Extend Range analysis to 64-bit range and mint operations

runtime/vm/intermediate_language_test.cc

Index: runtime/vm/intermediate_language_test.cc
diff --git a/runtime/vm/intermediate_language_test.cc b/runtime/vm/intermediate_language_test.cc
index 8400557f03afff8c26a9816fe4b6c83ef683b932..a6ed03fa13424a3c53442c4a0bf414cff65fd5e7 100644
--- a/runtime/vm/intermediate_language_test.cc
+++ b/runtime/vm/intermediate_language_test.cc
@@ -53,7 +53,6 @@ TEST_CASE(RangeTests) {
   Range* range_x = new Range(
       RangeBoundary::FromConstant(-15),
       RangeBoundary::FromConstant(100));
-  EXPECT(negative->IsNegative());
   EXPECT(positive->IsPositive());
   EXPECT(zero->Overlaps(0, 0));
   EXPECT(positive->Overlaps(0, 0));
@@ -119,8 +118,11 @@ TEST_CASE(RangeTests) {
 
 TEST_CASE(RangeTestsInfinity) {
   // +/- inf overflowed.
-  EXPECT(RangeBoundary::NegativeInfinity().Overflowed());
-  EXPECT(RangeBoundary::PositiveInfinity().Overflowed());
+  EXPECT(RangeBoundary::NegativeInfinity().OverflowedSmi());
+  EXPECT(RangeBoundary::PositiveInfinity().OverflowedSmi());
+
+  EXPECT(RangeBoundary::NegativeInfinity().OverflowedMint());
+  EXPECT(RangeBoundary::PositiveInfinity().OverflowedMint());
 
   Range* all = new Range(RangeBoundary::NegativeInfinity(),
                          RangeBoundary::PositiveInfinity());
@@ -130,14 +132,12 @@ TEST_CASE(RangeTestsInfinity) {
   Range* positive = new Range(RangeBoundary::FromConstant(0),
                               RangeBoundary::PositiveInfinity());
   EXPECT(positive->IsPositive());
-  EXPECT(!positive->IsNegative());
   EXPECT(positive->Overlaps(0, 1));
   EXPECT(positive->Overlaps(1, 100));
   EXPECT(positive->Overlaps(-1, 0));
   EXPECT(!positive->Overlaps(-2, -1));
   Range* negative = new Range(RangeBoundary::NegativeInfinity(),
                               RangeBoundary::FromConstant(-1));
-  EXPECT(negative->IsNegative());
   EXPECT(!negative->IsPositive());
   EXPECT(!negative->Overlaps(0, 1));
   EXPECT(!negative->Overlaps(1, 100));
@@ -145,7 +145,6 @@ TEST_CASE(RangeTestsInfinity) {
   EXPECT(negative->Overlaps(-2, -1));
   Range* negpos = new Range(RangeBoundary::NegativeInfinity(),
                             RangeBoundary::FromConstant(0));
-  EXPECT(!negpos->IsNegative());
   EXPECT(!negpos->IsPositive());
 
   Range* a = new Range(RangeBoundary::NegativeInfinity(),
@@ -174,4 +173,445 @@ TEST_CASE(RangeTestsInfinity) {
   EXPECT(unsatisfiable_left->IsUnsatisfiable());
 }
 
