st/mesa: fix lowered IO - don't call st_nir_assign_vs_in_locations twice
[mesa.git] / src / mesa / state_tracker / st_cb_queryobj.c
index 2281d10e99960c5b4c3f9f9792dac628a151ffb0..50a776df74566a8bcc2433fb167e5cfd57a7c961 100644 (file)
@@ -1,8 +1,8 @@
 /**************************************************************************
- * 
- * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ *
+ * Copyright 2007 VMware, Inc.
  * All Rights Reserved.
- * 
+ *
  * 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
  * distribute, sub license, 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 (including the
  * next paragraph) 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 NON-INFRINGEMENT.
- * IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS 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.
- * 
+ *
  **************************************************************************/
 
 
  */
 
 
-#include "main/imports.h"
+#include "util/compiler.h"
 #include "main/context.h"
-#include "main/image.h"
+#include "main/queryobj.h"
 
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
+#include "pipe/p_screen.h"
+#include "util/u_inlines.h"
 #include "st_context.h"
 #include "st_cb_queryobj.h"
+#include "st_cb_bitmap.h"
+#include "st_cb_bufferobjects.h"
+#include "st_util.h"
 
 
 static struct gl_query_object *
-st_NewQueryObject(GLcontext *ctx, GLuint id)
+st_NewQueryObject(struct gl_context *ctx, GLuint id)
 {
    struct st_query_object *stq = ST_CALLOC_STRUCT(st_query_object);
    if (stq) {
       stq->base.Id = id;
       stq->base.Ready = GL_TRUE;
       stq->pq = NULL;
+      stq->type = PIPE_QUERY_TYPES; /* an invalid value */
       return &stq->base;
    }
    return NULL;
 }
 
 
-
 static void
-st_DeleteQuery(GLcontext *ctx, struct gl_query_object *q)
+free_queries(struct pipe_context *pipe, struct st_query_object *stq)
 {
-   struct pipe_context *pipe = ctx->st->pipe;
-   struct st_query_object *stq = st_query_object(q);
-
    if (stq->pq) {
       pipe->destroy_query(pipe, stq->pq);
       stq->pq = NULL;
    }
 
-   _mesa_free(stq);
+   if (stq->pq_begin) {
+      pipe->destroy_query(pipe, stq->pq_begin);
+      stq->pq_begin = NULL;
+   }
 }
 
 
 static void
-st_BeginQuery(GLcontext *ctx, struct gl_query_object *q)
+st_DeleteQuery(struct gl_context *ctx, struct gl_query_object *q)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_query_object *stq = st_query_object(q);
+
+   free_queries(pipe, stq);
+
+   _mesa_delete_query(ctx, q);
+}
+
+static int
+target_to_index(const struct st_context *st, const struct gl_query_object *q)
+{
+   if (q->Target == GL_PRIMITIVES_GENERATED ||
+       q->Target == GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN ||
+       q->Target == GL_TRANSFORM_FEEDBACK_STREAM_OVERFLOW_ARB)
+      return q->Stream;
+
+   if (st->has_single_pipe_stat) {
+      switch (q->Target) {
+      case GL_VERTICES_SUBMITTED_ARB:
+         return PIPE_STAT_QUERY_IA_VERTICES;
+      case GL_PRIMITIVES_SUBMITTED_ARB:
+         return PIPE_STAT_QUERY_IA_PRIMITIVES;
+      case GL_VERTEX_SHADER_INVOCATIONS_ARB:
+         return PIPE_STAT_QUERY_VS_INVOCATIONS;
+      case GL_GEOMETRY_SHADER_INVOCATIONS:
+         return PIPE_STAT_QUERY_GS_INVOCATIONS;
+      case GL_GEOMETRY_SHADER_PRIMITIVES_EMITTED_ARB:
+         return PIPE_STAT_QUERY_GS_PRIMITIVES;
+      case GL_CLIPPING_INPUT_PRIMITIVES_ARB:
+         return PIPE_STAT_QUERY_C_INVOCATIONS;
+      case GL_CLIPPING_OUTPUT_PRIMITIVES_ARB:
+         return PIPE_STAT_QUERY_C_PRIMITIVES;
+      case GL_FRAGMENT_SHADER_INVOCATIONS_ARB:
+         return PIPE_STAT_QUERY_PS_INVOCATIONS;
+      case GL_TESS_CONTROL_SHADER_PATCHES_ARB:
+         return PIPE_STAT_QUERY_HS_INVOCATIONS;
+      case GL_TESS_EVALUATION_SHADER_INVOCATIONS_ARB:
+         return PIPE_STAT_QUERY_DS_INVOCATIONS;
+      case GL_COMPUTE_SHADER_INVOCATIONS_ARB:
+         return PIPE_STAT_QUERY_CS_INVOCATIONS;
+      default:
+         break;
+      }
+   }
+
+   return 0;
+}
+
+static void
+st_BeginQuery(struct gl_context *ctx, struct gl_query_object *q)
 {
-   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_context *st = st_context(ctx);
+   struct pipe_context *pipe = st->pipe;
    struct st_query_object *stq = st_query_object(q);
+   unsigned type;
+   bool ret = false;
+
+   st_flush_bitmap_cache(st_context(ctx));
 
