Index: src/runtime/runtime-numbers.cc |
diff --git a/src/runtime/runtime-numbers.cc b/src/runtime/runtime-numbers.cc |
new file mode 100644 |
index 0000000000000000000000000000000000000000..3286aa6f131a5047474093cf8d62d1d3e0012848 |
--- /dev/null |
+++ b/src/runtime/runtime-numbers.cc |
@@ -0,0 +1,565 @@ |
+// Copyright 2014 the V8 project authors. All rights reserved. |
+// Use of this source code is governed by a BSD-style license that can be |
+// found in the LICENSE file. |
+ |
+#include "src/v8.h" |
+ |
+#include "src/arguments.h" |
+#include "src/codegen.h" |
+#include "src/misc-intrinsics.h" |
+#include "src/runtime/runtime.h" |
+#include "src/runtime/runtime-utils.h" |
+ |
+ |
+#ifndef _STLP_VENDOR_CSTD |
+// STLPort doesn't import fpclassify and isless into the std namespace. |
+using std::fpclassify; |
+using std::isless; |
+#endif |
+ |
+namespace v8 { |
+namespace internal { |
+ |
+RUNTIME_FUNCTION(Runtime_NumberToRadixString) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ CONVERT_SMI_ARG_CHECKED(radix, 1); |
+ RUNTIME_ASSERT(2 <= radix && radix <= 36); |
+ |
+ // Fast case where the result is a one character string. |
+ if (args[0]->IsSmi()) { |
+ int value = args.smi_at(0); |
+ if (value >= 0 && value < radix) { |
+ // Character array used for conversion. |
+ static const char kCharTable[] = "0123456789abcdefghijklmnopqrstuvwxyz"; |
+ return *isolate->factory()->LookupSingleCharacterStringFromCode( |
+ kCharTable[value]); |
+ } |
+ } |
+ |
+ // Slow case. |
+ CONVERT_DOUBLE_ARG_CHECKED(value, 0); |
+ if (std::isnan(value)) { |
+ return isolate->heap()->nan_string(); |
+ } |
+ if (std::isinf(value)) { |
+ if (value < 0) { |
+ return isolate->heap()->minus_infinity_string(); |
+ } |
+ return isolate->heap()->infinity_string(); |
+ } |
+ char* str = DoubleToRadixCString(value, radix); |
+ Handle<String> result = isolate->factory()->NewStringFromAsciiChecked(str); |
+ DeleteArray(str); |
+ return *result; |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberToFixed) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(value, 0); |
+ CONVERT_DOUBLE_ARG_CHECKED(f_number, 1); |
+ int f = FastD2IChecked(f_number); |
+ // See DoubleToFixedCString for these constants: |
+ RUNTIME_ASSERT(f >= 0 && f <= 20); |
+ RUNTIME_ASSERT(!Double(value).IsSpecial()); |
+ char* str = DoubleToFixedCString(value, f); |
+ Handle<String> result = isolate->factory()->NewStringFromAsciiChecked(str); |
+ DeleteArray(str); |
+ return *result; |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberToExponential) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(value, 0); |
+ CONVERT_DOUBLE_ARG_CHECKED(f_number, 1); |
+ int f = FastD2IChecked(f_number); |
+ RUNTIME_ASSERT(f >= -1 && f <= 20); |
+ RUNTIME_ASSERT(!Double(value).IsSpecial()); |
+ char* str = DoubleToExponentialCString(value, f); |
+ Handle<String> result = isolate->factory()->NewStringFromAsciiChecked(str); |
+ DeleteArray(str); |
+ return *result; |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberToPrecision) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(value, 0); |
+ CONVERT_DOUBLE_ARG_CHECKED(f_number, 1); |
+ int f = FastD2IChecked(f_number); |
+ RUNTIME_ASSERT(f >= 1 && f <= 21); |
+ RUNTIME_ASSERT(!Double(value).IsSpecial()); |
+ char* str = DoubleToPrecisionCString(value, f); |
+ Handle<String> result = isolate->factory()->NewStringFromAsciiChecked(str); |
+ DeleteArray(str); |
+ return *result; |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_IsValidSmi) { |
+ SealHandleScope shs(isolate); |
