Merge remote-tracking branch 'origin/released'

Conflicts:
	mesalib/src/mapi/glapi/glapi_priv.h
	mesalib/src/mesa/drivers/dri/swrast/swrast.c
	xorg-server/dix/dispatch.c
	xorg-server/os/utils.c
	xorg-server/record/record.c

15 files changed, 722 insertions, 55 deletions

diff --git a/mesalib/src/mesa/state_tracker/st_atom_array.c b/mesalib/src/mesa/state_tracker/st_atom_array.c
index d4fb8b862..56b8019a3 100644
--- a/mesalib/src/mesa/state_tracker/st_atom_array.c
+++ b/mesalib/src/mesa/state_tracker/st_atom_array.c
@@ -44,7 +44,6 @@
 
 #include "cso_cache/cso_context.h"
 #include "util/u_math.h"
-
 #include "main/bufferobj.h"
 #include "main/glformats.h"
 
@@ -311,6 +310,18 @@ st_pipe_vertex_format(GLenum type, GLuint size, GLenum format,
    return PIPE_FORMAT_NONE; /* silence compiler warning */
 }
 
+static const struct gl_client_array *
+get_client_array(const struct st_vertex_program *vp,
+                 const struct gl_client_array **arrays,
+                 int attr)
+{
+   const GLuint mesaAttr = vp->index_to_input[attr];
+   /* st_program uses 0xffffffff to denote a double placeholder attribute */
+   if (mesaAttr == ST_DOUBLE_ATTRIB_PLACEHOLDER)
+      return NULL;
+   return arrays[mesaAttr];
+}
+
 /**
  * Examine the active arrays to determine if we have interleaved
  * vertex arrays all living in one VBO, or all living in user space.
@@ -327,11 +338,16 @@ is_interleaved_arrays(const struct st_vertex_program *vp,
    GLboolean userSpaceBuffer = GL_FALSE;
 
    for (attr = 0; attr < vpv->num_inputs; attr++) {
-      const GLuint mesaAttr = vp->index_to_input[attr];
-      const struct gl_client_array *array = arrays[mesaAttr];
-      const struct gl_buffer_object *bufObj = array->BufferObj;
-      const GLsizei stride = array->StrideB; /* in bytes */
+      const struct gl_client_array *array;
+      const struct gl_buffer_object *bufObj;
+      GLsizei stride;
+
+      array = get_client_array(vp, arrays, attr);
+      if (!array)
+	 continue;
 
+      stride = array->StrideB; /* in bytes */
+      bufObj = array->BufferObj;
       if (attr == 0) {
          /* save info about the first array */
          firstStride = stride;
@@ -358,6 +374,55 @@ is_interleaved_arrays(const struct st_vertex_program *vp,
    return GL_TRUE;
 }
 
+static void init_velement(struct pipe_vertex_element *velement,
+                          int src_offset, int format,
+                          int instance_divisor, int vbo_index)
+{
+   velement->src_offset = src_offset;
+   velement->src_format = format;
+   velement->instance_divisor = instance_divisor;
+   velement->vertex_buffer_index = vbo_index;
+   assert(velement->src_format);
+}
+
+static void init_velement_lowered(struct st_context *st,
+                                  struct pipe_vertex_element *velements,
+                                  int src_offset, int format,
+                                  int instance_divisor, int vbo_index,
+                                  int nr_components, GLboolean doubles,
+                                  GLuint *attr_idx)
+{
+   int idx = *attr_idx;
+   if (doubles) {
+      int lower_format;
+
+      if (nr_components == 1)
+         lower_format = PIPE_FORMAT_R32G32_UINT;
+      else if (nr_components >= 2)
+         lower_format = PIPE_FORMAT_R32G32B32A32_UINT;
+
+      init_velement(&velements[idx], src_offset,
+                    lower_format, instance_divisor, vbo_index);
+      idx++;
+
+      if (nr_components > 2) {
+         if (nr_components == 3)
+            lower_format = PIPE_FORMAT_R32G32_UINT;
+         else if (nr_components >= 4)
+            lower_format = PIPE_FORMAT_R32G32B32A32_UINT;
+
+         init_velement(&velements[idx], src_offset + 4 * sizeof(float),
+                       lower_format, instance_divisor, vbo_index);
+         idx++;
+      }
+   } else {
+      init_velement(&velements[idx], src_offset,
+                    format, instance_divisor, vbo_index);
+      idx++;
+   }
+   *attr_idx = idx;
+}
+
 /**
  * Set up for drawing interleaved arrays that all live in one VBO
  * or all live in user space.
@@ -365,13 +430,15 @@ is_interleaved_arrays(const struct st_vertex_program *vp,
  * \param velements  returns vertex element info
  */
 static boolean
-setup_interleaved_attribs(const struct st_vertex_program *vp,
+setup_interleaved_attribs(struct st_context *st,
+                          const struct st_vertex_program *vp,
                           const struct st_vp_variant *vpv,
                           const struct gl_client_array **arrays,
                           struct pipe_vertex_buffer *vbuffer,
-                          struct pipe_vertex_element velements[])
+                          struct pipe_vertex_element velements[],
+                          unsigned *num_velements)
 {
-   GLuint attr;
+   GLuint attr, attr_idx;
    const GLubyte *low_addr = NULL;
    GLboolean usingVBO;      /* all arrays in a VBO? */
    struct gl_buffer_object *bufobj;
@@ -381,8 +448,10 @@ setup_interleaved_attribs(const struct st_vertex_program *vp,
     * Init bufobj and stride.
     */
    if (vpv->num_inputs) {
-      const GLuint mesaAttr0 = vp->index_to_input[0];
-      const struct gl_client_array *array = arrays[mesaAttr0];
+      const struct gl_client_array *array;
+
+      array = get_client_array(vp, arrays, 0);
+      assert(array);
 
