intel: Add support for ARB_sync.

We currently weasel out of supporting the timeout parameter, but otherwise
this extension looks ready, and should make the common case happy.

1 files changed, 132 insertions, 0 deletions

diff --git a/src/mesa/drivers/dri/intel/intel_syncobj.c b/src/mesa/drivers/dri/intel/intel_syncobj.c
new file mode 100644
index 0000000000..d447b6a0fa
--- /dev/null
+++ b/src/mesa/drivers/dri/intel/intel_syncobj.c
@@ -0,0 +1,132 @@
+/*
+ * Copyright © 2008 Intel Corporation
+ *
+ * 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 (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 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.
+ *
+ * Authors:
+ *    Eric Anholt <eric@anholt.net>
+ *
+ */
+
+/** @file intel_syncobj.c
+ *
+ * Support for ARB_sync
+ *
+ * ARB_sync is implemented by flushing the current batchbuffer and keeping a
+ * reference on it.  We can then check for completion or wait for compeltion
+ * using the normal buffer object mechanisms.  This does mean that if an
+ * application is using many sync objects, it will emit small batchbuffers
+ * which may end up being a significant overhead.  In other tests of removing
+ * gratuitous batchbuffer syncs in Mesa, it hasn't appeared to be a significant
+ * performance bottleneck, though.
+ */
+
+#include "main/simple_list.h"
+#include "main/imports.h"
+
+#include "intel_context.h"
+#include "intel_batchbuffer.h"
+#include "intel_reg.h"
+
+static struct gl_sync_object *
+intel_new_sync_object(GLcontext *ctx, GLuint id)
+{
+   struct intel_sync_object *sync;
+
+   sync = _mesa_calloc(sizeof(struct intel_sync_object));
+
+   return &sync->Base;
+}
+
+static void
+intel_delete_sync_object(GLcontext *ctx, struct gl_sync_object *s)
+{
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   drm_intel_bo_unreference(sync->bo);
+   _mesa_free(sync);
+}
+
+static void
+intel_fence_sync(GLcontext *ctx, struct gl_sync_object *s,
+	       GLenum condition, GLbitfield flags)
+{
+   struct intel_context *intel = intel_context(ctx);
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   assert(condition == GL_SYNC_GPU_COMMANDS_COMPLETE);
+   intel_batchbuffer_emit_mi_flush(intel->batch);
+
+   sync->bo = intel->batch->buf;
+   drm_intel_bo_reference(sync->bo);
+
+   intelFlush(ctx);
+}
+
+/* We ignore the user-supplied timeout.  This is weaselly -- we're allowed to
+ * round to an implementation-dependent accuracy, and right now our
+ * implementation "rounds" to the wait-forever value.
+ *
+ * The fix would be a new kernel function to do the GTT transition with a
+ * timeout.
+ */
+static void intel_client_wait_sync(GLcontext *ctx, struct gl_sync_object *s,
+				 GLbitfield flags, GLuint64 timeout)
+{
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   if (sync->bo) {
+      drm_intel_bo_wait_rendering(sync->bo);
+      s->Status = 1;
+      drm_intel_bo_unreference(sync->bo);
+      sync->bo = NULL;
+   }
+}
+
+/* We have nothing to do for WaitSync.  Our GL command stream is sequential,
+ * so given that the sync object has already flushed the batchbuffer,
+ * any batchbuffers coming after this waitsync will naturally not occur until
+ * the previous one is done.
+ */
+static void intel_server_wait_sync(GLcontext *ctx, struct gl_sync_object *s,
+				 GLbitfield flags, GLuint64 timeout)
+{
+}
+
+static void intel_check_sync(GLcontext *ctx, struct gl_sync_object *s)
+{
+   struct intel_sync_object *sync = (struct intel_sync_object *)s;
+
+   if (sync->bo && drm_intel_bo_busy(sync->bo)) {
+      drm_intel_bo_unreference(sync->bo);
+      sync->bo = NULL;
+      s->Status = 1;
+   }
+}
+
+void intel_init_syncobj_functions(struct dd_function_table *functions)
+{
+   functions->NewSyncObject = intel_new_sync_object;
+   functions->DeleteSyncObject = intel_delete_sync_object;
+   functions->FenceSync = intel_fence_sync;
+   functions->CheckSync = intel_check_sync;
+   functions->ClientWaitSync = intel_client_wait_sync;
+   functions->ServerWaitSync = intel_server_wait_sync;
+}