gallium/util: replace pipe_thread_setname() with u_thread_setname()

[mesa.git] / src / gallium / drivers / llvmpipe / lp_rast.c

diff --git a/src/gallium/drivers/llvmpipe/lp_rast.c b/src/gallium/drivers/llvmpipe/lp_rast.c
index 7019acbda9d8c9cf4ac46f026e1b3eaf6e0c8d52..b25ade3d2ecf4f96cfc77e1c357925eedb2f2110 100644 (file)
--- a/src/gallium/drivers/llvmpipe/lp_rast.c
+++ b/src/gallium/drivers/llvmpipe/lp_rast.c
@@ -43,6 +43,7 @@
 #include "lp_query.h"
 #include "lp_rast.h"
 #include "lp_rast_priv.h"
+#include "gallivm/lp_bld_format.h"
 #include "gallivm/lp_bld_debug.h"
 #include "lp_scene.h"
 #include "lp_tex_sample.h"
@@ -82,7 +83,7 @@ lp_rast_end( struct lp_rasterizer *rast )
 
 
 /**
- * Begining rasterization of a tile.
+ * Beginning rasterization of a tile.
  * \param x  window X position of the tile, in pixels
  * \param y  window Y position of the tile, in pixels
  */
@@ -664,6 +665,17 @@ rasterize_scene(struct lp_rasterizer_task *task,
 {
    task->scene = scene;
 
+   /* Clear the cache tags. This should not always be necessary but
+      simpler for now. */
+#if LP_USE_TEXTURE_CACHE
+   memset(task->thread_data.cache->cache_tags, 0,
+          sizeof(task->thread_data.cache->cache_tags));
+#if LP_BUILD_FORMAT_CACHE_DEBUG
+   task->thread_data.cache->cache_access_total = 0;
+   task->thread_data.cache->cache_access_miss = 0;
+#endif
+#endif
+
    if (!task->rast->no_rast && !scene->discard) {
       /* loop over scene bins, rasterize each */
       {
@@ -679,6 +691,20 @@ rasterize_scene(struct lp_rasterizer_task *task,
    }
 
 
+#if LP_BUILD_FORMAT_CACHE_DEBUG
+   {
+      uint64_t total, miss;
+      total = task->thread_data.cache->cache_access_total;
+      miss = task->thread_data.cache->cache_access_miss;
+      if (total) {
+         debug_printf("thread %d cache access %llu miss %llu hit rate %f\n",
+                 task->thread_index, (long long unsigned)total,
+                 (long long unsigned)miss,
+                 (float)(total - miss)/(float)total);
+      }
+   }
+#endif
+
    if (scene->fence) {
       lp_fence_signal(scene->fence);
    }
@@ -755,7 +781,8 @@ lp_rast_finish( struct lp_rasterizer *rast )
  *   2. do work
  *   3. signal that we're done
  */
-static PIPE_THREAD_ROUTINE( thread_function, init_data )
+static int
+thread_function(void *init_data)
 {
    struct lp_rasterizer_task *task = (struct lp_rasterizer_task *) init_data;
    struct lp_rasterizer *rast = task->rast;
@@ -764,7 +791,7 @@ static PIPE_THREAD_ROUTINE( thread_function, init_data )
    unsigned fpstate;
 
    util_snprintf(thread_name, sizeof thread_name, "llvmpipe-%u", task->thread_index);
-   pipe_thread_setname(thread_name);
+   u_thread_setname(thread_name);
 
    /* Make sure that denorms are treated like zeros. This is 
     * the behavior required by D3D10. OpenGL doesn't care.
@@ -838,7 +865,7 @@ create_rast_threads(struct lp_rasterizer *rast)
    for (i = 0; i < rast->num_threads; i++) {
       pipe_semaphore_init(&rast->tasks[i].work_ready, 0);
       pipe_semaphore_init(&rast->tasks[i].work_done, 0);
-      rast->threads[i] = pipe_thread_create(thread_function,
+      rast->threads[i] = u_thread_create(thread_function,
                                             (void *) &rast->tasks[i]);
    }
 }
@@ -866,10 +893,15 @@ lp_rast_create( unsigned num_threads )
       goto no_full_scenes;
    }
 
-   for (i = 0; i < Elements(rast->tasks); i++) {
+   for (i = 0; i < MAX2(1, num_threads); i++) {
       struct lp_rasterizer_task *task = &rast->tasks[i];
       task->rast = rast;
       task->thread_index = i;
+      task->thread_data.cache = align_malloc(sizeof(struct lp_build_format_cache),
+                                             16);
+      if (!task->thread_data.cache) {
+         goto no_thread_data_cache;
+      }
    }
 
    rast->num_threads = num_threads;
@@ -879,12 +911,22 @@ lp_rast_create( unsigned num_threads )
    create_rast_threads(rast);
 
    /* for synchronizing rasterization threads */
-   pipe_barrier_init( &rast->barrier, rast->num_threads );
+   if (rast->num_threads > 0) {
+      pipe_barrier_init( &rast->barrier, rast->num_threads );
+   }
 
    memset(lp_dummy_tile, 0, sizeof lp_dummy_tile);
 
    return rast;
 
+no_thread_data_cache:
+   for (i = 0; i < MAX2(1, rast->num_threads); i++) {
+      if (rast->tasks[i].thread_data.cache) {
+         align_free(rast->tasks[i].thread_data.cache);
+      }
+   }
+
+   lp_scene_queue_destroy(rast->full_scenes);
 no_full_scenes:
    FREE(rast);
 no_rast:
@@ -914,7 +956,7 @@ void lp_rast_destroy( struct lp_rasterizer *rast )
 #ifdef _WIN32
       pipe_semaphore_wait(&rast->tasks[i].work_done);
 #else
-      pipe_thread_wait(rast->threads[i]);
+      thrd_join(rast->threads[i], NULL);
 #endif
    }
 
@@ -923,9 +965,14 @@ void lp_rast_destroy( struct lp_rasterizer *rast )
       pipe_semaphore_destroy(&rast->tasks[i].work_ready);
       pipe_semaphore_destroy(&rast->tasks[i].work_done);
    }
+   for (i = 0; i < MAX2(1, rast->num_threads); i++) {
+      align_free(rast->tasks[i].thread_data.cache);
+   }
 
    /* for synchronizing rasterization threads */
-   pipe_barrier_destroy( &rast->barrier );
+   if (rast->num_threads > 0) {
+      pipe_barrier_destroy( &rast->barrier );
+   }
 
    lp_scene_queue_destroy(rast->full_scenes);