       /* Since we're doing interleaved arrays, we know there'll be at most
        * one buffer object and the stride will be the same for all arrays.
@@ -394,7 +463,11 @@ setup_interleaved_attribs(const struct st_vertex_program *vp,
       low_addr = arrays[vp->index_to_input[0]]->Ptr;
 
       for (attr = 1; attr < vpv->num_inputs; attr++) {
-         const GLubyte *start = arrays[vp->index_to_input[attr]]->Ptr;
+         const GLubyte *start;
+         array = get_client_array(vp, arrays, attr);
+         if (!array)
+            continue;
+         start = array->Ptr;
          low_addr = MIN2(low_addr, start);
       }
    }
@@ -408,25 +481,33 @@ setup_interleaved_attribs(const struct st_vertex_program *vp,
    /* are the arrays in user space? */
    usingVBO = _mesa_is_bufferobj(bufobj);
 
+   attr_idx = 0;
    for (attr = 0; attr < vpv->num_inputs; attr++) {
-      const GLuint mesaAttr = vp->index_to_input[attr];
-      const struct gl_client_array *array = arrays[mesaAttr];
-      unsigned src_offset = (unsigned) (array->Ptr - low_addr);
+      const struct gl_client_array *array;
+      unsigned src_offset;
+      unsigned src_format;
+
+      array = get_client_array(vp, arrays, attr);
+      if (!array)
+         continue;
 
+      src_offset = (unsigned) (array->Ptr - low_addr);
       assert(array->_ElementSize ==
              _mesa_bytes_per_vertex_attrib(array->Size, array->Type));
 
-      velements[attr].src_offset = src_offset;
-      velements[attr].instance_divisor = array->InstanceDivisor;
-      velements[attr].vertex_buffer_index = 0;
-      velements[attr].src_format = st_pipe_vertex_format(array->Type,
-                                                         array->Size,
-                                                         array->Format,
-                                                         array->Normalized,
-                                                         array->Integer);
-      assert(velements[attr].src_format);
+      src_format = st_pipe_vertex_format(array->Type,
+                                         array->Size,
+                                         array->Format,
+                                         array->Normalized,
+                                         array->Integer);
+
+      init_velement_lowered(st, velements, src_offset, src_format,
+                            array->InstanceDivisor, 0,
+                            array->Size, array->Doubles, &attr_idx);
    }
 
+   *num_velements = attr_idx;
+
    /*
     * Return the vbuffer info and setup user-space attrib info, if needed.
     */
@@ -472,17 +553,25 @@ setup_non_interleaved_attribs(struct st_context *st,
                               const struct st_vp_variant *vpv,
                               const struct gl_client_array **arrays,
                               struct pipe_vertex_buffer vbuffer[],
-                              struct pipe_vertex_element velements[])
+                              struct pipe_vertex_element velements[],
+                              unsigned *num_velements)
 {
    struct gl_context *ctx = st->ctx;
-   GLuint attr;
+   GLuint attr, attr_idx = 0;
 
    for (attr = 0; attr < vpv->num_inputs; attr++) {
       const GLuint mesaAttr = vp->index_to_input[attr];
-      const struct gl_client_array *array = arrays[mesaAttr];
-      struct gl_buffer_object *bufobj = array->BufferObj;
-      GLsizei stride = array->StrideB;
+      const struct gl_client_array *array;
+      struct gl_buffer_object *bufobj;
+      GLsizei stride;
+      unsigned src_format;
 
+      array = get_client_array(vp, arrays, attr);
+      if (!array)
+         continue;
+
+      stride = array->StrideB;
+      bufobj = array->BufferObj;
       assert(array->_ElementSize ==
              _mesa_bytes_per_vertex_attrib(array->Size, array->Type));
 
@@ -524,16 +613,19 @@ setup_non_interleaved_attribs(struct st_context *st,
       /* common-case setup */
       vbuffer[attr].stride = stride; /* in bytes */
 
-      velements[attr].src_offset = 0;
-      velements[attr].instance_divisor = array->InstanceDivisor;
-      velements[attr].vertex_buffer_index = attr;
-      velements[attr].src_format = st_pipe_vertex_format(array->Type,
-                                                         array->Size,
-                                                         array->Format,
-                                                         array->Normalized,
-                                                         array->Integer);
-      assert(velements[attr].src_format);
+      src_format = st_pipe_vertex_format(array->Type,
+                                         array->Size,
+                                         array->Format,
+                                         array->Normalized,
+                                         array->Integer);
+
+      init_velement_lowered(st, velements, 0, src_format,
+                            array->InstanceDivisor, attr,
+                            array->Size, array->Doubles, &attr_idx);
+
    }
+
+   *num_velements = attr_idx;
    return TRUE;
 }
 
@@ -563,25 +655,23 @@ static void update_array(struct st_context *st)
     * Setup the vbuffer[] and velements[] arrays.
     */
    if (is_interleaved_arrays(vp, vpv, arrays)) {
-      if (!setup_interleaved_attribs(vp, vpv, arrays, vbuffer, velements)) {
+      if (!setup_interleaved_attribs(st, vp, vpv, arrays, vbuffer, velements, &num_velements)) {
          st->vertex_array_out_of_memory = TRUE;
          return;
       }
 
       num_vbuffers = 1;
-      num_velements = vpv->num_inputs;
       if (num_velements == 0)
          num_vbuffers = 0;
    }
    else {
       if (!setup_non_interleaved_attribs(st, vp, vpv, arrays, vbuffer,
-                                         velements)) {
+                                         velements, &num_velements)) {
          st->vertex_array_out_of_memory = TRUE;
          return;
       }
 