+
+TEST_CASE(RangeUtils) {
+  // [-inf, +inf].
+  const Range& range_0 = *(new Range(RangeBoundary::NegativeInfinity(),
+                                     RangeBoundary::PositiveInfinity()));
+  // [-inf, -1].
+  const Range& range_a = *(new Range(RangeBoundary::NegativeInfinity(),
+                                     RangeBoundary::FromConstant(-1)));
+  // [-inf, 0].
+  const Range& range_b = *(new Range(RangeBoundary::NegativeInfinity(),
+                                     RangeBoundary::FromConstant(0)));
+  // [-inf, 1].
+  const Range& range_c = *(new Range(RangeBoundary::NegativeInfinity(),
+                                     RangeBoundary::FromConstant(1)));
+  // [-1, +inf]
+  const Range& range_d = *(new Range(RangeBoundary::FromConstant(-1),
+                                     RangeBoundary::PositiveInfinity()));
+  // [0, +inf]
+  const Range& range_e = *(new Range(RangeBoundary::FromConstant(0),
+                                     RangeBoundary::PositiveInfinity()));
+  // [1, +inf].
+  const Range& range_f = *(new Range(RangeBoundary::FromConstant(1),
+                                     RangeBoundary::PositiveInfinity()));
+  // [1, 2].
+  const Range& range_g = *(new Range(RangeBoundary::FromConstant(1),
+                                     RangeBoundary::FromConstant(2)));
+  // [-1, -2].
+  const Range& range_h = *(new Range(RangeBoundary::FromConstant(-1),
+                                     RangeBoundary::FromConstant(-2)));
+  // [-1, 1].
+  const Range& range_i = *(new Range(RangeBoundary::FromConstant(-1),
+                                     RangeBoundary::FromConstant(1)));
+
+  // OnlyPositiveOrZero.
+  EXPECT(!Range::OnlyPositiveOrZero(range_a, range_b));
+  EXPECT(!Range::OnlyPositiveOrZero(range_b, range_c));
+  EXPECT(!Range::OnlyPositiveOrZero(range_c, range_d));
+  EXPECT(!Range::OnlyPositiveOrZero(range_d, range_e));
+  EXPECT(Range::OnlyPositiveOrZero(range_e, range_f));
+  EXPECT(!Range::OnlyPositiveOrZero(range_d, range_d));
+  EXPECT(Range::OnlyPositiveOrZero(range_e, range_e));
+  EXPECT(Range::OnlyPositiveOrZero(range_f, range_g));
+  EXPECT(!Range::OnlyPositiveOrZero(range_g, range_h));
+  EXPECT(!Range::OnlyPositiveOrZero(range_i, range_i));
+
+  // OnlyNegativeOrZero.
+  EXPECT(Range::OnlyNegativeOrZero(range_a, range_b));
+  EXPECT(!Range::OnlyNegativeOrZero(range_b, range_c));
+  EXPECT(Range::OnlyNegativeOrZero(range_b, range_b));
+  EXPECT(!Range::OnlyNegativeOrZero(range_c, range_c));
+  EXPECT(!Range::OnlyNegativeOrZero(range_c, range_d));
+  EXPECT(!Range::OnlyNegativeOrZero(range_d, range_e));
+  EXPECT(!Range::OnlyNegativeOrZero(range_e, range_f));
+  EXPECT(!Range::OnlyNegativeOrZero(range_f, range_g));
+  EXPECT(!Range::OnlyNegativeOrZero(range_g, range_h));
+  EXPECT(Range::OnlyNegativeOrZero(range_h, range_h));
+  EXPECT(!Range::OnlyNegativeOrZero(range_i, range_i));
+
+  // [-inf, +inf].
+  EXPECT(!Range::OnlyNegativeOrZero(range_0, range_0));
+  EXPECT(!Range::OnlyPositiveOrZero(range_0, range_0));
+
+  EXPECT(Range::ConstantAbsMax(&range_0) == RangeBoundary::kMax);
+  EXPECT(Range::ConstantAbsMax(&range_h) == 2);
+  EXPECT(Range::ConstantAbsMax(&range_i) == 1);
+
+  // RangeBOundary.Equals.
+  EXPECT(RangeBoundary::FromConstant(1).Equals(
+      RangeBoundary::FromConstant(1)));
+  EXPECT(!RangeBoundary::FromConstant(2).Equals(
+      RangeBoundary::FromConstant(1)));
+  EXPECT(RangeBoundary::PositiveInfinity().Equals(
+      RangeBoundary::PositiveInfinity()));
+  EXPECT(!RangeBoundary::PositiveInfinity().Equals(
+      RangeBoundary::NegativeInfinity()));
+  EXPECT(RangeBoundary::NegativeInfinity().Equals(
+      RangeBoundary::NegativeInfinity()));
+  EXPECT(!RangeBoundary::NegativeInfinity().Equals(
+      RangeBoundary::PositiveInfinity()));
+  EXPECT(!RangeBoundary::FromConstant(1).Equals(
+      RangeBoundary::NegativeInfinity()));
+  EXPECT(!RangeBoundary::FromConstant(1).Equals(
+      RangeBoundary::NegativeInfinity()));
+  EXPECT(!RangeBoundary::FromConstant(2).Equals(
+      RangeBoundary::PositiveInfinity()));
+}
+
+
+TEST_CASE(RangeBinaryOp) {
+  // Test that [0, +inf] + [-inf, 0] = [-inf, +inf].
+  Range* range_a = new Range(RangeBoundary::FromConstant(0),
+                             RangeBoundary::PositiveInfinity());
+  EXPECT(range_a->min().ConstantValue() == 0);
+  EXPECT(range_a->max().IsPositiveInfinity());
+  Range* range_b = new Range(RangeBoundary::NegativeInfinity(),
+                             RangeBoundary::FromConstant(0));
+  EXPECT(range_b->min().IsNegativeInfinity());