+   /* convert GL query type to Gallium query type */
    switch (q->Target) {
+   case GL_ANY_SAMPLES_PASSED:
+      type = PIPE_QUERY_OCCLUSION_PREDICATE;
+      break;
+   case GL_ANY_SAMPLES_PASSED_CONSERVATIVE:
+      type = PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE;
+      break;
    case GL_SAMPLES_PASSED_ARB:
-      if (!stq->pq)
-        stq->pq = pipe->create_query( pipe, PIPE_QUERY_OCCLUSION_COUNTER );
+      type = PIPE_QUERY_OCCLUSION_COUNTER;
+      break;
+   case GL_PRIMITIVES_GENERATED:
+      type = PIPE_QUERY_PRIMITIVES_GENERATED;
+      break;
+   case GL_TRANSFORM_FEEDBACK_PRIMITIVES_WRITTEN:
+      type = PIPE_QUERY_PRIMITIVES_EMITTED;
+      break;
+   case GL_TRANSFORM_FEEDBACK_STREAM_OVERFLOW_ARB:
+      type = PIPE_QUERY_SO_OVERFLOW_PREDICATE;
+      break;
+   case GL_TRANSFORM_FEEDBACK_OVERFLOW_ARB:
+      type = PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE;
+      break;
+   case GL_TIME_ELAPSED:
+      if (st->has_time_elapsed)
+         type = PIPE_QUERY_TIME_ELAPSED;
+      else
+         type = PIPE_QUERY_TIMESTAMP;
+      break;
+   case GL_VERTICES_SUBMITTED_ARB:
+   case GL_PRIMITIVES_SUBMITTED_ARB:
+   case GL_VERTEX_SHADER_INVOCATIONS_ARB:
+   case GL_TESS_CONTROL_SHADER_PATCHES_ARB:
+   case GL_TESS_EVALUATION_SHADER_INVOCATIONS_ARB:
+   case GL_GEOMETRY_SHADER_INVOCATIONS:
+   case GL_GEOMETRY_SHADER_PRIMITIVES_EMITTED_ARB:
+   case GL_FRAGMENT_SHADER_INVOCATIONS_ARB:
+   case GL_COMPUTE_SHADER_INVOCATIONS_ARB:
+   case GL_CLIPPING_INPUT_PRIMITIVES_ARB:
+   case GL_CLIPPING_OUTPUT_PRIMITIVES_ARB:
+      type = st->has_single_pipe_stat ? PIPE_QUERY_PIPELINE_STATISTICS_SINGLE
+                                      : PIPE_QUERY_PIPELINE_STATISTICS;
       break;
    default:
-      assert(0);
+      assert(0 && "unexpected query target in st_BeginQuery()");
       return;
    }
 