+ DCHECK(args.length() == 1); |
+ |
+ CONVERT_NUMBER_CHECKED(int32_t, number, Int32, args[0]); |
+ return isolate->heap()->ToBoolean(Smi::IsValid(number)); |
+} |
+ |
+ |
+static bool AreDigits(const uint8_t* s, int from, int to) { |
+ for (int i = from; i < to; i++) { |
+ if (s[i] < '0' || s[i] > '9') return false; |
+ } |
+ |
+ return true; |
+} |
+ |
+ |
+static int ParseDecimalInteger(const uint8_t* s, int from, int to) { |
+ DCHECK(to - from < 10); // Overflow is not possible. |
+ DCHECK(from < to); |
+ int d = s[from] - '0'; |
+ |
+ for (int i = from + 1; i < to; i++) { |
+ d = 10 * d + (s[i] - '0'); |
+ } |
+ |
+ return d; |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_StringToNumber) { |
+ HandleScope handle_scope(isolate); |
+ DCHECK(args.length() == 1); |
+ CONVERT_ARG_HANDLE_CHECKED(String, subject, 0); |
+ subject = String::Flatten(subject); |
+ |
+ // Fast case: short integer or some sorts of junk values. |
+ if (subject->IsSeqOneByteString()) { |
+ int len = subject->length(); |
+ if (len == 0) return Smi::FromInt(0); |
+ |
+ DisallowHeapAllocation no_gc; |
+ uint8_t const* data = Handle<SeqOneByteString>::cast(subject)->GetChars(); |
+ bool minus = (data[0] == '-'); |
+ int start_pos = (minus ? 1 : 0); |
+ |
+ if (start_pos == len) { |
+ return isolate->heap()->nan_value(); |
+ } else if (data[start_pos] > '9') { |
+ // Fast check for a junk value. A valid string may start from a |
+ // whitespace, a sign ('+' or '-'), the decimal point, a decimal digit |
+ // or the 'I' character ('Infinity'). All of that have codes not greater |
+ // than '9' except 'I' and . |
+ if (data[start_pos] != 'I' && data[start_pos] != 0xa0) { |
+ return isolate->heap()->nan_value(); |
+ } |
+ } else if (len - start_pos < 10 && AreDigits(data, start_pos, len)) { |
+ // The maximal/minimal smi has 10 digits. If the string has less digits |
+ // we know it will fit into the smi-data type. |
+ int d = ParseDecimalInteger(data, start_pos, len); |
+ if (minus) { |
+ if (d == 0) return isolate->heap()->minus_zero_value(); |
+ d = -d; |
+ } else if (!subject->HasHashCode() && len <= String::kMaxArrayIndexSize && |
+ (len == 1 || data[0] != '0')) { |
+ // String hash is not calculated yet but all the data are present. |
+ // Update the hash field to speed up sequential convertions. |
+ uint32_t hash = StringHasher::MakeArrayIndexHash(d, len); |
+#ifdef DEBUG |
+ subject->Hash(); // Force hash calculation. |
+ DCHECK_EQ(static_cast<int>(subject->hash_field()), |
+ static_cast<int>(hash)); |
+#endif |
+ subject->set_hash_field(hash); |
+ } |
+ return Smi::FromInt(d); |
+ } |
+ } |
+ |
+ // Slower case. |
+ int flags = ALLOW_HEX; |
+ if (FLAG_harmony_numeric_literals) { |
+ // The current spec draft has not updated "ToNumber Applied to the String |
+ // Type", https://bugs.ecmascript.org/show_bug.cgi?id=1584 |
+ flags |= ALLOW_OCTAL | ALLOW_BINARY; |
+ } |
+ |
+ return *isolate->factory()->NewNumber( |
+ StringToDouble(isolate->unicode_cache(), *subject, flags)); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_StringParseInt) { |
+ HandleScope handle_scope(isolate); |
+ DCHECK(args.length() == 2); |
+ CONVERT_ARG_HANDLE_CHECKED(String, subject, 0); |
+ CONVERT_NUMBER_CHECKED(int, radix, Int32, args[1]); |
+ RUNTIME_ASSERT(radix == 0 || (2 <= radix && radix <= 36)); |
+ |
+ subject = String::Flatten(subject); |
+ double value; |
+ |
+ { |
+ DisallowHeapAllocation no_gc; |
+ String::FlatContent flat = subject->GetFlatContent(); |
+ |
+ // ECMA-262 section 15.1.2.3, empty string is NaN |