[Florian Schneider, 2014/06/17 10:50:11]

Where does it make sense to have +/- infinity as an input range?

If an operation has an infinity as a result range, it signals that it may
overflow. But I think the range will always be clamped before it's  used as
input in another operation.

How about asserting that input ranges are always finite?

[Cutch, 2014/06/17 21:49:04]

Agreed. I've modified Range::BinaryOp to assert that both the left and right
ranges != +/- infinity.

I've also modified BinarySmiOpInstr and BinaryMintOpInstr to clamp the resulting
ranges. In the case of BinarySmiOpInstr I first determine if an overflow has
occurred before clamping the range.

+  EXPECT(range_b->max().ConstantValue() == 0);
+  Range* result = Range::BinaryOp(Token::kADD,
+                                  range_a,
+                                  range_b,
+                                  NULL);
+  ASSERT(result != NULL);
+  EXPECT(result->min().IsNegativeInfinity());
+  EXPECT(result->max().IsPositiveInfinity());
+
+  // Test that [5, 10] + [0, 5] = [5, 15].
+  Range* range_c = new Range(RangeBoundary::FromConstant(5),
+                             RangeBoundary::FromConstant(10));
+  Range* range_d = new Range(RangeBoundary::FromConstant(0),
+                             RangeBoundary::FromConstant(5));
+  result = Range::BinaryOp(Token::kADD,
+                           range_c,
+                           range_d,
+                           NULL);
+  ASSERT(result != NULL);
+  EXPECT(result->min().ConstantValue() == 5);
+  EXPECT(result->max().ConstantValue() == 15);
+
+
+  // Test that [0xff, 0xfff] & [0xf, 0xf] = [0x0, 0xf].
+  Range* range_e = new Range(RangeBoundary::FromConstant(0xff),
+                             RangeBoundary::FromConstant(0xfff));
+  Range* range_f = new Range(RangeBoundary::FromConstant(0xf),
+                             RangeBoundary::FromConstant(0xf));
+  result = Range::BinaryOp(Token::kBIT_AND,
+                           range_e,
+                           range_f,
+                           NULL);
+  ASSERT(result != NULL);
+  EXPECT(result->min().ConstantValue() == 0x0);
+  EXPECT(result->max().ConstantValue() == 0xf);
+}
+
+
+TEST_CASE(RangeAdd) {
+#define TEST_RANGE_ADD(l_min, l_max, r_min, r_max, result_min, result_max)     \

[Florian Schneider, 2014/06/17 10:50:11]

Can you re-use the TEST_RANGE_OP macro from above here? or something similar?