-   pipe->begin_query(pipe, stq->pq);
+   if (stq->type != type) {
+      /* free old query of different type */
+      free_queries(pipe, stq);
+      stq->type = PIPE_QUERY_TYPES; /* an invalid value */
+   }
+
+   if (q->Target == GL_TIME_ELAPSED &&
+       type == PIPE_QUERY_TIMESTAMP) {
+      /* Determine time elapsed by emitting two timestamp queries. */
+      if (!stq->pq_begin) {
+         stq->pq_begin = pipe->create_query(pipe, type, 0);
+         stq->type = type;
+      }
+      if (stq->pq_begin)
+         ret = pipe->end_query(pipe, stq->pq_begin);
+   } else {
+      if (!stq->pq) {
+         stq->pq = pipe->create_query(pipe, type, target_to_index(st, q));
+         stq->type = type;
+      }
+      if (stq->pq)
+         ret = pipe->begin_query(pipe, stq->pq);
+   }
+
+   if (!ret) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glBeginQuery");
+
+      free_queries(pipe, stq);
+      q->Active = GL_FALSE;
+      return;
+   }
+
+   if (stq->type != PIPE_QUERY_TIMESTAMP)
+      st->active_queries++;
+
+   assert(stq->type == type);
 }
 
 
 static void
-st_EndQuery(GLcontext *ctx, struct gl_query_object *q)
+st_EndQuery(struct gl_context *ctx, struct gl_query_object *q)
 {
-   struct pipe_context *pipe = ctx->st->pipe;
+   struct st_context *st = st_context(ctx);
+   struct pipe_context *pipe = st->pipe;
    struct st_query_object *stq = st_query_object(q);
+   bool ret = false;
+
+   st_flush_bitmap_cache(st_context(ctx));
+
+   if ((q->Target == GL_TIMESTAMP ||
+        q->Target == GL_TIME_ELAPSED) &&
+       !stq->pq) {
+      stq->pq = pipe->create_query(pipe, PIPE_QUERY_TIMESTAMP, 0);
+      stq->type = PIPE_QUERY_TIMESTAMP;
+   }
+
+   if (stq->pq)
+      ret = pipe->end_query(pipe, stq->pq);
+
+   if (!ret) {
+      _mesa_error(ctx, GL_OUT_OF_MEMORY, "glEndQuery");
+      return;
+   }
+
+   if (stq->type != PIPE_QUERY_TIMESTAMP)
+      st->active_queries--;
+}
+
+
+static boolean
+get_query_result(struct pipe_context *pipe,
+                 struct st_query_object *stq,
+                 boolean wait)
+{
+   union pipe_query_result data;
+
+   if (!stq->pq) {
+      /* Only needed in case we failed to allocate the gallium query earlier.
+       * Return TRUE so we don't spin on this forever.
+       */
+      return TRUE;
+   }
+
+   if (!pipe->get_query_result(pipe, stq->pq, wait, &data))
+      return FALSE;
+
+   switch (stq->type) {
+   case PIPE_QUERY_PIPELINE_STATISTICS:
+      switch (stq->base.Target) {
+      case GL_VERTICES_SUBMITTED_ARB:
+         stq->base.Result = data.pipeline_statistics.ia_vertices;
+         break;
+      case GL_PRIMITIVES_SUBMITTED_ARB:
+         stq->base.Result = data.pipeline_statistics.ia_primitives;
+         break;
+      case GL_VERTEX_SHADER_INVOCATIONS_ARB:
+         stq->base.Result = data.pipeline_statistics.vs_invocations;
+         break;
+      case GL_TESS_CONTROL_SHADER_PATCHES_ARB:
+         stq->base.Result = data.pipeline_statistics.hs_invocations;
+         break;
+      case GL_TESS_EVALUATION_SHADER_INVOCATIONS_ARB:
+         stq->base.Result = data.pipeline_statistics.ds_invocations;
+         break;
+      case GL_GEOMETRY_SHADER_INVOCATIONS:
+         stq->base.Result = data.pipeline_statistics.gs_invocations;
+         break;
+      case GL_GEOMETRY_SHADER_PRIMITIVES_EMITTED_ARB:
+         stq->base.Result = data.pipeline_statistics.gs_primitives;
+         break;
+      case GL_FRAGMENT_SHADER_INVOCATIONS_ARB:
+         stq->base.Result = data.pipeline_statistics.ps_invocations;
+         break;
+      case GL_COMPUTE_SHADER_INVOCATIONS_ARB:
+         stq->base.Result = data.pipeline_statistics.cs_invocations;
+         break;
+      case GL_CLIPPING_INPUT_PRIMITIVES_ARB:
+         stq->base.Result = data.pipeline_statistics.c_invocations;
+         break;
+      case GL_CLIPPING_OUTPUT_PRIMITIVES_ARB:
+         stq->base.Result = data.pipeline_statistics.c_primitives;
+         break;
+      default:
+         unreachable("invalid pipeline statistics counter");
+      }
+      break;
+   case PIPE_QUERY_OCCLUSION_PREDICATE:
+   case PIPE_QUERY_OCCLUSION_PREDICATE_CONSERVATIVE:
+   case PIPE_QUERY_SO_OVERFLOW_PREDICATE:
+   case PIPE_QUERY_SO_OVERFLOW_ANY_PREDICATE:
+      stq->base.Result = !!data.b;
+      break;
+   default:
+      stq->base.Result = data.u64;
+      break;
+   }
+
+   if (stq->base.Target == GL_TIME_ELAPSED &&
+       stq->type == PIPE_QUERY_TIMESTAMP) {
+      /* Calculate the elapsed time from the two timestamp queries */
+      GLuint64EXT Result0 = 0;
+      assert(stq->pq_begin);
+      pipe->get_query_result(pipe, stq->pq_begin, TRUE, (void *)&Result0);
+      stq->base.Result -= Result0;
+   } else {
+      assert(!stq->pq_begin);
+   }
 