+ if (flat.IsOneByte()) { |
+ value = |
+ StringToInt(isolate->unicode_cache(), flat.ToOneByteVector(), radix); |
+ } else { |
+ value = StringToInt(isolate->unicode_cache(), flat.ToUC16Vector(), radix); |
+ } |
+ } |
+ |
+ return *isolate->factory()->NewNumber(value); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_StringParseFloat) { |
+ HandleScope shs(isolate); |
+ DCHECK(args.length() == 1); |
+ CONVERT_ARG_HANDLE_CHECKED(String, subject, 0); |
+ |
+ subject = String::Flatten(subject); |
+ double value = StringToDouble(isolate->unicode_cache(), *subject, |
+ ALLOW_TRAILING_JUNK, base::OS::nan_value()); |
+ |
+ return *isolate->factory()->NewNumber(value); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberToStringRT) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 1); |
+ CONVERT_NUMBER_ARG_HANDLE_CHECKED(number, 0); |
+ |
+ return *isolate->factory()->NumberToString(number); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberToStringSkipCache) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 1); |
+ CONVERT_NUMBER_ARG_HANDLE_CHECKED(number, 0); |
+ |
+ return *isolate->factory()->NumberToString(number, false); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberToInteger) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 1); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(number, 0); |
+ return *isolate->factory()->NewNumber(DoubleToInteger(number)); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberToIntegerMapMinusZero) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 1); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(number, 0); |
+ double double_value = DoubleToInteger(number); |
+ // Map both -0 and +0 to +0. |
+ if (double_value == 0) double_value = 0; |
+ |
+ return *isolate->factory()->NewNumber(double_value); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberToJSUint32) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 1); |
+ |
+ CONVERT_NUMBER_CHECKED(int32_t, number, Uint32, args[0]); |
+ return *isolate->factory()->NewNumberFromUint(number); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberToJSInt32) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 1); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(number, 0); |
+ return *isolate->factory()->NewNumberFromInt(DoubleToInt32(number)); |
+} |
+ |
+ |
+// Converts a Number to a Smi, if possible. Returns NaN if the number is not |
+// a small integer. |
+RUNTIME_FUNCTION(Runtime_NumberToSmi) { |
+ SealHandleScope shs(isolate); |
+ DCHECK(args.length() == 1); |
+ CONVERT_ARG_CHECKED(Object, obj, 0); |
+ if (obj->IsSmi()) { |
+ return obj; |
+ } |
+ if (obj->IsHeapNumber()) { |
+ double value = HeapNumber::cast(obj)->value(); |
+ int int_value = FastD2I(value); |
+ if (value == FastI2D(int_value) && Smi::IsValid(int_value)) { |
+ return Smi::FromInt(int_value); |
+ } |
+ } |
+ return isolate->heap()->nan_value(); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberAdd) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(x, 0); |
+ CONVERT_DOUBLE_ARG_CHECKED(y, 1); |
+ return *isolate->factory()->NewNumber(x + y); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberSub) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(x, 0); |
+ CONVERT_DOUBLE_ARG_CHECKED(y, 1); |
+ return *isolate->factory()->NewNumber(x - y); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberMul) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(x, 0); |
+ CONVERT_DOUBLE_ARG_CHECKED(y, 1); |