[Cutch, 2014/06/17 21:49:03]

I can't reuse for add and subtract because they take an extra argument
(Definition* left_defn).

+  {                                                                            \
+    RangeBoundary min, max;                                                    \
+    Range* left_range = new Range(                                             \
+      RangeBoundary::FromConstant(l_min),                                      \
+      RangeBoundary::FromConstant(l_max));                                     \
+    Range* right_range = new Range(                                            \
+      RangeBoundary::FromConstant(r_min),                                      \
+      RangeBoundary::FromConstant(r_max));                                     \
+    EXPECT(left_range->min().ConstantValue() == l_min);                        \
+    EXPECT(left_range->max().ConstantValue() == l_max);                        \
+    EXPECT(right_range->min().ConstantValue() == r_min);                       \
+    EXPECT(right_range->max().ConstantValue() == r_max);                       \
+    Range::Add(left_range, right_range, &min, &max, NULL);                     \
+    EXPECT(min.Equals(result_min));                                            \
+    if (!min.Equals(result_min)) {                                             \
+      OS::Print("%s != %s\n", min.ToCString(), result_min.ToCString());        \
+    }                                                                          \
+    EXPECT(max.Equals(result_max));                                            \
+    if (!max.Equals(result_max)) {                                             \
+      OS::Print("%s != %s\n", max.ToCString(), result_max.ToCString());        \
+    }                                                                          \
+  }
+
+  // [kMaxInt32, kMaxInt32 + 15] + [10, 20] = [kMaxInt32 + 10, kMaxInt32 + 35].
+  TEST_RANGE_ADD(static_cast<int64_t>(kMaxInt32),
+                 static_cast<int64_t>(kMaxInt32) + 15,
+                 static_cast<int64_t>(10),
+                 static_cast<int64_t>(20),
+                 RangeBoundary(static_cast<int64_t>(kMaxInt32) + 10),
+                 RangeBoundary(static_cast<int64_t>(kMaxInt32) + 35));
+
+  // [kMaxInt32 - 15, kMaxInt32 + 15] + [15, -15] = [kMaxInt32, kMaxInt32].
+  TEST_RANGE_ADD(static_cast<int64_t>(kMaxInt32) - 15,
+                 static_cast<int64_t>(kMaxInt32) + 15,
+                 static_cast<int64_t>(15),
+                 static_cast<int64_t>(-15),
+                 RangeBoundary(static_cast<int64_t>(kMaxInt32)),
+                 RangeBoundary(static_cast<int64_t>(kMaxInt32)));
+
+  // [kMaxInt32, kMaxInt32 + 15] + [10, kMaxInt64] = [kMaxInt32 + 10, +inf].
+  TEST_RANGE_ADD(static_cast<int64_t>(kMaxInt32),
+                 static_cast<int64_t>(kMaxInt32) + 15,
+                 static_cast<int64_t>(10),
+                 static_cast<int64_t>(kMaxInt64),
+                 RangeBoundary(static_cast<int64_t>(kMaxInt32) + 10),
+                 RangeBoundary::PositiveInfinity());
+
+  // [kMaxInt64, kMaxInt32 + 15] + [10, kMaxInt64] = [-inf, kMaxInt32 + 35].

[Florian Schneider, 2014/06/17 10:50:10]

Lower bound should be smaller than upper bound.

Maybe s/kMaxInt64/kMinInt64/ here ?

[Cutch, 2014/06/17 21:49:03]

Done.

