *) shmat(shminfo->shmid, 0, 0);
-   if (shminfo->shmaddr == (char *) -1) {
-      shmctl(shminfo->shmid, IPC_RMID, 0);
-      return GL_FALSE;
-   }
-
-   shminfo->readOnly = False;
-   return GL_TRUE;
-}
-
-
-/**
- * Allocate a shared memory XImage back buffer for the given XMesaBuffer.
- */
-static void
-alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb,
-                 unsigned width, unsigned height)
-{
-   /*
-    * We have to do a _lot_ of error checking here to be sure we can
-    * really use the XSHM extension.  It seems different servers trigger
-    * errors at different points if the extension won't work.  Therefore
-    * we have to be very careful...
-    */
-#if 0
-   GC gc;
-#endif
-   int (*old_handler)(XMesaDisplay *, XErrorEvent *);
-
-   b->tempImage = XShmCreateImage(xmb->xm_visual->display,
-                                  xmb->xm_visual->visinfo->visual,
-                                  xmb->xm_visual->visinfo->depth,
-                                  ZPixmap,
-                                  NULL,
-                                  &b->shminfo,
-                                  width, height);
-   if (b->tempImage == NULL) {
-      b->shm = 0;
-      return;
-   }
-
-
-   mesaXErrorFlag = 0;
-   old_handler = XSetErrorHandler(mesaHandleXError);
-   /* This may trigger the X protocol error we're ready to catch: */
-   XShmAttach(xmb->xm_visual->display, &b->shminfo);
-   XSync(xmb->xm_visual->display, False);
-
-   if (mesaXErrorFlag) {
-      /* we are on a remote display, this error is normal, don't print it */
-      XFlush(xmb->xm_visual->display);
-      mesaXErrorFlag = 0;
-      XDestroyImage(b->tempImage);
-      b->tempImage = NULL;
-      b->shm = 0;
-      (void) XSetErrorHandler(old_handler);
-      return;
-   }
-
-
-   /* Finally, try an XShmPutImage to be really sure the extension works */
-#if 0
-   gc = XCreateGC(xmb->xm_visual->display, xmb->drawable, 0, NULL);
-   XShmPutImage(xmb->xm_visual->display, xmb->drawable, gc,
-                b->tempImage, 0, 0, 0, 0, 1, 1 /*one pixel*/, False);
-   XSync(xmb->xm_visual->display, False);
-   XFreeGC(xmb->xm_visual->display, gc);
-   (void) XSetErrorHandler(old_handler);
-   if (mesaXErrorFlag) {
-      XFlush(xmb->xm_visual->display);
-      mesaXErrorFlag = 0;
-      XDestroyImage(b->tempImage);
-      b->tempImage = NULL;
-      b->shm = 0;
-      return;
-   }
-#endif
-}
-#else
-static void
-alloc_shm_ximage(struct xm_buffer *b, struct xmesa_buffer *xmb,
-                 unsigned width, unsigned height)
-{
-   b->shm = 0;
-}
-#endif
-
-
 static struct pipe_buffer *
 xm_buffer_create(struct pipe_winsys *pws, 
                  unsigned alignment, 
@@ -602,7 +608,7 @@ xmesa_get_pipe_winsys_aub(struct xmesa_visual *xm_vis)
 }
 