-   pipe->end_query(pipe, stq->pq);
+   return TRUE;
 }
 
 
 static void
-st_WaitQuery(GLcontext *ctx, struct gl_query_object *q)
+st_WaitQuery(struct gl_context *ctx, struct gl_query_object *q)
 {
-   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_context *pipe = st_context(ctx)->pipe;
    struct st_query_object *stq = st_query_object(q);
 
    /* this function should only be called if we don't have a ready result */
    assert(!stq->base.Ready);
 
    while (!stq->base.Ready &&
-         !pipe->get_query_result(pipe, 
-                                 stq->pq,
-                                 TRUE,
-                                 &q->Result))
+         !get_query_result(pipe, stq, TRUE))
    {
       /* nothing */
    }
-                           
+
    q->Ready = GL_TRUE;
 }
 
 
 static void
-st_CheckQuery(GLcontext *ctx, struct gl_query_object *q)
+st_CheckQuery(struct gl_context *ctx, struct gl_query_object *q)
 {
-   struct pipe_context *pipe = ctx->st->pipe;
+   struct pipe_context *pipe = st_context(ctx)->pipe;
    struct st_query_object *stq = st_query_object(q);
+   assert(!q->Ready);   /* we should not get called if Ready is TRUE */
+   q->Ready = get_query_result(pipe, stq, FALSE);
+}
+
 
-   if (!q->Ready) {
-      q->Ready = pipe->get_query_result(pipe, 
-                                       stq->pq,
-                                       FALSE,
-                                       &q->Result);
+static uint64_t
+st_GetTimestamp(struct gl_context *ctx)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct pipe_screen *screen = pipe->screen;
+
+   /* Prefer the per-screen function */
+   if (screen->get_timestamp) {
+      return screen->get_timestamp(screen);
+   }
+   else {
+      /* Fall back to the per-context function */
+      assert(pipe->get_timestamp);
+      return pipe->get_timestamp(pipe);
    }
 }
 
+static void
+st_StoreQueryResult(struct gl_context *ctx, struct gl_query_object *q,
+                    struct gl_buffer_object *buf, intptr_t offset,
+                    GLenum pname, GLenum ptype)
+{
+   struct pipe_context *pipe = st_context(ctx)->pipe;
+   struct st_query_object *stq = st_query_object(q);
+   struct st_buffer_object *stObj = st_buffer_object(buf);
+   boolean wait = pname == GL_QUERY_RESULT;
+   enum pipe_query_value_type result_type;
+   int index;
 
