Fast algorithm for double->string conversion.

test/cctest/test-double.cc

+// Copyright 2006-2008 the V8 project authors. All rights reserved.
+
+#include <stdlib.h>
+
+#include "v8.h"
+
+#include "platform.h"
+#include "cctest.h"
+#include "diy_fp.h"
+#include "double.h"
+
+using namespace v8::internal;
+
+
+TEST(Uint64Conversions) {
+  // Start by checking the byte-order.
+  uint64_t ordered = V8_2PART_UINT64_C(0x01234567, 89ABCDEF);
+  CHECK_EQ(3512700564088504e-318, Double(ordered).value());
+
+  uint64_t min_double64 = V8_2PART_UINT64_C(0x00000000, 00000001);
+  CHECK_EQ(5e-324, Double(min_double64).value());
+
+  uint64_t max_double64 = V8_2PART_UINT64_C(0x7fefffff, ffffffff);
+  CHECK_EQ(1.7976931348623157e308, Double(max_double64).value());
+}
+
+TEST(AsDiyFp) {
+  uint64_t ordered = V8_2PART_UINT64_C(0x01234567, 89ABCDEF);
+  DiyFp diy_fp = Double(ordered).AsDiyFp();
+  CHECK_EQ(0x12 - 0x3FF - 52, diy_fp.e());
+  // The last 52 bits + 1 hidden bit a position 52.
+  CHECK_EQ(V8_2PART_UINT64_C(0x00134567, 89ABCDEF), diy_fp.f());
+
+  uint64_t min_double64 = V8_2PART_UINT64_C(0x00000000, 00000001);
+  diy_fp = Double(min_double64).AsDiyFp();
+  CHECK_EQ(-0x3FF - 52 + 1, diy_fp.e());
+  // This is a denormal; so no hidden bit.
+  uint64_t correct_f = 1;
+  CHECK_EQ(correct_f, diy_fp.f());
+
+  uint64_t max_double64 = V8_2PART_UINT64_C(0x7fefffff, ffffffff);
+  diy_fp = Double(max_double64).AsDiyFp();
+  CHECK_EQ(0x7FE - 0x3FF - 52, diy_fp.e());
+  correct_f = V8_2PART_UINT64_C(0x001fffff, ffffffff);
+  CHECK_EQ(correct_f, diy_fp.f());
+}
+
+
+TEST(AsNormalizedDiyFp) {
+  uint64_t ordered = V8_2PART_UINT64_C(0x01234567, 89ABCDEF);
+  DiyFp diy_fp = Double(ordered).AsNormalizedDiyFp();
+  CHECK_EQ(0x12 - 0x3FF - 52 - 11, diy_fp.e());
+  uint64_t correct_f = V8_2PART_UINT64_C(0x00134567, 89ABCDEF) << 11;
+  CHECK_EQ(correct_f, diy_fp.f());
+
+  uint64_t min_double64 = V8_2PART_UINT64_C(0x00000000, 00000001);
+  diy_fp = Double(min_double64).AsNormalizedDiyFp();
+  CHECK_EQ(-0x3FF - 52 + 1 - 63, diy_fp.e());
+  // This is a denormal; so no hidden bit.
+  correct_f = V8_2PART_UINT64_C(0x80000000, 00000000);
+  CHECK_EQ(correct_f, diy_fp.f());
+
+  uint64_t max_double64 = V8_2PART_UINT64_C(0x7fefffff, ffffffff);
+  diy_fp = Double(max_double64).AsNormalizedDiyFp();
+  CHECK_EQ(0x7FE - 0x3FF - 52 - 11, diy_fp.e());
+  correct_f = V8_2PART_UINT64_C(0x001fffff, ffffffff) << 11;
+  CHECK_EQ(correct_f, diy_fp.f());
+}
+
+TEST(IsDenormal) {
+  uint64_t min_double64 = V8_2PART_UINT64_C(0x00000000, 00000001);
+  CHECK(Double(min_double64).IsDenormal());
+  uint64_t bits = V8_2PART_UINT64_C(0x000FFFFF, FFFFFFFF);
+  CHECK(Double(bits).IsDenormal());
+  bits = V8_2PART_UINT64_C(0x00100000, 00000000);
+  CHECK(!Double(bits).IsDenormal());
+}
+
+TEST(IsSpecial) {
+  CHECK(Double(V8_INFINITY).IsSpecial());
+  CHECK(Double(-V8_INFINITY).IsSpecial());
+  CHECK(Double(OS::nan_value()).IsSpecial());
+  uint64_t bits = V8_2PART_UINT64_C(0xFFF12345, 00000000);
+  CHECK(Double(bits).IsSpecial());
+  // Denormals are not special:
+  CHECK(!Double(5e-324).IsSpecial());
+  // And some random numbers:
+  CHECK(!Double(0.0).IsSpecial());
+  CHECK(!Double(1.0).IsSpecial());
+  CHECK(!Double(1000000.0).IsSpecial());
+  CHECK(!Double(1e23).IsSpecial());
+  CHECK(!Double(1.7976931348623157e308).IsSpecial());
+}
+
+TEST(NormalizedBoundaries) {
+  DiyFp boundary_plus;
+  DiyFp boundary_minus;
+  DiyFp diy_fp = Double(1.5).AsNormalizedDiyFp();
+  Double(1.5).NormalizedBoundaries(&boundary_minus, &boundary_plus);
+  CHECK_EQ(diy_fp.e(), boundary_minus.e());
+  CHECK_EQ(diy_fp.e(), boundary_plus.e());
+  // 1.5 does not have a significand of the form 2^p (for some p).