       num_vbuffers = vpv->num_inputs;
-      num_velements = vpv->num_inputs;
    }
 
    cso_set_vertex_buffers(st->cso_context, 0, num_vbuffers, vbuffer);
diff --git a/mesalib/src/mesa/state_tracker/st_atom_framebuffer.c b/mesalib/src/mesa/state_tracker/st_atom_framebuffer.c
index b195c55b3..ae883a253 100644
--- a/mesalib/src/mesa/state_tracker/st_atom_framebuffer.c
+++ b/mesalib/src/mesa/state_tracker/st_atom_framebuffer.c
@@ -134,7 +134,10 @@ update_framebuffer_state( struct st_context *st )
    else {
       strb = st_renderbuffer(fb->Attachment[BUFFER_STENCIL].Renderbuffer);
       if (strb) {
-         assert(strb->surface);
+         if (strb->is_rtt) {
+            /* rendering to a GL texture, may have to update surface */
+            st_update_renderbuffer_surface(st, strb);
+         }
          pipe_surface_reference(&framebuffer->zsbuf, strb->surface);
          update_framebuffer_size(framebuffer, strb->surface);
       }
diff --git a/mesalib/src/mesa/state_tracker/st_cb_fbo.c b/mesalib/src/mesa/state_tracker/st_cb_fbo.c
index 296ea1e0d..0399eef72 100644
--- a/mesalib/src/mesa/state_tracker/st_cb_fbo.c
+++ b/mesalib/src/mesa/state_tracker/st_cb_fbo.c
@@ -842,7 +842,7 @@ void st_init_fbo_functions(struct dd_function_table *functions)
    functions->NewFramebuffer = st_new_framebuffer;
    functions->NewRenderbuffer = st_new_renderbuffer;
    functions->BindFramebuffer = st_bind_framebuffer;
-   functions->FramebufferRenderbuffer = _mesa_framebuffer_renderbuffer;
+   functions->FramebufferRenderbuffer = _mesa_FramebufferRenderbuffer_sw;
    functions->RenderTexture = st_render_texture;
    functions->FinishRenderTexture = st_finish_render_texture;
    functions->ValidateFramebuffer = st_validate_framebuffer;
diff --git a/mesalib/src/mesa/state_tracker/st_cb_flush.c b/mesalib/src/mesa/state_tracker/st_cb_flush.c
index ca51eeee3..82affd2de 100644
--- a/mesalib/src/mesa/state_tracker/st_cb_flush.c
+++ b/mesalib/src/mesa/state_tracker/st_cb_flush.c
@@ -141,11 +141,44 @@ static void st_glFinish(struct gl_context *ctx)
 }
 
 
-void st_init_flush_functions(struct dd_function_table *functions)
+/**
+ * Query information about GPU resets observed by this context
+ *
+ * Called via \c dd_function_table::GetGraphicsResetStatus.
+ */
+static GLenum
+st_get_graphics_reset_status(struct gl_context *ctx)
+{
+   struct st_context *st = st_context(ctx);
+   enum pipe_reset_status status;
+
+   status = st->pipe->get_device_reset_status(st->pipe);
+
+   switch (status) {
+   case PIPE_NO_RESET:
+      return GL_NO_ERROR;
+   case PIPE_GUILTY_CONTEXT_RESET:
+      return GL_GUILTY_CONTEXT_RESET_ARB;
+   case PIPE_INNOCENT_CONTEXT_RESET:
+      return GL_INNOCENT_CONTEXT_RESET_ARB;
+   case PIPE_UNKNOWN_CONTEXT_RESET:
+      return GL_UNKNOWN_CONTEXT_RESET_ARB;
+   default:
+      assert(0);
+      return GL_NO_ERROR;
+   }
+}
+
+
+void st_init_flush_functions(struct pipe_screen *screen,
+                             struct dd_function_table *functions)
 {
    functions->Flush = st_glFlush;
    functions->Finish = st_glFinish;
 
+   if (screen->get_param(screen, PIPE_CAP_DEVICE_RESET_STATUS_QUERY))
+      functions->GetGraphicsResetStatus = st_get_graphics_reset_status;
+
    /* Windows opengl32.dll calls glFinish prior to every swapbuffers.
     * This is unnecessary and degrades performance.  Luckily we have some
     * scope to work around this, as the externally-visible behaviour of
diff --git a/mesalib/src/mesa/state_tracker/st_cb_flush.h b/mesalib/src/mesa/state_tracker/st_cb_flush.h
index 84ffc63ae..f92dcd56b 100644
--- a/mesalib/src/mesa/state_tracker/st_cb_flush.h
+++ b/mesalib/src/mesa/state_tracker/st_cb_flush.h
@@ -37,7 +37,8 @@ struct pipe_fence_handle;
 struct st_context;
 
 extern void
-st_init_flush_functions(struct dd_function_table *functions);
+st_init_flush_functions(struct pipe_screen *screen,
+                        struct dd_function_table *functions);
 