+  TEST_RANGE_ADD(static_cast<int64_t>(kMaxInt64),
+                 static_cast<int64_t>(kMaxInt32) + 15,
+                 static_cast<int64_t>(10),
+                 static_cast<int64_t>(20),
+                 RangeBoundary::NegativeInfinity(),
+                 RangeBoundary(static_cast<int64_t>(kMaxInt32) + 35));
+
+  // [0, 0] + [kMinInt64, kMaxInt64] = [kMinInt64, kMaxInt64].
+  TEST_RANGE_ADD(static_cast<int64_t>(0),
+                 static_cast<int64_t>(0),
+                 static_cast<int64_t>(kMinInt64),
+                 static_cast<int64_t>(kMaxInt64),
+                 RangeBoundary(kMinInt64),
+                 RangeBoundary(kMaxInt64));
+
+
+  // [-1, 1] + [kMinInt64, kMaxInt64] = [-inf, +inf].
+  TEST_RANGE_ADD(static_cast<int64_t>(-1),
+                 static_cast<int64_t>(1),
+                 static_cast<int64_t>(kMinInt64),
+                 static_cast<int64_t>(kMaxInt64),
+                 RangeBoundary::NegativeInfinity(),
+                 RangeBoundary::PositiveInfinity());
+
+#undef TEST_RANGE_ADD
+}
+
+
+TEST_CASE(RangeSub) {
+#define TEST_RANGE_SUB(l_min, l_max, r_min, r_max, result_min, result_max)     \
+  {                                                                            \
+    RangeBoundary min, max;                                                    \
+    Range* left_range = new Range(                                             \
+      RangeBoundary::FromConstant(l_min),                                      \
+      RangeBoundary::FromConstant(l_max));                                     \
+    Range* right_range = new Range(                                            \
+      RangeBoundary::FromConstant(r_min),                                      \
+      RangeBoundary::FromConstant(r_max));                                     \
+    EXPECT(left_range->min().ConstantValue() == l_min);                        \
+    EXPECT(left_range->max().ConstantValue() == l_max);                        \
+    EXPECT(right_range->min().ConstantValue() == r_min);                       \
+    EXPECT(right_range->max().ConstantValue() == r_max);                       \
+    Range::Sub(left_range, right_range, &min, &max, NULL);                     \
+    EXPECT(min.Equals(result_min));                                            \
+    if (!min.Equals(result_min)) {                                             \
+      OS::Print("%s != %s\n", min.ToCString(), result_min.ToCString());        \
+    }                                                                          \
+    EXPECT(max.Equals(result_max));                                            \
+    if (!max.Equals(result_max)) {                                             \
+      OS::Print("%s != %s\n", max.ToCString(), result_max.ToCString());        \
+    }                                                                          \
+  }
+
+  // [kMaxInt32, kMaxInt32 + 15] - [10, 20] = [kMaxInt32 - 20, kMaxInt32 + 5].
+  TEST_RANGE_SUB(static_cast<int64_t>(kMaxInt32),

[Florian Schneider, 2014/06/17 10:50:10]

Maybe move the static_cast<int64_t> into the macro for conciseness?

[Cutch, 2014/06/17 21:49:03]

I can't because kMaxInt32 is an int32_t and it must be cast to int64_t before
the addition.

+                 static_cast<int64_t>(kMaxInt32) + 15,
+                 static_cast<int64_t>(10),
+                 static_cast<int64_t>(20),
+                 RangeBoundary(static_cast<int64_t>(kMaxInt32) - 20),
+                 RangeBoundary(static_cast<int64_t>(kMaxInt32) + 5));
+
+  // [kMintInt64, kMintInt64] - [1, 1] = [-inf, +inf].
+  TEST_RANGE_SUB(static_cast<int64_t>(kMinInt64),
+                 static_cast<int64_t>(kMinInt64),
+                 static_cast<int64_t>(1),
+                 static_cast<int64_t>(1),
+                 RangeBoundary::NegativeInfinity(),
+                 RangeBoundary::PositiveInfinity());
+
+  // [1, 1] - [kMintInt64, kMintInt64] = [-inf, +inf].
+  TEST_RANGE_SUB(static_cast<int64_t>(1),
+                 static_cast<int64_t>(1),
+                 static_cast<int64_t>(kMinInt64),
+                 static_cast<int64_t>(kMinInt64),
+                 RangeBoundary::NegativeInfinity(),
+                 RangeBoundary::PositiveInfinity());

[Florian Schneider, 2014/06/17 10:50:10]

I'd add also the test cases that you have for ADD to verify that

[a, b] - [c, d] results the same as [a, b] + [-d, -c]

[Cutch, 2014/06/17 21:49:03]

Done.