+ return *isolate->factory()->NewNumber(x * y); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberUnaryMinus) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 1); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(x, 0); |
+ return *isolate->factory()->NewNumber(-x); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberDiv) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(x, 0); |
+ CONVERT_DOUBLE_ARG_CHECKED(y, 1); |
+ return *isolate->factory()->NewNumber(x / y); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberMod) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(x, 0); |
+ CONVERT_DOUBLE_ARG_CHECKED(y, 1); |
+ return *isolate->factory()->NewNumber(modulo(x, y)); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberImul) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ // We rely on implementation-defined behavior below, but at least not on |
+ // undefined behavior. |
+ CONVERT_NUMBER_CHECKED(uint32_t, x, Int32, args[0]); |
+ CONVERT_NUMBER_CHECKED(uint32_t, y, Int32, args[1]); |
+ int32_t product = static_cast<int32_t>(x * y); |
+ return *isolate->factory()->NewNumberFromInt(product); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberOr) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_NUMBER_CHECKED(int32_t, x, Int32, args[0]); |
+ CONVERT_NUMBER_CHECKED(int32_t, y, Int32, args[1]); |
+ return *isolate->factory()->NewNumberFromInt(x | y); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberAnd) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_NUMBER_CHECKED(int32_t, x, Int32, args[0]); |
+ CONVERT_NUMBER_CHECKED(int32_t, y, Int32, args[1]); |
+ return *isolate->factory()->NewNumberFromInt(x & y); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberXor) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_NUMBER_CHECKED(int32_t, x, Int32, args[0]); |
+ CONVERT_NUMBER_CHECKED(int32_t, y, Int32, args[1]); |
+ return *isolate->factory()->NewNumberFromInt(x ^ y); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberShl) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_NUMBER_CHECKED(int32_t, x, Int32, args[0]); |
+ CONVERT_NUMBER_CHECKED(int32_t, y, Int32, args[1]); |
+ return *isolate->factory()->NewNumberFromInt(x << (y & 0x1f)); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberShr) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_NUMBER_CHECKED(uint32_t, x, Uint32, args[0]); |
+ CONVERT_NUMBER_CHECKED(int32_t, y, Int32, args[1]); |
+ return *isolate->factory()->NewNumberFromUint(x >> (y & 0x1f)); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberSar) { |
+ HandleScope scope(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_NUMBER_CHECKED(int32_t, x, Int32, args[0]); |
+ CONVERT_NUMBER_CHECKED(int32_t, y, Int32, args[1]); |
+ return *isolate->factory()->NewNumberFromInt( |
+ ArithmeticShiftRight(x, y & 0x1f)); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberEquals) { |
+ SealHandleScope shs(isolate); |
+ DCHECK(args.length() == 2); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(x, 0); |
+ CONVERT_DOUBLE_ARG_CHECKED(y, 1); |
+ if (std::isnan(x)) return Smi::FromInt(NOT_EQUAL); |
+ if (std::isnan(y)) return Smi::FromInt(NOT_EQUAL); |
+ if (x == y) return Smi::FromInt(EQUAL); |
+ Object* result; |
+ if ((fpclassify(x) == FP_ZERO) && (fpclassify(y) == FP_ZERO)) { |
+ result = Smi::FromInt(EQUAL); |
+ } else { |
+ result = Smi::FromInt(NOT_EQUAL); |
+ } |
+ return result; |
+} |
+ |
+ |
+RUNTIME_FUNCTION(Runtime_NumberCompare) { |
+ SealHandleScope shs(isolate); |
+ DCHECK(args.length() == 3); |
+ |
+ CONVERT_DOUBLE_ARG_CHECKED(x, 0); |