 extern void
 st_flush(struct st_context *st,
diff --git a/mesalib/src/mesa/state_tracker/st_cb_perfmon.c b/mesalib/src/mesa/state_tracker/st_cb_perfmon.c
new file mode 100644
index 000000000..1bb5be397
--- /dev/null
+++ b/mesalib/src/mesa/state_tracker/st_cb_perfmon.c
@@ -0,0 +1,425 @@
+/*
+ * Copyright (C) 2013 Christoph Bumiller
+ * Copyright (C) 2015 Samuel Pitoiset
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+/**
+ * Performance monitoring counters interface to gallium.
+ */
+
+#include "st_debug.h"
+#include "st_context.h"
+#include "st_cb_bitmap.h"
+#include "st_cb_perfmon.h"
+
+#include "util/bitset.h"
+
+#include "pipe/p_context.h"
+#include "pipe/p_screen.h"
+#include "util/u_memory.h"
+
+/**
+ * Return a PIPE_QUERY_x type >= PIPE_QUERY_DRIVER_SPECIFIC, or -1 if
+ * the driver-specific query doesn't exist.
+ */
+static int
+find_query_type(struct pipe_screen *screen, const char *name)
+{
+   int num_queries;
+   int type = -1;
+   int i;
+
+   num_queries = screen->get_driver_query_info(screen, 0, NULL);
+   if (!num_queries)
+      return type;
+
+   for (i = 0; i < num_queries; i++) {
+      struct pipe_driver_query_info info;
+
+      if (!screen->get_driver_query_info(screen, i, &info))
+         continue;
+
+      if (!strncmp(info.name, name, strlen(name))) {
+         type = info.query_type;
+         break;
+      }
+   }
+   return type;
+}
+
+/**
+ * Return TRUE if the underlying driver expose GPU counters.
+ */
+static bool
+has_gpu_counters(struct pipe_screen *screen)
+{
+   int num_groups, gid;
+
+   num_groups = screen->get_driver_query_group_info(screen, 0, NULL);
+   for (gid = 0; gid < num_groups; gid++) {
+      struct pipe_driver_query_group_info group_info;
+
+      if (!screen->get_driver_query_group_info(screen, gid, &group_info))
+         continue;
+
+      if (group_info.type == PIPE_DRIVER_QUERY_GROUP_TYPE_GPU)
+         return true;
+   }
+   return false;
+}
+
+static bool
+init_perf_monitor(struct gl_context *ctx, struct gl_perf_monitor_object *m)
+{
+   struct st_perf_monitor_object *stm = st_perf_monitor_object(m);
+   struct pipe_screen *screen = st_context(ctx)->pipe->screen;
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   int gid, cid;
+
+   st_flush_bitmap_cache(st_context(ctx));
+
+   /* Create a query for each active counter. */
+   for (gid = 0; gid < ctx->PerfMonitor.NumGroups; gid++) {
+      const struct gl_perf_monitor_group *g = &ctx->PerfMonitor.Groups[gid];
+
+      if (m->ActiveGroups[gid] > g->MaxActiveCounters) {
+         /* Maximum number of counters reached. Cannot start the session. */
+         if (ST_DEBUG & DEBUG_MESA) {
+            debug_printf("Maximum number of counters reached. "
+                         "Cannot start the session!\n");
+         }
+         return false;
+      }
+
+      for (cid = 0; cid < g->NumCounters; cid++) {
+         const struct gl_perf_monitor_counter *c = &g->Counters[cid];
+         struct st_perf_counter_object *cntr;
+         int query_type;
+
+         if (!BITSET_TEST(m->ActiveCounters[gid], cid))
+            continue;
+
+         query_type = find_query_type(screen, c->Name);
+         assert(query_type != -1);
+
+         cntr = CALLOC_STRUCT(st_perf_counter_object);
+         if (!cntr)
+            return false;
+
+         cntr->query    = pipe->create_query(pipe, query_type, 0);
+         cntr->id       = cid;
+         cntr->group_id = gid;
+
+         list_addtail(&cntr->list, &stm->active_counters);
+      }
+   }
+   return true;
+}
+
+static void
+reset_perf_monitor(struct st_perf_monitor_object *stm,
+                   struct pipe_context *pipe)
+{
+   struct st_perf_counter_object *cntr, *tmp;
+
+   LIST_FOR_EACH_ENTRY_SAFE(cntr, tmp, &stm->active_counters, list) {
+      if (cntr->query)
+         pipe->destroy_query(pipe, cntr->query);
+      list_del(&cntr->list);
+      free(cntr);
+   }
+}
+
+static struct gl_perf_monitor_object *
+st_NewPerfMonitor(struct gl_context *ctx)
+{
+   struct st_perf_monitor_object *stq = ST_CALLOC_STRUCT(st_perf_monitor_object);
+   if (stq) {
+      list_inithead(&stq->active_counters);
+      return &stq->base;
+   }
+   return NULL;
+}
+
+static void
+st_DeletePerfMonitor(struct gl_context *ctx, struct gl_perf_monitor_object *m)
+{
+   struct st_perf_monitor_object *stm = st_perf_monitor_object(m);
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+
+   reset_perf_monitor(stm, pipe);
+   FREE(stm);
+}
+
+static GLboolean
+st_BeginPerfMonitor(struct gl_context *ctx, struct gl_perf_monitor_object *m)
+{
+   struct st_perf_monitor_object *stm = st_perf_monitor_object(m);
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_perf_counter_object *cntr;
+
+   if (LIST_IS_EMPTY(&stm->active_counters)) {
+      /* Create a query for each active counter before starting
+       * a new monitoring session. */
+      if (!init_perf_monitor(ctx, m))
+         goto fail;
+   }
+
+   /* Start the query for each active counter. */
+   LIST_FOR_EACH_ENTRY(cntr, &stm->active_counters, list) {
+      if (!pipe->begin_query(pipe, cntr->query))
+          goto fail;
+   }
+   return true;
+
+fail:
+   /* Failed to start the monitoring session. */
+   reset_perf_monitor(stm, pipe);
+   return false;
+}
+
+static void
+st_EndPerfMonitor(struct gl_context *ctx, struct gl_perf_monitor_object *m)
+{
+   struct st_perf_monitor_object *stm = st_perf_monitor_object(m);
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_perf_counter_object *cntr;
+
+   /* Stop the query for each active counter. */
+   LIST_FOR_EACH_ENTRY(cntr, &stm->active_counters, list)
+      pipe->end_query(pipe, cntr->query);
+}
+
+static void
+st_ResetPerfMonitor(struct gl_context *ctx, struct gl_perf_monitor_object *m)
+{
+   struct st_perf_monitor_object *stm = st_perf_monitor_object(m);
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+
+   if (!m->Ended)
+      st_EndPerfMonitor(ctx, m);
+
+   reset_perf_monitor(stm, pipe);
+
+   if (m->Active)
+      st_BeginPerfMonitor(ctx, m);
+}
+
+static GLboolean
+st_IsPerfMonitorResultAvailable(struct gl_context *ctx,
+                                struct gl_perf_monitor_object *m)
+{
+   struct st_perf_monitor_object *stm = st_perf_monitor_object(m);
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_perf_counter_object *cntr;
+
+   if (LIST_IS_EMPTY(&stm->active_counters))
+      return false;
+
+   /* The result of a monitoring session is only available if the query of
+    * each active counter is idle. */
+   LIST_FOR_EACH_ENTRY(cntr, &stm->active_counters, list) {
+      union pipe_query_result result;
+      if (!pipe->get_query_result(pipe, cntr->query, FALSE, &result)) {
+         /* The query is busy. */
+         return false;
+      }
+   }
+   return true;
+}
+
+static void
+st_GetPerfMonitorResult(struct gl_context *ctx,
+                        struct gl_perf_monitor_object *m,
+                        GLsizei dataSize,
+                        GLuint *data,
+                        GLint *bytesWritten)
+{
+   struct st_perf_monitor_object *stm = st_perf_monitor_object(m);
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_perf_counter_object *cntr;
+
+   /* Copy data to the supplied array (data).
+    *
+    * The output data format is: <group ID, counter ID, value> for each
+    * active counter. The API allows counters to appear in any order.
+    */
+   GLsizei offset = 0;
+
+   /* Read query results for each active counter. */
+   LIST_FOR_EACH_ENTRY(cntr, &stm->active_counters, list) {
+      union pipe_query_result result = { 0 };