+
+
+#undef TEST_RANGE_SUB
+}
+
+
+TEST_CASE(RangeAnd) {
+#define TEST_RANGE_AND(l_min, l_max, r_min, r_max, result_min, result_max)     \
+  {                                                                            \
+    RangeBoundary min, max;                                                    \
+    Range* left_range = new Range(                                             \
+      RangeBoundary::FromConstant(l_min),                                      \
+      RangeBoundary::FromConstant(l_max));                                     \
+    Range* right_range = new Range(                                            \
+      RangeBoundary::FromConstant(r_min),                                      \
+      RangeBoundary::FromConstant(r_max));                                     \
+    EXPECT(left_range->min().ConstantValue() == l_min);                        \
+    EXPECT(left_range->max().ConstantValue() == l_max);                        \
+    EXPECT(right_range->min().ConstantValue() == r_min);                       \
+    EXPECT(right_range->max().ConstantValue() == r_max);                       \
+    Range::And(left_range, right_range, &min, &max);                           \
+    EXPECT(min.Equals(result_min));                                            \
+    if (!min.Equals(result_min)) {                                             \
+      OS::Print("%s != %s\n", min.ToCString(), result_min.ToCString());        \
+    }                                                                          \
+    EXPECT(max.Equals(result_max));                                            \
+    if (!max.Equals(result_max)) {                                             \
+      OS::Print("%s != %s\n", max.ToCString(), result_max.ToCString());        \
+    }                                                                          \
+  }
+
+  // [0xff, 0xfff] & [0xf, 0xf] = [0x0, 0xf].
+  TEST_RANGE_AND(static_cast<int64_t>(0xff),
+                 static_cast<int64_t>(0xfff),
+                 static_cast<int64_t>(0xf),
+                 static_cast<int64_t>(0xf),
+                 RangeBoundary(0),
+                 RangeBoundary(0xf));
+
+  // [0xffffffff, 0xffffffff] & [0xfffffffff, 0xfffffffff] = [0x0, 0xfffffffff].
+  TEST_RANGE_AND(static_cast<int64_t>(0xffffffff),
+                 static_cast<int64_t>(0xffffffff),
+                 static_cast<int64_t>(0xfffffffff),
+                 static_cast<int64_t>(0xfffffffff),
+                 RangeBoundary(0),
+                 RangeBoundary(static_cast<int64_t>(0xfffffffff)));
+
+  // [0xffffffff, 0xffffffff] & [-20, 20] = [0x0, 0xffffffff].
+  TEST_RANGE_AND(static_cast<int64_t>(0xffffffff),
+                 static_cast<int64_t>(0xffffffff),
+                 static_cast<int64_t>(-20),
+                 static_cast<int64_t>(20),
+                 RangeBoundary(0),
+                 RangeBoundary(static_cast<int64_t>(0xffffffff)));
+
+  // [-20, 20] & [0xffffffff, 0xffffffff] = [0x0, 0xffffffff].
+  TEST_RANGE_AND(static_cast<int64_t>(-20),
+                 static_cast<int64_t>(20),
+                 static_cast<int64_t>(0xffffffff),
+                 static_cast<int64_t>(0xffffffff),
+                 RangeBoundary(0),
+                 RangeBoundary(static_cast<int64_t>(0xffffffff)));

[Florian Schneider, 2014/06/17 10:50:10]

Please add a test case where both, left and right range can be negative: e.g.
[-20, 20] & [-20, 20]

[Cutch, 2014/06/17 21:49:03]

Done.

