Fix HIP_SYNC_NULL_STREAM=0 mode.

- Fix null-stream sync
- hipStreamDestroy of null stream returns hipErrorInvalidResourceHandle
- Update documentation.
- Add tests for null stream sync, hipEventElapsedTime.
- Rename internal enum hipEventStatusRecorded to hipEventStatusComplete
- refactor hipStreamWaitEvent to streamline control-flow

src/hip_stream.cpp

@@ -93,20 +93,17 @@ hipError_t hipStreamWaitEvent(hipStream_t stream, hipEvent_t event, unsigned int
 
     } else if (event->_state != hipEventStatusUnitialized) {
 
-        bool fastWait = false;
-
         if (stream != hipStreamNull) {
+
+            // This will user create_blocking_marker to wait on the specified queue.
             stream->locked_waitEvent(event);
 
-            fastWait = true; // don't use the slow host-side synchronization.
-        }
-
-        if (!fastWait) {
+        } else {
             // TODO-hcc Convert to use create_blocking_marker(...) functionality.
             // Currently we have a super-conservative version of this - block on host, and drain the queue.
             // This should create a barrier packet in the target queue.
+            // TODO-HIP_SYNC_NULL_STREAM
             stream->locked_wait();
-            e = hipSuccess;
         }
     } // else event not recorded, return immediately and don't create marker.
 
@@ -150,6 +147,7 @@ hipError_t hipStreamSynchronize(hipStream_t stream)
         ihipCtx_t *ctx = ihipGetTlsDefaultCtx();
         ctx->locked_syncDefaultStream(true/*waitOnSelf*/, true/*syncToHost*/);
     } else {
+		// note this does not synchornize with the NULL stream:
         stream->locked_wait();
         e = hipSuccess;
     }
@@ -171,20 +169,18 @@ hipError_t hipStreamDestroy(hipStream_t stream)
 
     //--- Drain the stream:
     if (stream == NULL) {
-        ihipCtx_t *ctx = ihipGetTlsDefaultCtx();
-        ctx->locked_syncDefaultStream(true/*waitOnSelf*/, true /*syncToHost*/);
+        e = hipErrorInvalidResourceHandle; // TODO - review - what happens if try to destroy null stream
     } else {
         stream->locked_wait();
-        e = hipSuccess;
-    }
 
-    ihipCtx_t *ctx = stream->getCtx();
+        ihipCtx_t *ctx = stream->getCtx();
 
-    if (ctx) {
-        ctx->locked_removeStream(stream);
-        delete stream;
-    } else {
-        e = hipErrorInvalidResourceHandle;
+        if (ctx) {
+            ctx->locked_removeStream(stream);
+            delete stream;
+        } else {
+            e = hipErrorInvalidResourceHandle;
+        }
     }
 
     return ihipLogStatus(e);