winsys/amdgpu: make RADEON_ALL_BOS a debug only feature

Improves performance in SPECviewperf13 snx. e.g.: test10 fps evolution: 235 -> 270. Extract from "pahole radeonsi_dri.so -C amdgpu_winsys_bo", before: struct amdgpu_winsys_bo { struct pb_buffer base; /* 0 32 */ union { struct { struct pb_cache_entry cache_entry; /* 32 56 */ /* XXX last struct has 4 bytes of padding */ /* --- cacheline 1 boundary (64 bytes) was 24 bytes ago --- */ amdgpu_va_handle va_handle; /* 88 8 */ int map_count; /* 96 4 */ _Bool use_reusable_pool; /* 100 1 */ /* XXX 3 bytes hole, try to pack */ struct list_head global_list_item; /* 104 16 */ uint32_t kms_handle; /* 120 4 */ } real; [...] } u; /* 32 96 */ [...] /* size: 200, cachelines: 4, members: 15 */ }; After: struct amdgpu_winsys_bo { struct pb_buffer base; /* 0 32 */ union { struct { struct pb_cache_entry cache_entry; /* 32 56 */ /* XXX last struct has 4 bytes of padding */ /* --- cacheline 1 boundary (64 bytes) was 24 bytes ago --- */ amdgpu_va_handle va_handle; /* 88 8 */ int map_count; /* 96 4 */ _Bool use_reusable_pool; /* 100 1 */ /* XXX 3 bytes hole, try to pack */ uint32_t kms_handle; /* 104 4 */ } real; /* 32 80 */ } u; /* 32 80 */ /* --- cacheline 1 boundary (64 bytes) was 48 bytes ago --- */ [...] /* size: 184, cachelines: 3, members: 15 */ }; Reviewed-by: Marek Olšák <marek.olsak@amd.com> Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/7532>
2020-11-13 12:19:14 +01:00
parent 631e18d427
commit 111a1b2e1c
5 changed files with 23 additions and 5 deletions
@@ -176,12 +176,14 @@ void amdgpu_bo_destroy(struct pb_buffer *_buf)
   }
   assert(bo->is_user_ptr || bo->u.real.map_count == 0);

+#if DEBUG
   if (ws->debug_all_bos) {
      simple_mtx_lock(&ws->global_bo_list_lock);
      list_del(&bo->u.real.global_list_item);
      ws->num_buffers--;
      simple_mtx_unlock(&ws->global_bo_list_lock);
   }
+#endif

   /* Close all KMS handles retrieved for other DRM file descriptions */
   simple_mtx_lock(&ws->sws_list_lock);
@@ -431,6 +433,7 @@ static const struct pb_vtbl amdgpu_winsys_bo_vtbl = {

 static void amdgpu_add_buffer_to_global_list(struct amdgpu_winsys_bo *bo)
 {
+#if DEBUG
   struct amdgpu_winsys *ws = bo->ws;

   assert(bo->bo);
@@ -441,6 +444,7 @@ static void amdgpu_add_buffer_to_global_list(struct amdgpu_winsys_bo *bo)
      ws->num_buffers++;
      simple_mtx_unlock(&ws->global_bo_list_lock);
   }
+#endif
 }

 static unsigned amdgpu_get_optimal_alignment(struct amdgpu_winsys *ws,
@@ -64,9 +64,9 @@ struct amdgpu_winsys_bo {
         amdgpu_va_handle va_handle;
         int map_count;
         bool use_reusable_pool;
-
+#if DEBUG
         struct list_head global_list_item;
-
+#endif
         uint32_t kms_handle;
      } real;
      struct {
@@ -1430,6 +1430,7 @@ static void amdgpu_cs_submit_ib(void *job, int thread_index)
   bool use_bo_list_create = ws->info.drm_minor < 27;
   struct drm_amdgpu_bo_list_in bo_list_in;

+#if DEBUG
   /* Prepare the buffer list. */
   if (ws->debug_all_bos) {
      /* The buffer list contains all buffers. This is a slow path that
@@ -1453,7 +1454,9 @@ static void amdgpu_cs_submit_ib(void *job, int thread_index)
         fprintf(stderr, "amdgpu: buffer list creation failed (%d)\n", r);
         goto cleanup;
      }
-   } else {
+   } else
+#endif
+   {
      if (!amdgpu_add_sparse_backing_buffers(cs)) {
         fprintf(stderr, "amdgpu: amdgpu_add_sparse_backing_buffers failed\n");
         r = -ENOMEM;
@@ -47,7 +47,9 @@
 static struct hash_table *dev_tab = NULL;
 static simple_mtx_t dev_tab_mutex = _SIMPLE_MTX_INITIALIZER_NP;

+#if DEBUG
 DEBUG_GET_ONCE_BOOL_OPTION(all_bos, "RADEON_ALL_BOS", false)
+#endif

 static void handle_env_var_force_family(struct amdgpu_winsys *ws)
 {
@@ -108,7 +110,9 @@ static bool do_winsys_init(struct amdgpu_winsys *ws,

   ws->check_vm = strstr(debug_get_option("R600_DEBUG", ""), "check_vm") != NULL ||
                  strstr(debug_get_option("AMD_DEBUG", ""), "check_vm") != NULL;
+#if DEBUG
   ws->debug_all_bos = debug_get_option_all_bos();
+#endif
   ws->reserve_vmid = strstr(debug_get_option("R600_DEBUG", ""), "reserve_vmid") != NULL ||
                      strstr(debug_get_option("AMD_DEBUG", ""), "reserve_vmid") != NULL;
   ws->zero_all_vram_allocs = strstr(debug_get_option("R600_DEBUG", ""), "zerovram") != NULL ||
@@ -138,7 +142,9 @@ static void do_winsys_deinit(struct amdgpu_winsys *ws)
   pb_cache_deinit(&ws->bo_cache);
   _mesa_hash_table_destroy(ws->bo_export_table, NULL);
   simple_mtx_destroy(&ws->sws_list_lock);
+#if DEBUG
   simple_mtx_destroy(&ws->global_bo_list_lock);
+#endif
   simple_mtx_destroy(&ws->bo_export_table_lock);

   ac_addrlib_destroy(ws->addrlib);
@@ -468,12 +474,15 @@ amdgpu_winsys_create(int fd, const struct pipe_screen_config *config,

      /* init reference */
      pipe_reference_init(&aws->reference, 1);
-
+#if DEBUG
      list_inithead(&aws->global_bo_list);
+#endif
      aws->bo_export_table = util_hash_table_create_ptr_keys();

      (void) simple_mtx_init(&aws->sws_list_lock, mtx_plain);
+#if DEBUG
      (void) simple_mtx_init(&aws->global_bo_list_lock, mtx_plain);
+#endif
      (void) simple_mtx_init(&aws->bo_fence_lock, mtx_plain);
      (void) simple_mtx_init(&aws->bo_export_table_lock, mtx_plain);

@@ -82,14 +82,16 @@ struct amdgpu_winsys {
   struct ac_addrlib *addrlib;

   bool check_vm;
-   bool debug_all_bos;
   bool reserve_vmid;
   bool zero_all_vram_allocs;
+#if DEBUG
+   bool debug_all_bos;

   /* List of all allocated buffers */
   simple_mtx_t global_bo_list_lock;
   struct list_head global_bo_list;
   unsigned num_buffers;
+#endif

   /* Single-linked list of all structs amdgpu_screen_winsys referencing this
    * struct amdgpu_winsys