+   /* GL_QUERY_TARGET is a bit of an extension since it has nothing to
+    * do with the GPU end of the query. Write it in "by hand".
+    */
+   if (pname == GL_QUERY_TARGET) {
+      /* Assume that the data must be LE. The endianness situation wrt CPU and
+       * GPU is incredibly confusing, but the vast majority of GPUs are
+       * LE. When a BE one comes along, this needs some form of resolution.
+       */
+      unsigned data[2] = { CPU_TO_LE32(q->Target), 0 };
+      pipe_buffer_write(pipe, stObj->buffer, offset,
+                        (ptype == GL_INT64_ARB ||
+                         ptype == GL_UNSIGNED_INT64_ARB) ? 8 : 4,
+                        data);
+      return;
+   }
 
+   switch (ptype) {
+   case GL_INT:
+      result_type = PIPE_QUERY_TYPE_I32;
+      break;
+   case GL_UNSIGNED_INT:
+      result_type = PIPE_QUERY_TYPE_U32;
+      break;
+   case GL_INT64_ARB:
+      result_type = PIPE_QUERY_TYPE_I64;
+      break;
+   case GL_UNSIGNED_INT64_ARB:
+      result_type = PIPE_QUERY_TYPE_U64;
+      break;
+   default:
+      unreachable("Unexpected result type");
+   }
+
+   if (pname == GL_QUERY_RESULT_AVAILABLE) {
+      index = -1;
+   } else if (stq->type == PIPE_QUERY_PIPELINE_STATISTICS) {
+      switch (q->Target) {
+      case GL_VERTICES_SUBMITTED_ARB:
+         index = PIPE_STAT_QUERY_IA_VERTICES;
+         break;
+      case GL_PRIMITIVES_SUBMITTED_ARB:
+         index = PIPE_STAT_QUERY_IA_PRIMITIVES;
+         break;
+      case GL_VERTEX_SHADER_INVOCATIONS_ARB:
+         index = PIPE_STAT_QUERY_VS_INVOCATIONS;
+         break;
+      case GL_GEOMETRY_SHADER_INVOCATIONS:
+         index = PIPE_STAT_QUERY_GS_INVOCATIONS;
+         break;
+      case GL_GEOMETRY_SHADER_PRIMITIVES_EMITTED_ARB:
+         index = PIPE_STAT_QUERY_GS_PRIMITIVES;
+         break;
+      case GL_CLIPPING_INPUT_PRIMITIVES_ARB:
+         index = PIPE_STAT_QUERY_C_INVOCATIONS;
+         break;
+      case GL_CLIPPING_OUTPUT_PRIMITIVES_ARB:
+         index = PIPE_STAT_QUERY_C_PRIMITIVES;
+         break;
+      case GL_FRAGMENT_SHADER_INVOCATIONS_ARB:
+         index = PIPE_STAT_QUERY_PS_INVOCATIONS;
+         break;
+      case GL_TESS_CONTROL_SHADER_PATCHES_ARB:
+         index = PIPE_STAT_QUERY_HS_INVOCATIONS;
+         break;
+      case GL_TESS_EVALUATION_SHADER_INVOCATIONS_ARB:
+         index = PIPE_STAT_QUERY_DS_INVOCATIONS;
+         break;
+      case GL_COMPUTE_SHADER_INVOCATIONS_ARB:
+         index = PIPE_STAT_QUERY_CS_INVOCATIONS;
+         break;
+      default:
+         unreachable("Unexpected target");
+      }
+   } else {
+      index = 0;
+   }
+
+   pipe->get_query_result_resource(pipe, stq->pq, wait, result_type, index,
+                                   stObj->buffer, offset);
+}
 
 void st_init_query_functions(struct dd_function_table *functions)
 {
@@ -150,4 +487,6 @@ void st_init_query_functions(struct dd_function_table *functions)
    functions->EndQuery = st_EndQuery;
    functions->WaitQuery = st_WaitQuery;
    functions->CheckQuery = st_CheckQuery;
+   functions->GetTimestamp = st_GetTimestamp;
+   functions->StoreQueryResult = st_StoreQueryResult;
 }