+      int gid, cid;
+      GLenum type;
+
+      cid  = cntr->id;
+      gid  = cntr->group_id;
+      type = ctx->PerfMonitor.Groups[gid].Counters[cid].Type;
+
+      if (!pipe->get_query_result(pipe, cntr->query, TRUE, &result))
+         continue;
+
+      data[offset++] = gid;
+      data[offset++] = cid;
+      switch (type) {
+      case GL_UNSIGNED_INT64_AMD:
+         *(uint64_t *)&data[offset] = result.u64;
+         offset += sizeof(uint64_t) / sizeof(GLuint);
+         break;
+      case GL_UNSIGNED_INT:
+         *(uint32_t *)&data[offset] = result.u32;
+         offset += sizeof(uint32_t) / sizeof(GLuint);
+         break;
+      case GL_FLOAT:
+      case GL_PERCENTAGE_AMD:
+         *(GLfloat *)&data[offset] = result.f;
+         offset += sizeof(GLfloat) / sizeof(GLuint);
+         break;
+      }
+   }
+
+   if (bytesWritten)
+      *bytesWritten = offset * sizeof(GLuint);
+}
+
+
+bool
+st_init_perfmon(struct st_context *st)
+{
+   struct gl_perf_monitor_state *perfmon = &st->ctx->PerfMonitor;
+   struct pipe_screen *screen = st->pipe->screen;
+   struct gl_perf_monitor_group *groups = NULL;
+   int num_counters, num_groups;
+   int gid, cid;
+
+   if (!screen->get_driver_query_info || !screen->get_driver_query_group_info)
+      return false;
+
+   if (!has_gpu_counters(screen)) {
+      /* According to the spec, GL_AMD_performance_monitor must only
+       * expose GPU counters. */
+      return false;
+   }
+
+   /* Get the number of available queries. */
+   num_counters = screen->get_driver_query_info(screen, 0, NULL);
+   if (!num_counters)
+      return false;
+
+   /* Get the number of available groups. */
+   num_groups = screen->get_driver_query_group_info(screen, 0, NULL);
+   if (num_groups)
+      groups = CALLOC(num_groups, sizeof(*groups));
+   if (!groups)
+      return false;
+
+   for (gid = 0; gid < num_groups; gid++) {
+      struct gl_perf_monitor_group *g = &groups[perfmon->NumGroups];
+      struct pipe_driver_query_group_info group_info;
+      struct gl_perf_monitor_counter *counters = NULL;
+
+      if (!screen->get_driver_query_group_info(screen, gid, &group_info))
+         continue;
+
+      if (group_info.type != PIPE_DRIVER_QUERY_GROUP_TYPE_GPU)
+         continue;
+
+      g->Name = group_info.name;
+      g->MaxActiveCounters = group_info.max_active_queries;
+      g->NumCounters = 0;
+      g->Counters = NULL;
+
+      if (group_info.num_queries)
+         counters = CALLOC(group_info.num_queries, sizeof(*counters));
+      if (!counters)
+         goto fail;
+
+      for (cid = 0; cid < num_counters; cid++) {
+         struct gl_perf_monitor_counter *c = &counters[g->NumCounters];
+         struct pipe_driver_query_info info;
+
+         if (!screen->get_driver_query_info(screen, cid, &info))
+            continue;
+         if (info.group_id != gid)
+            continue;
+
+         c->Name = info.name;
+         switch (info.type) {
+            case PIPE_DRIVER_QUERY_TYPE_UINT64:
+               c->Minimum.u64 = 0;
+               c->Maximum.u64 = info.max_value.u64 ? info.max_value.u64 : -1;
+               c->Type = GL_UNSIGNED_INT64_AMD;
+               break;
+            case PIPE_DRIVER_QUERY_TYPE_UINT:
+               c->Minimum.u32 = 0;
+               c->Maximum.u32 = info.max_value.u32 ? info.max_value.u32 : -1;
+               c->Type = GL_UNSIGNED_INT;
+               break;
+            case PIPE_DRIVER_QUERY_TYPE_FLOAT:
+               c->Minimum.f = 0.0;
+               c->Maximum.f = info.max_value.f ? info.max_value.f : -1;
+               c->Type = GL_FLOAT;
+               break;
+            case PIPE_DRIVER_QUERY_TYPE_PERCENTAGE:
+               c->Minimum.f = 0.0f;
+               c->Maximum.f = 100.0f;
+               c->Type = GL_PERCENTAGE_AMD;
+               break;
+            default:
+               unreachable("Invalid driver query type!");
+         }
+         g->NumCounters++;
+      }
+      g->Counters = counters;
+      perfmon->NumGroups++;
+   }
+   perfmon->Groups = groups;
+
+   return true;
+
+fail:
+   for (gid = 0; gid < num_groups; gid++)
+      FREE((void *)groups[gid].Counters);
+   FREE(groups);
+   return false;
+}
+
+void
+st_destroy_perfmon(struct st_context *st)
+{
+   struct gl_perf_monitor_state *perfmon = &st->ctx->PerfMonitor;
+   int gid;
+
+   for (gid = 0; gid < perfmon->NumGroups; gid++)
+      FREE((void *)perfmon->Groups[gid].Counters);
+   FREE((void *)perfmon->Groups);
+}
+
+void st_init_perfmon_functions(struct dd_function_table *functions)
+{
+   functions->NewPerfMonitor = st_NewPerfMonitor;
+   functions->DeletePerfMonitor = st_DeletePerfMonitor;
+   functions->BeginPerfMonitor = st_BeginPerfMonitor;
+   functions->EndPerfMonitor = st_EndPerfMonitor;
+   functions->ResetPerfMonitor = st_ResetPerfMonitor;
+   functions->IsPerfMonitorResultAvailable = st_IsPerfMonitorResultAvailable;
+   functions->GetPerfMonitorResult = st_GetPerfMonitorResult;
+}
diff --git a/mesalib/src/mesa/state_tracker/st_cb_perfmon.h b/mesalib/src/mesa/state_tracker/st_cb_perfmon.h
new file mode 100644
index 000000000..13d3627de
--- /dev/null
+++ b/mesalib/src/mesa/state_tracker/st_cb_perfmon.h
@@ -0,0 +1,64 @@
+/*
+ * Copyright (C) 2013 Christoph Bumiller
+ * Copyright (C) 2015 Samuel Pitoiset
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the "Software"),
+ * to deal in the Software without restriction, including without limitation
+ * the rights to use, copy, modify, merge, publish, distribute, sublicense,
+ * and/or sell copies of the Software, and to permit persons to whom the
+ * Software is furnished to do so, subject to the following conditions:
+ *
+ * The above copyright notice and this permission notice shall be included in
+ * all copies or substantial portions of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+ * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+ * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT.  IN NO EVENT SHALL
+ * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR
+ * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE,
+ * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR
+ * OTHER DEALINGS IN THE SOFTWARE.
+ */
+
+#ifndef ST_CB_PERFMON_H
+#define ST_CB_PERFMON_H
+
+#include "util/list.h"
+
+/**
+ * Subclass of gl_perf_monitor_object
+ */
+struct st_perf_monitor_object
+{
+   struct gl_perf_monitor_object base;
+   struct list_head active_counters;
+};
+
+struct st_perf_counter_object
+{
+   struct list_head list;
+   struct pipe_query *query;
+   int id;
+   int group_id;
+};
+
+/**
+ * Cast wrapper
+ */
+static INLINE struct st_perf_monitor_object *
+st_perf_monitor_object(struct gl_perf_monitor_object *q)
+{
+   return (struct st_perf_monitor_object *)q;
+}
+
+bool
+st_init_perfmon(struct st_context *st);
+
+void
+st_destroy_perfmon(struct st_context *st);
+
+extern void
+st_init_perfmon_functions(struct dd_function_table *functions);
+
+#endif
diff --git a/mesalib/src/mesa/state_tracker/st_context.c b/mesalib/src/mesa/state_tracker/st_context.c
index 5fe132ac2..69e0f929d 100644
--- a/mesalib/src/mesa/state_tracker/st_context.c
+++ b/mesalib/src/mesa/state_tracker/st_context.c
@@ -51,6 +51,7 @@
 #include "st_cb_fbo.h"
 #include "st_cb_feedback.h"
 #include "st_cb_msaa.h"
+#include "st_cb_perfmon.h"
 #include "st_cb_program.h"
 #include "st_cb_queryobj.h"
 #include "st_cb_readpixels.h"
@@ -116,6 +117,7 @@ st_destroy_context_priv(struct st_context *st)
    st_destroy_bitmap(st);
    st_destroy_drawpix(st);
    st_destroy_drawtex(st);
+   st_destroy_perfmon(st);
 