+  // Therefore its boundaries are at the same distance.
+  CHECK_EQ(diy_fp.f() - boundary_minus.f(), boundary_plus.f() - diy_fp.f());
+  CHECK_EQ(static_cast<uint64_t>(1) << 10, diy_fp.f() - boundary_minus.f());
+
+  diy_fp = Double(1.0).AsNormalizedDiyFp();
+  Double(1.0).NormalizedBoundaries(&boundary_minus, &boundary_plus);
+  CHECK_EQ(diy_fp.e(), boundary_minus.e());
+  CHECK_EQ(diy_fp.e(), boundary_plus.e());
+  // 1.0 does have a significand of the form 2^p (for some p).
+  // Therefore its lower boundary is twice as close as the upper boundary.
+  CHECK_GT(boundary_plus.f() - diy_fp.f(), diy_fp.f() - boundary_minus.f());
+  CHECK_EQ(static_cast<uint64_t>(1) << 9, diy_fp.f() - boundary_minus.f());
+  CHECK_EQ(static_cast<uint64_t>(1) << 10, boundary_plus.f() - diy_fp.f());
+
+  uint64_t min_double64 = V8_2PART_UINT64_C(0x00000000, 00000001);
+  diy_fp = Double(min_double64).AsNormalizedDiyFp();
+  Double(min_double64).NormalizedBoundaries(&boundary_minus, &boundary_plus);
+  CHECK_EQ(diy_fp.e(), boundary_minus.e());
+  CHECK_EQ(diy_fp.e(), boundary_plus.e());
+  // min-value does not have a significand of the form 2^p (for some p).
+  // Therefore its boundaries are at the same distance.
+  CHECK_EQ(diy_fp.f() - boundary_minus.f(), boundary_plus.f() - diy_fp.f());
+  // Denormals have their boundaries much closer.
+  CHECK_EQ(static_cast<uint64_t>(1) << 62, diy_fp.f() - boundary_minus.f());
+
+  uint64_t smallest_normal64 = V8_2PART_UINT64_C(0x00100000, 00000000);
+  diy_fp = Double(smallest_normal64).AsNormalizedDiyFp();
+  Double(smallest_normal64).NormalizedBoundaries(&boundary_minus,
+                                                 &boundary_plus);
+  CHECK_EQ(diy_fp.e(), boundary_minus.e());
+  CHECK_EQ(diy_fp.e(), boundary_plus.e());
+  // Even though the significand is of the form 2^p (for some p), its boundaries
+  // are at the same distance. (This is the only exception).
+  CHECK_EQ(diy_fp.f() - boundary_minus.f(), boundary_plus.f() - diy_fp.f());
+  CHECK_EQ(static_cast<uint64_t>(1) << 10, diy_fp.f() - boundary_minus.f());
+
+  uint64_t largest_denormal64 = V8_2PART_UINT64_C(0x000FFFFF, FFFFFFFF);
+  diy_fp = Double(largest_denormal64).AsNormalizedDiyFp();
+  Double(largest_denormal64).NormalizedBoundaries(&boundary_minus,
+                                                  &boundary_plus);
+  CHECK_EQ(diy_fp.e(), boundary_minus.e());
+  CHECK_EQ(diy_fp.e(), boundary_plus.e());
+  CHECK_EQ(diy_fp.f() - boundary_minus.f(), boundary_plus.f() - diy_fp.f());
+  CHECK_EQ(static_cast<uint64_t>(1) << 11, diy_fp.f() - boundary_minus.f());
+
+  uint64_t max_double64 = V8_2PART_UINT64_C(0x7fefffff, ffffffff);
+  diy_fp = Double(max_double64).AsNormalizedDiyFp();
+  Double(max_double64).NormalizedBoundaries(&boundary_minus, &boundary_plus);
+  CHECK_EQ(diy_fp.e(), boundary_minus.e());
+  CHECK_EQ(diy_fp.e(), boundary_plus.e());
+  // max-value does not have a significand of the form 2^p (for some p).
+  // Therefore its boundaries are at the same distance.
+  CHECK_EQ(diy_fp.f() - boundary_minus.f(), boundary_plus.f() - diy_fp.f());
+  CHECK_EQ(static_cast<uint64_t>(1) << 10, diy_fp.f() - boundary_minus.f());
+}