diff --git a/lib/cudnn/CUDNN.jl b/lib/cudnn/CUDNN.jl
index f7f34399a8..3a9bce21c7 100644
--- a/lib/cudnn/CUDNN.jl
+++ b/lib/cudnn/CUDNN.jl
@@ -12,6 +12,7 @@ using ..APIUtils
 using ..CUDA
 using ..CUDA: CUstream, libraryPropertyType
 using ..CUDA: libcudnn, @retry_reclaim, isdebug, @context!
+using BFloat16s
 
 using CEnum: @cenum
 
diff --git a/lib/cudnn/util.jl b/lib/cudnn/util.jl
index 8fa8ff7bcc..6400947b70 100644
--- a/lib/cudnn/util.jl
+++ b/lib/cudnn/util.jl
@@ -2,6 +2,7 @@
 cptr(x,a::DenseCuArray{Float64})=Float64[x]
 cptr(x,a::DenseCuArray{Float32})=Float32[x]
 cptr(x,a::DenseCuArray{Float16})=Float32[x]
+cptr(x,a::DenseCuArray{BFloat16s.BFloat16})=BFloat16s.BFloat16[x]
 
 # Conversion between Julia and CUDNN datatypes
 cudnnDataType(::Type{Float16})=CUDNN_DATA_HALF
@@ -10,6 +11,7 @@ cudnnDataType(::Type{Float64})=CUDNN_DATA_DOUBLE
 cudnnDataType(::Type{Int8}) = CUDNN_DATA_INT8
 cudnnDataType(::Type{UInt8}) = CUDNN_DATA_UINT8
 cudnnDataType(::Type{Int32}) = CUDNN_DATA_INT32
+cudnnDataType(::Type{BFloat16s.BFloat16}) = CUDNN_DATA_BFLOAT16
 # The following are 32-bit elements each composed of 4 8-bit integers, only supported with CUDNN_TENSOR_NCHW_VECT_C
 # CUDNN_DATA_INT8x4,
 # CUDNN_DATA_UINT8x4,
@@ -19,7 +21,9 @@ juliaDataType(a)=(a==CUDNN_DATA_HALF ? Float16 :
                   a==CUDNN_DATA_DOUBLE ? Float64 :
                   a==CUDNN_DATA_INT8 ? Int8 :
                   a==CUDNN_DATA_UINT8 ? UInt8 :
-                  a==CUDNN_DATA_INT32 ? Int32 : error())
+                  a==CUDNN_DATA_INT32 ? Int32 : 
+                  a==CUDNN_DATA_BFLOAT16 ? BFloat16s.BFloat16 :
+                  error())
 
 tuple_strides(A::Tuple) = _strides((1,), A)
 _strides(out::Tuple{Int}, A::Tuple{}) = ()