    for (shader = 0; shader < ARRAY_SIZE(st->state.sampler_views); shader++) {
       for (i = 0; i < ARRAY_SIZE(st->state.sampler_views[0]); i++) {
@@ -250,6 +252,12 @@ st_create_context_priv( struct gl_context *ctx, struct pipe_context *pipe,
    st_init_extensions(st->pipe->screen, &ctx->Const,
                       &ctx->Extensions, &st->options, ctx->Mesa_DXTn);
 
+   if (st_init_perfmon(st)) {
+      /* GL_AMD_performance_monitor is only enabled when the underlying
+       * driver expose GPU hardware performance counters. */
+      ctx->Extensions.AMD_performance_monitor = GL_TRUE;
+   }
+
    /* Enable shader-based fallbacks for ARB_color_buffer_float if needed. */
    if (screen->get_param(screen, PIPE_CAP_VERTEX_COLOR_UNCLAMPED)) {
       if (!screen->get_param(screen, PIPE_CAP_VERTEX_COLOR_CLAMPED)) {
@@ -313,7 +321,7 @@ struct st_context *st_create_context(gl_api api, struct pipe_context *pipe,
    struct st_context *st;
 
    memset(&funcs, 0, sizeof(funcs));
-   st_init_driver_functions(&funcs);
+   st_init_driver_functions(pipe->screen, &funcs);
 
    ctx = _mesa_create_context(api, visual, shareCtx, &funcs);
    if (!ctx) {
@@ -393,7 +401,8 @@ void st_destroy_context( struct st_context *st )
 }
 