+ CONVERT_DOUBLE_ARG_CHECKED(y, 1); |
+ CONVERT_ARG_HANDLE_CHECKED(Object, uncomparable_result, 2) |
+ if (std::isnan(x) || std::isnan(y)) return *uncomparable_result; |
+ if (x == y) return Smi::FromInt(EQUAL); |
+ if (isless(x, y)) return Smi::FromInt(LESS); |
+ return Smi::FromInt(GREATER); |
+} |
+ |
+ |
+// Compare two Smis as if they were converted to strings and then |
+// compared lexicographically. |
+RUNTIME_FUNCTION(Runtime_SmiLexicographicCompare) { |
+ SealHandleScope shs(isolate); |
+ DCHECK(args.length() == 2); |
+ CONVERT_SMI_ARG_CHECKED(x_value, 0); |
+ CONVERT_SMI_ARG_CHECKED(y_value, 1); |
+ |
+ // If the integers are equal so are the string representations. |
+ if (x_value == y_value) return Smi::FromInt(EQUAL); |
+ |
+ // If one of the integers is zero the normal integer order is the |
+ // same as the lexicographic order of the string representations. |
+ if (x_value == 0 || y_value == 0) |
+ return Smi::FromInt(x_value < y_value ? LESS : GREATER); |
+ |
+ // If only one of the integers is negative the negative number is |
+ // smallest because the char code of '-' is less than the char code |
+ // of any digit. Otherwise, we make both values positive. |
+ |
+ // Use unsigned values otherwise the logic is incorrect for -MIN_INT on |
+ // architectures using 32-bit Smis. |
+ uint32_t x_scaled = x_value; |
+ uint32_t y_scaled = y_value; |
+ if (x_value < 0 || y_value < 0) { |
+ if (y_value >= 0) return Smi::FromInt(LESS); |
+ if (x_value >= 0) return Smi::FromInt(GREATER); |
+ x_scaled = -x_value; |
+ y_scaled = -y_value; |
+ } |
+ |
+ static const uint32_t kPowersOf10[] = { |
+ 1, 10, 100, 1000, |
+ 10 * 1000, 100 * 1000, 1000 * 1000, 10 * 1000 * 1000, |
+ 100 * 1000 * 1000, 1000 * 1000 * 1000}; |
+ |
+ // If the integers have the same number of decimal digits they can be |
+ // compared directly as the numeric order is the same as the |
+ // lexicographic order. If one integer has fewer digits, it is scaled |
+ // by some power of 10 to have the same number of digits as the longer |
+ // integer. If the scaled integers are equal it means the shorter |
+ // integer comes first in the lexicographic order. |
+ |
+ // From http://graphics.stanford.edu/~seander/bithacks.html#IntegerLog10 |
+ int x_log2 = IntegerLog2(x_scaled); |
+ int x_log10 = ((x_log2 + 1) * 1233) >> 12; |
+ x_log10 -= x_scaled < kPowersOf10[x_log10]; |
+ |
+ int y_log2 = IntegerLog2(y_scaled); |
+ int y_log10 = ((y_log2 + 1) * 1233) >> 12; |
+ y_log10 -= y_scaled < kPowersOf10[y_log10]; |
+ |
+ int tie = EQUAL; |
+ |
+ if (x_log10 < y_log10) { |
+ // X has fewer digits. We would like to simply scale up X but that |
+ // might overflow, e.g when comparing 9 with 1_000_000_000, 9 would |
+ // be scaled up to 9_000_000_000. So we scale up by the next |
+ // smallest power and scale down Y to drop one digit. It is OK to |
+ // drop one digit from the longer integer since the final digit is |
+ // past the length of the shorter integer. |
+ x_scaled *= kPowersOf10[y_log10 - x_log10 - 1]; |
+ y_scaled /= 10; |
+ tie = LESS; |
+ } else if (y_log10 < x_log10) { |
+ y_scaled *= kPowersOf10[x_log10 - y_log10 - 1]; |
+ x_scaled /= 10; |
+ tie = GREATER; |
+ } |
+ |
+ if (x_scaled < y_scaled) return Smi::FromInt(LESS); |
+ if (x_scaled > y_scaled) return Smi::FromInt(GREATER); |
+ return Smi::FromInt(tie); |
+} |
+ |
+ |
+RUNTIME_FUNCTION(RuntimeReference_NumberToString) { |
+ SealHandleScope shs(isolate); |
+ return __RT_impl_Runtime_NumberToStringRT(args, isolate); |
+} |
+} |
+} // namespace v8::internal |