+
+#undef TEST_RANGE_AND
+}
+
+
+TEST_CASE(RangeMinMax) {
+  // Constants.
+  // MIN(0, 1) == 0
+  EXPECT(RangeBoundary::Min(
+      RangeBoundary::FromConstant(0),
+      RangeBoundary::FromConstant(1),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == 0);
+  // MIN(0, -1) == -1
+  EXPECT(RangeBoundary::Min(
+      RangeBoundary::FromConstant(0),
+      RangeBoundary::FromConstant(-1),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == -1);
+
+  // MIN(1, 0) == 0
+  EXPECT(RangeBoundary::Min(
+      RangeBoundary::FromConstant(1),
+      RangeBoundary::FromConstant(0),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == 0);
+  // MIN(-1, 0) == -1
+  EXPECT(RangeBoundary::Min(
+      RangeBoundary::FromConstant(-1),
+      RangeBoundary::FromConstant(0),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == -1);
+
+  // MAX(0, 1) == 1
+  EXPECT(RangeBoundary::Max(
+      RangeBoundary::FromConstant(0),
+      RangeBoundary::FromConstant(1),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == 1);
+
+  // MAX(0, -1) == 0
+  EXPECT(RangeBoundary::Max(
+      RangeBoundary::FromConstant(0),
+      RangeBoundary::FromConstant(-1),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == 0);
+
+  // MAX(1, 0) == 1
+  EXPECT(RangeBoundary::Max(
+      RangeBoundary::FromConstant(1),
+      RangeBoundary::FromConstant(0),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == 1);
+  // MAX(-1, 0) == 0
+  EXPECT(RangeBoundary::Max(
+      RangeBoundary::FromConstant(-1),
+      RangeBoundary::FromConstant(0),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == 0);
+
+  RangeBoundary n_infinity = RangeBoundary::NegativeInfinity();
+  RangeBoundary p_infinity = RangeBoundary::PositiveInfinity();
+
+  // Constants vs. infinity.
+  EXPECT(RangeBoundary::Max(
+      n_infinity,
+      RangeBoundary::FromConstant(-1),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == -1);
+
+  EXPECT(RangeBoundary::Max(
+      RangeBoundary::FromConstant(-1),
+      n_infinity,
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == -1);
+
+  EXPECT(RangeBoundary::Max(
+      RangeBoundary::FromConstant(1),
+      n_infinity,
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == 1);
+
+  EXPECT(RangeBoundary::Max(
+      n_infinity,
+      RangeBoundary::FromConstant(1),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == 1);
+
+  EXPECT(RangeBoundary::Min(
+      p_infinity,
+      RangeBoundary::FromConstant(-1),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == -1);
+
+  EXPECT(RangeBoundary::Min(
+      RangeBoundary::FromConstant(-1),
+      p_infinity,
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == -1);
+
+  EXPECT(RangeBoundary::Min(
+      RangeBoundary::FromConstant(1),
+      p_infinity,
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == 1);
+
+  EXPECT(RangeBoundary::Min(
+      p_infinity,
+      RangeBoundary::FromConstant(1),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == 1);
+
+  // 64-bit values.
+  EXPECT(RangeBoundary::Min(
+      RangeBoundary(static_cast<int64_t>(kMinInt64)),
+      RangeBoundary(static_cast<int64_t>(kMinInt32)),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == kMinInt64);
+
+  EXPECT(RangeBoundary::Max(
+      RangeBoundary(static_cast<int64_t>(kMinInt64)),
+      RangeBoundary(static_cast<int64_t>(kMinInt32)),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == kMinInt32);
+
+  EXPECT(RangeBoundary::Min(
+      RangeBoundary(static_cast<int64_t>(kMaxInt64)),
+      RangeBoundary(static_cast<int64_t>(kMaxInt32)),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == kMaxInt32);
+
+  EXPECT(RangeBoundary::Max(
+      RangeBoundary(static_cast<int64_t>(kMaxInt64)),
+      RangeBoundary(static_cast<int64_t>(kMaxInt32)),
+      RangeBoundary::kRangeBoundaryInt64).ConstantValue() == kMaxInt64);
+}
+
 }  // namespace dart