 
-void st_init_driver_functions(struct dd_function_table *functions)
+void st_init_driver_functions(struct pipe_screen *screen,
+                              struct dd_function_table *functions)
 {
    _mesa_init_shader_object_functions(functions);
    _mesa_init_sampler_object_functions(functions);
@@ -414,13 +423,14 @@ void st_init_driver_functions(struct dd_function_table *functions)
    st_init_fbo_functions(functions);
    st_init_feedback_functions(functions);
    st_init_msaa_functions(functions);
+   st_init_perfmon_functions(functions);
    st_init_program_functions(functions);
    st_init_query_functions(functions);
    st_init_cond_render_functions(functions);
    st_init_readpixels_functions(functions);
    st_init_texture_functions(functions);
    st_init_texture_barrier_functions(functions);
-   st_init_flush_functions(functions);
+   st_init_flush_functions(screen, functions);
    st_init_string_functions(functions);
    st_init_viewport_functions(functions);
 
diff --git a/mesalib/src/mesa/state_tracker/st_context.h b/mesalib/src/mesa/state_tracker/st_context.h
index 8a9504bb7..dac5a4b90 100644
--- a/mesalib/src/mesa/state_tracker/st_context.h
+++ b/mesalib/src/mesa/state_tracker/st_context.h
@@ -237,7 +237,8 @@ struct st_framebuffer
 };
 
 
-extern void st_init_driver_functions(struct dd_function_table *functions);
+extern void st_init_driver_functions(struct pipe_screen *screen,
+                                     struct dd_function_table *functions);
 
 void st_invalidate_state(struct gl_context * ctx, GLuint new_state);
 
diff --git a/mesalib/src/mesa/state_tracker/st_extensions.c b/mesalib/src/mesa/state_tracker/st_extensions.c
index 82e4a3009..23a45883d 100644
--- a/mesalib/src/mesa/state_tracker/st_extensions.c
+++ b/mesalib/src/mesa/state_tracker/st_extensions.c
@@ -650,6 +650,12 @@ void st_init_extensions(struct pipe_screen *screen,
                           ARRAY_SIZE(vertex_mapping), PIPE_BUFFER,
                           PIPE_BIND_VERTEX_BUFFER);
 
+   /* ARB_direct_state_access requires OpenGL 2.0. Assume that all drivers
+    * that support NPOT textures are able to support GL 2.0.
+    */
+   if (extensions->ARB_texture_non_power_of_two)
+      extensions->ARB_direct_state_access = GL_TRUE;
+
    if (extensions->ARB_stencil_texturing)
       extensions->ARB_texture_stencil8 = GL_TRUE;
 
@@ -909,6 +915,8 @@ void st_init_extensions(struct pipe_screen *screen,
    if (screen->get_shader_param(screen, PIPE_SHADER_VERTEX,
                                 PIPE_SHADER_CAP_DOUBLES) &&
        screen->get_shader_param(screen, PIPE_SHADER_FRAGMENT,
-                                PIPE_SHADER_CAP_DOUBLES))
+                                PIPE_SHADER_CAP_DOUBLES)) {
       extensions->ARB_gpu_shader_fp64 = GL_TRUE;
+      extensions->ARB_vertex_attrib_64bit = GL_TRUE;
+   }
 }
diff --git a/mesalib/src/mesa/state_tracker/st_format.c b/mesalib/src/mesa/state_tracker/st_format.c
index 181465dd8..db7b5b71d 100644
--- a/mesalib/src/mesa/state_tracker/st_format.c
+++ b/mesalib/src/mesa/state_tracker/st_format.c
@@ -991,7 +991,7 @@ static const struct format_mapping format_map[] = {
    {
       { GL_RGB10, 0 },
       { PIPE_FORMAT_B10G10R10X2_UNORM, PIPE_FORMAT_B10G10R10A2_UNORM,
-        DEFAULT_RGB_FORMATS }
+        PIPE_FORMAT_R10G10B10A2_UNORM, DEFAULT_RGB_FORMATS }
    },
    {
       { GL_RGB10_A2, 0 },
diff --git a/mesalib/src/mesa/state_tracker/st_glsl_to_tgsi.cpp b/mesalib/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
index 93671ba9c..f0f2a77d0 100644
--- a/mesalib/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
+++ b/mesalib/src/mesa/state_tracker/st_glsl_to_tgsi.cpp
@@ -88,6 +88,7 @@ public:
       this->reladdr = NULL;
       this->reladdr2 = NULL;
       this->has_index2 = false;
+      this->double_reg2 = false;
    }
 
    st_src_reg(gl_register_file file, int index, int type)
@@ -101,6 +102,7 @@ public:
       this->reladdr = NULL;
       this->reladdr2 = NULL;
       this->has_index2 = false;
+      this->double_reg2 = false;
    }
 
    st_src_reg(gl_register_file file, int index, int type, int index2D)
@@ -114,6 +116,7 @@ public:
       this->reladdr = NULL;
       this->reladdr2 = NULL;
       this->has_index2 = false;
+      this->double_reg2 = false;
    }
 
    st_src_reg()
@@ -127,6 +130,7 @@ public:
       this->reladdr = NULL;
       this->reladdr2 = NULL;
       this->has_index2 = false;
+      this->double_reg2 = false;
    }
 
    explicit st_src_reg(st_dst_reg reg);
@@ -141,6 +145,11 @@ public:
    st_src_reg *reladdr;
    st_src_reg *reladdr2;
    bool has_index2;
+   /*
+    * Is this the second half of a double register pair?
+    * currently used for input mapping only.
+    */
+   bool double_reg2;
 };
 
 class st_dst_reg {
@@ -197,6 +206,7 @@ st_src_reg::st_src_reg(st_dst_reg reg)
    this->index2D = 0;
    this->reladdr2 = NULL;
    this->has_index2 = false;
+   this->double_reg2 = false;
 }
 
 st_dst_reg::st_dst_reg(st_src_reg reg)
@@ -677,8 +687,10 @@ glsl_to_tgsi_visitor::emit(ir_instruction *ir, unsigned op,
 
             if (dinst->src[j].type == GLSL_TYPE_DOUBLE) {
                dinst->src[j].index = initial_src_idx[j];
-               if (swz > 1)
+               if (swz > 1) {
+                  dinst->src[j].double_reg2 = true;
                   dinst->src[j].index++;
+	       }
 
