r300g: offload the CS ioctl to another thread

This is a multi-threading optimization which hides the kernel overhead
behind a thread. It improves performance in CPU-limited apps by 2-15%.
Of course you must have at least 2 cores for it to make any difference.

It can be disabled with:

export RADEON_THREAD=0

2 files changed, 13 insertions, 2 deletions

diff --git a/src/gallium/drivers/r300/r300_flush.c b/src/gallium/drivers/r300/r300_flush.c
index bfc15ceae7..c4bb332aec 100644
--- a/src/gallium/drivers/r300/r300_flush.c
+++ b/src/gallium/drivers/r300/r300_flush.c
@@ -94,6 +94,10 @@ static void r300_flush(struct pipe_context* pipe,
             r300->rws->cs_flush(r300->cs);
         }
     }
+
+    if (flags & PIPE_FLUSH_FRAME) {
+        r300->rws->cs_sync_flush(r300->cs);
+    }
 }
 
 void r300_init_flush_functions(struct r300_context* r300)
diff --git a/src/gallium/drivers/r300/r300_winsys.h b/src/gallium/drivers/r300/r300_winsys.h
index bf1dd5c980..6733253ccc 100644
--- a/src/gallium/drivers/r300/r300_winsys.h
+++ b/src/gallium/drivers/r300/r300_winsys.h
@@ -45,8 +45,8 @@ struct r300_winsys_screen;
 struct r300_winsys_cs_handle;   /* for write_reloc etc. */
 
 struct r300_winsys_cs {
-    unsigned cdw;                           /* Number of used dwords. */
-    uint32_t buf[R300_MAX_CMDBUF_DWORDS];   /* The command buffer. */
+    unsigned cdw;  /* Number of used dwords. */
+    uint32_t *buf; /* The command buffer. */
 };
 
 enum r300_value_id {
@@ -269,6 +269,13 @@ struct r300_winsys_screen {
     void (*cs_flush)(struct r300_winsys_cs *cs);
 
     /**
+     * Wait until the last flush is completed.
+     *
+     * \param cs        A command stream.
+     */
+    void (*cs_sync_flush)(struct r300_winsys_cs *cs);
+
+    /**
      * Set a flush callback which is called from winsys when flush is
      * required.
      *