 
-struct pipe_winsys *
+static struct pipe_winsys *
 xmesa_get_pipe_winsys(struct xmesa_visual *xm_vis)
 {
    static struct xmesa_pipe_winsys *ws = NULL;
diff --git a/src/mesa/main/imports.c b/src/mesa/main/imports.c
index d8d35af15e..d798f80e25 100644
--- a/src/mesa/main/imports.c
+++ b/src/mesa/main/imports.c
@@ -542,26 +542,24 @@ int
 _mesa_ffs(int i)
 {
 #if (defined(_WIN32) && !defined(__MINGW32__) ) || defined(__IBMC__) || defined(__IBMCPP__)
-   register int bit = 0;
-   if (i != 0) {
-      if ((i & 0xffff) == 0) {
-         bit += 16;
-         i >>= 16;
-      }
-      if ((i & 0xff) == 0) {
-         bit += 8;
-         i >>= 8;
-      }
-      if ((i & 0xf) == 0) {
-         bit += 4;
-         i >>= 4;
-      }
-      while ((i & 1) == 0) {
-         bit++;
-         i >>= 1;
-      }
+   register int bit = 1;
+   if ((i & 0xffff) == 0) {
+      bit += 16;
+      i >>= 16;
+   }
+   if ((i & 0xff) == 0) {
+      bit += 8;
+      i >>= 8;
+   }
+   if ((i & 0xf) == 0) {
+      bit += 4;
+      i >>= 4;
+   }
+   if ((i & 0x3) == 0) {
+      bit += 2;
+      i >>= 2;
    }
-   return bit;
+   return (i) ? (bit + ((i + 1) & 0x01)) : 0;
 #else
    return ffs(i);
 #endif
diff --git a/src/mesa/state_tracker/st_atom.h b/src/mesa/state_tracker/st_atom.h
index 3a63e2dec0..c6c6eba812 100644
--- a/src/mesa/state_tracker/st_atom.h
+++ b/src/mesa/state_tracker/st_atom.h
@@ -44,20 +44,20 @@ void st_destroy_atoms( struct st_context *st );
 void st_validate_state( struct st_context *st );
 
 
-const struct st_tracked_state st_update_framebuffer;
-const struct st_tracked_state st_update_clip;
-const struct st_tracked_state st_update_depth_stencil_alpha;
-const struct st_tracked_state st_update_shader;
-const struct st_tracked_state st_update_rasterizer;
-const struct st_tracked_state st_update_polygon_stipple;
-const struct st_tracked_state st_update_viewport;
-const struct st_tracked_state st_update_scissor;
-const struct st_tracked_state st_update_blend;
-const struct st_tracked_state st_update_sampler;
-const struct st_tracked_state st_update_texture;
-const struct st_tracked_state st_update_fs_constants;
-const struct st_tracked_state st_update_vs_constants;
-const struct st_tracked_state st_update_pixel_transfer;
+extern const struct st_tracked_state st_update_framebuffer;
+extern const struct st_tracked_state st_update_clip;
+extern const struct st_tracked_state st_update_depth_stencil_alpha;
+extern const struct st_tracked_state st_update_shader;
+extern const struct st_tracked_state st_update_rasterizer;
+extern const struct st_tracked_state st_update_polygon_stipple;
+extern const struct st_tracked_state st_update_viewport;
+extern const struct st_tracked_state st_update_scissor;
+extern const struct st_tracked_state st_update_blend;
+extern const struct st_tracked_state st_update_sampler;
+extern const struct st_tracked_state st_update_texture;
+extern const struct st_tracked_state st_update_fs_constants;
+extern const struct st_tracked_state st_update_vs_constants;
+extern const struct st_tracked_state st_update_pixel_transfer;
 
 
 uint st_compare_func_to_pipe(GLenum func);
diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c
index 0c22e03883..4ce7c41e90 100644
--- a/src/mesa/state_tracker/st_atom_sampler.c
+++ b/src/mesa/state_tracker/st_atom_sampler.c
@@ -32,12 +32,14 @@
   */
  
 
+#include "main/macros.h"
+
 #include "st_context.h"
 #include "st_atom.h"
 #include "st_program.h"
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
-#include "pipe/p_util.h"
+
 #include "cso_cache/cso_context.h"
 
 
diff --git a/src/mesa/state_tracker/st_atom_texture.c b/src/mesa/state_tracker/st_atom_texture.c
index f42b2f8d66..767654f3d0 100644
--- a/src/mesa/state_tracker/st_atom_texture.c
+++ b/src/mesa/state_tracker/st_atom_texture.c
@@ -90,6 +90,8 @@ update_textures(struct st_context *st)
             }
 
             st->state.num_textures = su + 1;
+
+            stObj->teximage_realloc = TRUE;
          }
 
          pt = st_get_stobj_texture(stObj);
diff --git a/src/mesa/state_tracker/st_cb_strings.c b/src/mesa/state_tracker/st_cb_strings.c
index 247519ab3d..066124f8f3 100644
--- a/src/mesa/state_tracker/st_cb_strings.c
+++ b/src/mesa/state_tracker/st_cb_strings.c
@@ -53,21 +53,21 @@ st_get_string(GLcontext * ctx, GLenum name)
       const char *vendor = screen->get_vendor( screen );
       const char *tungsten = "Tungsten Graphics, Inc.";
 
