Add uint32 and uint64 support for PopulationCount

PopulationCount is part of the bitwise operations (BitwiseAnd/Or/Xor/etc)
though unlike other bitwise operatios, PopulationCount does not
have uint32 and uint64 support.

This PR add uint32 and uint64 support and fixes 46676.

Signed-off-by: Yong Tang <yong.tang.github@outlook.com>

tensorflow/core/kernels/population_count_op.cc

@@ -65,7 +65,9 @@ TF_CALL_int8(REGISTER_POPULATION_COUNT);
 TF_CALL_uint16(REGISTER_POPULATION_COUNT);
 TF_CALL_int16(REGISTER_POPULATION_COUNT);
 TF_CALL_int32(REGISTER_POPULATION_COUNT);
+TF_CALL_uint32(REGISTER_POPULATION_COUNT);
 TF_CALL_int64(REGISTER_POPULATION_COUNT);
+TF_CALL_uint64(REGISTER_POPULATION_COUNT);
 
 #undef REGISTER_POPULATION_COUNT
 
@@ -87,7 +89,9 @@ POPCNT(uint8, 8);
 POPCNT(int16, 16);
 POPCNT(uint16, 16);
 POPCNT(int32, 32);
+POPCNT(uint32, 32);
 POPCNT(int64, 64);
+POPCNT(uint64, 64);
 
 #undef POPCNT
 

tensorflow/python/ops/bitwise_ops_test.py

@@ -54,7 +54,8 @@ def testBinaryOps(self):
   def testPopulationCountOp(self):
     dtype_list = [dtypes.int8, dtypes.int16,
                   dtypes.int32, dtypes.int64,
-                  dtypes.uint8, dtypes.uint16]
+                  dtypes.uint8, dtypes.uint16,
+                  dtypes.uint32, dtypes.uint64]
     raw_inputs = [0, 1, -1, 3, -3, 5, -5, 14, -14,
                   127, 128, 255, 256, 65535, 65536,
                   2**31 - 1, 2**31, 2**32 - 1, 2**32, -2**32 + 1, -2**32,