                if (swz & 1)
                   dinst->src[j].swizzle = MAKE_SWIZZLE4(SWIZZLE_Z, SWIZZLE_W, SWIZZLE_Z, SWIZZLE_W);
@@ -1941,7 +1953,7 @@ glsl_to_tgsi_visitor::visit(ir_expression *ir)
       break;
    case ir_unop_i2b:
       if (native_integers)
-         emit(ir, TGSI_OPCODE_INEG, result_dst, op[0]);
+         emit(ir, TGSI_OPCODE_USNE, result_dst, op[0], st_src_reg_for_int(0));
       else
          emit(ir, TGSI_OPCODE_SNE, result_dst, op[0], st_src_reg_for_float(0.0));
       break;
@@ -2611,6 +2623,7 @@ glsl_to_tgsi_visitor::visit(ir_assignment *ir)
       assert(!ir->lhs->type->is_scalar() && !ir->lhs->type->is_vector());
       l.writemask = WRITEMASK_XYZW;
    } else if (ir->lhs->type->is_scalar() &&
+              !ir->lhs->type->is_double() &&
               ir->lhs->variable_referenced()->data.mode == ir_var_shader_out) {
       /* FINISHME: This hack makes writing to gl_FragDepth, which lives in the
        * FINISHME: W component of fragment shader output zero, work correctly.
@@ -3704,6 +3717,7 @@ glsl_to_tgsi_visitor::copy_propagate(void)
             } else {
                if (first->src[0].file != copy_chan->src[0].file ||
                    first->src[0].index != copy_chan->src[0].index ||
+                   first->src[0].double_reg2 != copy_chan->src[0].double_reg2 ||
                    first->src[0].index2D != copy_chan->src[0].index2D) {
                   good = false;
                   break;
@@ -3719,6 +3733,7 @@ glsl_to_tgsi_visitor::copy_propagate(void)
             inst->src[r].index = first->src[0].index;
             inst->src[r].index2D = first->src[0].index2D;
             inst->src[r].has_index2 = first->src[0].has_index2;
+            inst->src[r].double_reg2 = first->src[0].double_reg2;
 
             int swizzle = 0;
             for (int i = 0; i < 4; i++) {
@@ -4551,6 +4566,9 @@ dst_register(struct st_translate *t,
 static struct ureg_src
 src_register(struct st_translate *t, const st_src_reg *reg)
 {
+   int index = reg->index;
+   int double_reg2 = reg->double_reg2 ? 1 : 0;
+
    switch(reg->file) {
    case PROGRAM_UNDEFINED:
       return ureg_imm4f(t->ureg, 0, 0, 0, 0);
@@ -4576,8 +4594,12 @@ src_register(struct st_translate *t, const st_src_reg *reg)
       return t->immediates[reg->index];
 
    case PROGRAM_INPUT:
-      assert(t->inputMapping[reg->index] < ARRAY_SIZE(t->inputs));
-      return t->inputs[t->inputMapping[reg->index]];
+      /* GLSL inputs are 64-bit containers, so we have to
+       * map back to the original index and add the offset after
+       * mapping. */
+      index -= double_reg2;
+      assert(t->inputMapping[index] < ARRAY_SIZE(t->inputs));
+      return t->inputs[t->inputMapping[index] + double_reg2];
 
    case PROGRAM_OUTPUT:
       assert(t->outputMapping[reg->index] < ARRAY_SIZE(t->outputs));
diff --git a/mesalib/src/mesa/state_tracker/st_manager.c b/mesalib/src/mesa/state_tracker/st_manager.c
index 840f76a13..0376954f7 100644
--- a/mesalib/src/mesa/state_tracker/st_manager.c
+++ b/mesalib/src/mesa/state_tracker/st_manager.c
@@ -680,6 +680,10 @@ st_api_create_context(struct st_api *stapi, struct st_manager *smapi,
 
    if (attribs->flags & ST_CONTEXT_FLAG_FORWARD_COMPATIBLE)
       st->ctx->Const.ContextFlags |= GL_CONTEXT_FLAG_FORWARD_COMPATIBLE_BIT;
+   if (attribs->flags & ST_CONTEXT_FLAG_ROBUST_ACCESS)
+      st->ctx->Const.ContextFlags |= GL_CONTEXT_FLAG_ROBUST_ACCESS_BIT_ARB;
+   if (attribs->flags & ST_CONTEXT_FLAG_RESET_NOTIFICATION_ENABLED)
+      st->ctx->Const.ResetStrategy = GL_LOSE_CONTEXT_ON_RESET_ARB;
 
    /* need to perform version check */
    if (attribs->major > 1 || attribs->minor > 0) {
diff --git a/mesalib/src/mesa/state_tracker/st_program.c b/mesalib/src/mesa/state_tracker/st_program.c
index d93b3c7bc..a9110d3c6 100644
--- a/mesalib/src/mesa/state_tracker/st_program.c
+++ b/mesalib/src/mesa/state_tracker/st_program.c
@@ -194,6 +194,11 @@ st_prepare_vertex_program(struct gl_context *ctx,
          stvp->input_to_index[attr] = stvp->num_inputs;
          stvp->index_to_input[stvp->num_inputs] = attr;
          stvp->num_inputs++;
+         if ((stvp->Base.Base.DoubleInputsRead & BITFIELD64_BIT(attr)) != 0) {
+            /* add placeholder for second part of a double attribute */
+            stvp->index_to_input[stvp->num_inputs] = ST_DOUBLE_ATTRIB_PLACEHOLDER;
+            stvp->num_inputs++;
+         }
       }
    }
    /* bit of a hack, presetup potentially unused edgeflag input */
diff --git a/mesalib/src/mesa/state_tracker/st_program.h b/mesalib/src/mesa/state_tracker/st_program.h
index b2c86faec..a2c56062d 100644
--- a/mesalib/src/mesa/state_tracker/st_program.h
+++ b/mesalib/src/mesa/state_tracker/st_program.h
@@ -45,6 +45,7 @@
 extern "C" {
 #endif
 
+#define ST_DOUBLE_ATTRIB_PLACEHOLDER 0xffffffff
 
 /** Fragment program variant key */
 struct st_fp_variant_key