diff --git a/cuda-mode.el b/cuda-mode.el
index 48e1591..0db1590 100644
--- a/cuda-mode.el
+++ b/cuda-mode.el
@@ -100,6 +100,166 @@ contain type identifiers."
 	    ;; MSVC extension.
 	    "__declspec"))
 
+(defconst cuda-builtins
+  '(;; atom
+    "atomicAdd"
+    "atomicAnd"
+    "atomicCAS"
+    "atomicDec"
+    "atomicExch"
+    "atomicInc"
+    "atomicMax"
+    "atomicMin"
+    "atomicOr"
+    "atomicSub"
+    "atomicXor"
+    ;; dev
+    "tex1D"
+    "tex1Dfetch"
+    "tex2D"
+    "__float_as_int"
+    "__int_as_float"
+    "__float2int_rn"
+    "__float2int_rz"
+    "__float2int_ru"
+    "__float2int_rd"
+    "__float2uint_rn"
+    "__float2uint_rz"
+    "__float2uint_ru"
+    "__float2uint_rd"
+    "__int2float_rn"
+    "__int2float_rz"
+    "__int2float_ru"
+    "__int2float_rd"
+    "__uint2float_rn"
+    "__uint2float_rz"
+    "__uint2float_ru"
+    "__uint2float_rd"
+    "__fadd_rz"
+    "__fmul_rz"
+    "__fdividef"
+    "__mul24"
+    "__umul24"
+    "__mulhi"
+    "__umulhi"
+    "__mul64hi"
+    "__umul64hi"
+    "min"
+    "umin"
+    "fminf"
+    "fmin"
+    "max"
+    "umax"
+    "fmaxf"
+    "fmax"
+    "abs"
+    "fabsf"
+    "fabs"
+    "sqrtf"
+    "sqrt"
+    "sinf"
+    "__sinf"
+    "sin"
+    "cosf"
+    "__cosf"
+    "cos"
+    "sincosf"
+    "__sincosf"
+    "expf"
+    "__expf"
+    "exp"
+    "logf"
+    "__logf"
+    "log"
+    ;; runtime
+    "cudaBindTexture"
+    "cudaBindTextureToArray"
+    "cudaChooseDevice"
+    "cudaConfigureCall"
+    "cudaCreateChannelDesc"
+    "cudaD3D10GetDevice"
+    "cudaD3D10MapResources"
+    "cudaD3D10RegisterResource"
+    "cudaD3D10ResourceGetMappedArray"
+    "cudaD3D10ResourceGetMappedPitch"
+    "cudaD3D10ResourceGetMappedPointer"
+    "cudaD3D10ResourceGetMappedSize"
+    "cudaD3D10ResourceGetSurfaceDimensions"
+    "cudaD3D10ResourceSetMapFlags"
+    "cudaD3D10SetDirect3DDevice"
+    "cudaD3D10UnmapResources"
+    "cudaD3D10UnregisterResource"
+    "cudaD3D9GetDevice"
+    "cudaD3D9GetDirect3DDevice"
+    "cudaD3D9MapResources"
+    "cudaD3D9RegisterResource"
+    "cudaD3D9ResourceGetMappedArray"
+    "cudaD3D9ResourceGetMappedPitch"
+    "cudaD3D9ResourceGetMappedPointer"
+    "cudaD3D9ResourceGetMappedSize"
+    "cudaD3D9ResourceGetSurfaceDimensions"
+    "cudaD3D9ResourceSetMapFlags"
+    "cudaD3D9SetDirect3DDevice"
+    "cudaD3D9UnmapResources"
+    "cudaD3D9UnregisterResource"
+    "cudaEventCreate"
+    "cudaEventDestroy"
+    "cudaEventElapsedTime"
+    "cudaEventQuery"
+    "cudaEventRecord"
+    "cudaEventSynchronize"
+    "cudaFree"
+    "cudaFreeArray"
+    "cudaFreeHost "
+    "cudaGetChannelDesc"
+    "cudaGetDevice"
+    "cudaGetDeviceCount"
+    "cudaGetDeviceProperties"
+    "cudaGetErrorString"
+    "cudaGetLastError"
+    "cudaGetSymbolAddress"
+    "cudaGetSymbolSize"
+    "cudaGetTextureAlignmentOffset"
+    "cudaGetTextureReference"
+    "cudaGLMapBufferObject"
+    "cudaGLRegisterBufferObject"
+    "cudaGLSetGLDevice"
+    "cudaGLUnmapBufferObject"
+    "cudaGLUnregisterBufferObject"
+    "cudaLaunch"
+    "cudaMalloc"
+    "cudaMalloc3D"
+    "cudaMalloc3DArray"
+    "cudaMallocArray"
+    "cudaMallocHost"
+    "cudaMallocPitch"
+    "cudaMemcpy"
+    "cudaMemcpy2D"
+    "cudaMemcpy2DArrayToArray"
+    "cudaMemcpy2DFromArray"
+    "cudaMemcpy2DToArray"
+    "cudaMemcpy3D"
+    "cudaMemcpyArrayToArray"
+    "cudaMemcpyFromArray"
+    "cudaMemcpyFromSymbol"
+    "cudaMemcpyToArray"
+    "cudaMemcpyToSymbol"
+    "cudaMemset"
+    "cudaMemset2D"
+    "cudaMemset3D"
+    "cudaSetDevice"
+    "cudaSetupArgument"
+    "cudaStreamCreate"
+    "cudaStreamDestroy"
+    "cudaStreamQuery"
+    "cudaStreamSynchronize"
+    "cudaThreadExit"
+    "cudaThreadSynchronize"
+    "cudaUnbindTexture"
+    ;; other
+    "__syncthreads")
+  "Names of built-in cuda functions.")
+
 (defcustom cuda-font-lock-extra-types nil
   "*List of extra types to recognize in Cuda mode.
 Each list item should be a regexp matching a single identifier."
@@ -117,7 +277,12 @@ Each list item should be a regexp matching a single identifier."
   (c-lang-const c-matchers-3 cuda)
   "Accurate normal highlighting for CUDA mode.")
 
-(defvar cuda-font-lock-keywords cuda-font-lock-keywords-3
+(defvar cuda-font-lock-keywords
+  (append cuda-font-lock-keywords-3
+          (list
+           (list
+            (regexp-opt cuda-builtins 'symbols)
+            (0 'font-lock-builtin-face))))
   "Default expressions to highlight in CUDA mode.")
 
 (defvar cuda-mode-syntax-table nil