-      /* Tungsten developed the state_tracker module (and much of
-       * Mesa), but the driver itself may come from elsewhere.  The
-       * additional string allows "and XyzCorp" to reflect this.
+      /* Tungsten Graphics, Inc. developed the state_tracker module
+       * (and much of Mesa), but the driver itself may come from elsewhere.
+       * The additional string allows "and XyzCorp" to reflect this.
        */
       if (vendor && strcmp(vendor, tungsten) != 0)
-	 snprintf(st->vendor, sizeof(st->vendor),
+	 util_snprintf(st->vendor, sizeof(st->vendor),
                   "%s and %s", tungsten, vendor);
       else
-	 snprintf(st->vendor, sizeof(st->vendor), "%s", tungsten);
+	 util_snprintf(st->vendor, sizeof(st->vendor), "%s", tungsten);
 
       return (GLubyte *) st->vendor;
    }
 
    case GL_RENDERER:
-      snprintf(st->renderer, sizeof(st->renderer), "Gallium %s, %s on %s", 
+      util_snprintf(st->renderer, sizeof(st->renderer), "Gallium %s, %s on %s", 
                ST_VERSION_STRING,
 	       screen->get_name( screen ),
 	       screen->winsys->get_name( screen->winsys ));
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index 21c0141ac3..c877e4382d 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -557,15 +557,17 @@ st_TexImage(GLcontext * ctx,
     * waiting on any outstanding fences.
     */
    if (stObj->pt &&
-       /*stObj->pt->first_level == level &&*/
-       stObj->pt->last_level == level &&
-       stObj->pt->target != PIPE_TEXTURE_CUBE &&
-       !st_texture_match_image(stObj->pt, &stImage->base,
-                                  stImage->face, stImage->level)) {
+       (stObj->teximage_realloc ||
+        (/*stObj->pt->first_level == level &&*/
+         stObj->pt->last_level == level &&
+         stObj->pt->target != PIPE_TEXTURE_CUBE &&
+         !st_texture_match_image(stObj->pt, &stImage->base,
+                                 stImage->face, stImage->level)))) {
 
       DBG("release it\n");
       pipe_texture_release(&stObj->pt);
       assert(!stObj->pt);
+      stObj->teximage_realloc = FALSE;
    }
 
    if (!stObj->pt) {
@@ -1445,17 +1447,6 @@ st_finalize_texture(GLcontext *ctx,
    calculate_first_last_level(stObj);
    firstImage = st_texture_image(stObj->base.Image[0][stObj->base.BaseLevel]);
 
-#if 0
-   /* Fallback case:
-    */
-   if (firstImage->base.Border) {
-      if (stObj->pt) {
-         pipe_texture_release(&stObj->pt);
-      }
-      return GL_FALSE;
-   }
-#endif
-
    /* If both firstImage and stObj point to a texture which can contain
     * all active images, favour firstImage.  Note that because of the
     * completeness requirement, we know that the image dimensions
@@ -1479,8 +1470,8 @@ st_finalize_texture(GLcontext *ctx,
       cpp = firstImage->base.TexFormat->TexelBytes;
    }
 
-   /* Check texture can hold all active levels.  Check texture matches
-    * target, imageFormat, etc.
+   /* If we already have a gallium texture, check that it matches the texture
+    * object's format, target, size, num_levels, etc.
     */
    if (stObj->pt &&
        (stObj->pt->target != gl_target_to_pipe(stObj->base.Target) ||
@@ -1495,8 +1486,7 @@ st_finalize_texture(GLcontext *ctx,
       pipe_texture_release(&stObj->pt);
    }
 
-
-   /* May need to create a new texture:
+   /* May need to create a new gallium texture:
     */
    if (!stObj->pt) {
       stObj->pt = st_texture_create(ctx->st,
diff --git a/src/mesa/state_tracker/st_texture.h b/src/mesa/state_tracker/st_texture.h
index 6a9f08ec6b..55d1a367b5 100644
--- a/src/mesa/state_tracker/st_texture.h
+++ b/src/mesa/state_tracker/st_texture.h
@@ -68,9 +68,7 @@ struct st_texture_object
     */
    struct pipe_texture *pt;
 
-   GLboolean imageOverride;
-   GLint depthOverride;
-   GLuint pitchOverride;
+   GLboolean teximage_realloc;
 };