diff --git a/va/va.c b/va/va.c
index 8b0bee6f0..8b1f0f27c 100644
--- a/va/va.c
+++ b/va/va.c
@@ -2247,3 +2247,41 @@ vaQueryVideoProcPipelineCaps(
     VA_TRACE_RET(dpy, status);
     return status;
 }
+
+VAStatus
+vaHWAccelCopy(
+    VADisplay         dpy,
+    VAHWCpObject      *dst,
+    VAHWCpObject      *src,
+    uint32_t          option,
+    uintptr_t*        cphandle
+)
+{
+  VAStatus va_status;
+  VADriverContextP ctx;
+  CHECK_DISPLAY(dpy);
+  ctx = CTX(dpy);
+
+  if(ctx->vtable->vaHWAccelCopy  == NULL)
+      va_status = VA_STATUS_ERROR_UNIMPLEMENTED;
+  else
+      va_status = ctx->vtable->vaHWAccelCopy( ctx, dst, src, option, cphandle);
+  return va_status;
+}
+
+VAStatus
+vaHWAccelCopyWait(
+    VADisplay           dpy,
+    uintptr_t           cphandle
+)
+{
+  VAStatus va_status;
+  VADriverContextP ctx;
+  CHECK_DISPLAY(dpy);
+  ctx = CTX(dpy);
+  if(ctx->vtable->vaHWAccelCopyWait  == NULL)
+      va_status = VA_STATUS_ERROR_UNIMPLEMENTED;
+  else
+      va_status = ctx->vtable->vaHWAccelCopyWait( ctx, cphandle);
+  return va_status;
+}
diff --git a/va/va.h b/va/va.h
index 8e5f72107..2a0c4826b 100755
--- a/va/va.h
+++ b/va/va.h
@@ -4620,6 +4620,36 @@ typedef struct _VAPictureHEVC
  */
 #define VA_PICTURE_HEVC_RPS_LT_CURR             0x00000040
 
+typedef enum{
+    VAHWAccelCopyObjSurface = 0,
+    VAHWAccelCopyObjBuffer  = 1,
+    VAHWAccelCopyObjUserPTR = 2
+}VAHWAccelCopyObjType;
+
+typedef struct _VAHWCpObject{
+    union
+    {
+        VASurfaceID surface_id;
+        VABufferID  Buffer_id;
+        uint8_t *   userptr;       //pointer to system memory
+    }obj;
+    VAHWAccelCopyObjType  obj_type;          //type of object.
+    uint32_t    size;              // for userptr, it should be same with surface/buffer size
+    uint32_t    width_stride;      // for user ptr only,
+    uint32_t    height_stride;     // for user ptr only
+    uint32_t    va_reserved[VA_PADDING_LOW];
+}VAHWCpObject;
+
+#define VA_GPU_ACCEL_COPY_BLOCK    0x00000001
+#define VA_GPU_ACCEL_COPY_NONBLOCK 0x00000002
+
+/** GPU accelarate copy
+ * it can be block copy or non block copy
+ * if it is non block copy, a handle is returned for wait funtion*/
+VAStatus vaHWAccelCopy(VADisplay dpy,VAHWCpObject * dst, VAHWCpObject * src, uint32_t option, uintptr_t* cphandle);
+
+VAStatus vaHWAccelCopyWait(VADisplay dpy, uintptr_t cphandle);
+
 #include <va/va_dec_hevc.h>
 #include <va/va_dec_jpeg.h>
 #include <va/va_dec_vp8.h>
diff --git a/va/va_backend.h b/va/va_backend.h
index f2cd7b0f3..c0a256e0e 100644
--- a/va/va_backend.h
+++ b/va/va_backend.h
@@ -484,9 +484,23 @@ struct VADriverVTable
             uint32_t            flags,          /* in */
             void               *descriptor      /* out */
         );
+        VAStatus
+        (*vaHWAccelCopy)(
+            VADriverContextP    ctx,            /* in */
+            VAHWCpObject        *dst,           /* in */
+            VAHWCpObject        *src,           /* in */
+            uint32_t            option,         /* in */
+            uintptr_t*          cphandle       /* out */
+        );
+
+        VAStatus
+        (*vaHWAccelCopyWait)(
+            VADriverContextP    ctx,            /* in */
+            uintptr_t           cphandle       /* in */
+        );
 
         /** \brief Reserved bytes for future use, must be zero */
-        unsigned long reserved[57];
+        unsigned long reserved[55];
 };
 
 struct VADriverContext