cell: implement fencing for texture buffers

If we delete a texture, we need to keep the underlying tiled data buffer
around until any rendering that references it has completed.
Keep a list of buffers referenced by a rendering batch.  Unref/free them when
the associated batch's fence is executed/signalled.

2 files changed, 36 insertions, 4 deletions

diff --git a/src/gallium/drivers/cell/spu/spu_command.c b/src/gallium/drivers/cell/spu/spu_command.c
index 9c853c0961a..a6ed29ea631 100644
--- a/src/gallium/drivers/cell/spu/spu_command.c
+++ b/src/gallium/drivers/cell/spu/spu_command.c
@@ -76,9 +76,10 @@ static void
 release_buffer(uint buffer)
 {
    /* Evidently, using less than a 16-byte status doesn't work reliably */
-   static const uint status[4] ALIGN16_ATTRIB
-      = {CELL_BUFFER_STATUS_FREE, 0, 0, 0};
-
+   static const vector unsigned int status = {CELL_BUFFER_STATUS_FREE,
+                                              CELL_BUFFER_STATUS_FREE,
+                                              CELL_BUFFER_STATUS_FREE,
+                                              CELL_BUFFER_STATUS_FREE};
    const uint index = 4 * (spu.init.id * CELL_NUM_BUFFERS + buffer);
    uint *dst = spu.init.buffer_status + index;
 
@@ -93,6 +94,29 @@ release_buffer(uint buffer)
 }
 
 
+/**
+ * Write CELL_FENCE_SIGNALLED back to the fence status qword in main memory.
+ * There's a qword of status per SPU.
+ */
+static void
+cmd_fence(struct cell_command_fence *fence_cmd)
+{
+   static const vector unsigned int status = {CELL_FENCE_SIGNALLED,
+                                              CELL_FENCE_SIGNALLED,
+                                              CELL_FENCE_SIGNALLED,
+                                              CELL_FENCE_SIGNALLED};
+   uint *dst = (uint *) fence_cmd->fence;
+   dst += 4 * spu.init.id;  /* main store/memory address, not local store */
+
+   mfc_put((void *) &status,    /* src in local memory */
+           (unsigned int) dst,  /* dst in main memory */
+           sizeof(status),      /* size */
+           TAG_FENCE,           /* tag */
+           0, /* tid */
+           0  /* rid */);
+}
+
+
 static void
 cmd_clear_surface(const struct cell_command_clear_surface *clear)
 {
@@ -637,6 +661,14 @@ cmd_batch(uint opcode)
          cmd_finish();
          pos += 1;
          break;
+      case CELL_CMD_FENCE:
+         {
+            struct cell_command_fence *fence_cmd =
+               (struct cell_command_fence *) &buffer[pos];
+            cmd_fence(fence_cmd);
+            pos += sizeof(*fence_cmd) / 8;
+         }
+         break;
       case CELL_CMD_RELEASE_VERTS:
          {
             struct cell_command_release_verts *release
diff --git a/src/gallium/drivers/cell/spu/spu_main.h b/src/gallium/drivers/cell/spu/spu_main.h
index 95ef4c92448..668af10be25 100644
--- a/src/gallium/drivers/cell/spu/spu_main.h
+++ b/src/gallium/drivers/cell/spu/spu_main.h
@@ -210,7 +210,7 @@ extern struct spu_global spu;
 #define TAG_DCACHE1           21
 #define TAG_DCACHE2           22
 #define TAG_DCACHE3           23
-
+#define TAG_FENCE             24
 
 
 static INLINE void