Extended the range of value inputs for TensorIntDiv to support tensors with more than 4 billion elements.

1 files changed, 42 insertions, 4 deletions

diff --git a/unsupported/test/cxx11_tensor_intdiv.cpp b/unsupported/test/cxx11_tensor_intdiv.cpp
index 134329034..b005238bd 100644
--- a/unsupported/test/cxx11_tensor_intdiv.cpp
+++ b/unsupported/test/cxx11_tensor_intdiv.cpp
@@ -67,10 +67,46 @@ void test_unsigned_64bit()
   }
 }
 
+void test_powers_32bit() {
+  for (int expon = 1; expon < 31; expon++) {
+    int32_t div = (1ull << expon);
+    for (int num_expon = 0; num_expon < 32; num_expon++) {
+      int32_t start_num = (1ull << num_expon) - 100;
+      int32_t end_num = (1ull << num_expon) + 100;
+      if (start_num < 0)
+        start_num = 0;
+      for (int64_t num = start_num; num < end_num; num++) {
+        Eigen::internal::TensorIntDivisor<int32_t> divider =
+          Eigen::internal::TensorIntDivisor<int32_t>(div);
+        int32_t result = num/div;
+        int32_t result_op = divider.divide(num);
+        VERIFY_IS_EQUAL(result_op, result);
+      }
+    }
+  }
+}
 
-void test_specific()
-{
-  // A particular combination that exposed a bug in the past.
+void test_powers_64bit() {
+  for (int expon = 0; expon < 63; expon++) {
+    int64_t div = (1ull << expon);
+    for (int num_expon = 0; num_expon < 63; num_expon++) {
+      int64_t start_num = (1ull << num_expon) - 10;
+      int64_t end_num = (1ull << num_expon) + 10;
+      if (start_num < 0)
+        start_num = 0;
+      for (int64_t num = start_num; num < end_num; num++) {
+        Eigen::internal::TensorIntDivisor<int64_t> divider =
+          Eigen::internal::TensorIntDivisor<int64_t>(div);
+        int64_t result = num/div;
+        int64_t result_op = divider.divide(num);
+        VERIFY_IS_EQUAL(result_op, result);
+      }
+    }
+  }
+}
+
+void test_specific() {
+  // A particular combination that was previously failing
   int64_t div = 209715200;
   int64_t num = 3238002688;
   Eigen::internal::TensorIntDivisor<int64_t> divider =
@@ -86,5 +122,7 @@ void test_cxx11_tensor_intdiv()
   CALL_SUBTEST_2(test_unsigned_32bit());
   CALL_SUBTEST_3(test_signed_64bit());
   CALL_SUBTEST_4(test_unsigned_64bit());
-  CALL_SUBTEST_5(test_specific());
+  CALL_SUBTEST_5(test_powers_32bit());
+  CALL_SUBTEST_6(test_powers_64bit());
+  CALL_SUBTEST_7(test_specific());
 }