fluencelabs/musl
1
0
Fork 0
mirror of https://github.com/fluencelabs/musl synced 2025-06-07 12:01:35 +00:00
Code Issues Projects Releases Wiki Activity
musl/src/internal/floatscan.c

459 lines
8.6 KiB
C
Raw Normal View History

new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
#include <stdint.h>
#include <stdio.h>
#include <math.h>
#include <float.h>
#include <limits.h>
set errno properly when parsing floating point
2012-04-10 22:38:21 -04:00
#include <errno.h>
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
#include "shgetc.h"
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
#include "floatscan.h"
#if LDBL_MANT_DIG == 53 && LDBL_MAX_EXP == 1024
#define LD_B1B_DIG 2
#define LD_B1B_MAX 9007199, 254740991
#define KMAX 128
#else /* LDBL_MANT_DIG == 64 && LDBL_MAX_EXP == 16384 */
#define LD_B1B_DIG 3
#define LD_B1B_MAX 18, 446744073, 709551615
#define KMAX 2048
#endif
#define MASK (KMAX-1)
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
static long long scanexp(FILE *f, int pok)
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
{
int c;
int x;
long long y;
int neg = 0;
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
c = shgetc(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (c=='+' || c=='-') {
neg = (c=='-');
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
c = shgetc(f);
if (c-'0'>=10U && pok) shunget(f);
}
if (c-'0'>=10U) {
shunget(f);
return LLONG_MIN;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
}
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
for (x=0; c-'0'<10U && x<INT_MAX/10; c = shgetc(f))
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
x = 10*x + c-'0';
fix potential overflow in exponent reading note that there's no need for a precise cutoff, because exponents this large will always result in overflow or underflow (it's impossible to read enough digits to compensate for the exponent magnitude; even at a few nanoseconds per digit it would take hundreds of years).
2012-04-10 23:05:16 -04:00
for (y=x; c-'0'<10U && x<LLONG_MAX/100; c = shgetc(f))
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
y = 10*y + c-'0';
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
for (; c-'0'<10U; c = shgetc(f));
shunget(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
return neg ? -y : y;
}
fix bug parsing lone zero followed by junk, and hex float over-reading
2012-04-11 00:18:57 -04:00
static long double decfloat(FILE *f, int c, int bits, int emin, int sign, int pok)
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
{
uint32_t x[KMAX];
static const uint32_t th[] = { LD_B1B_MAX };
int i, j, k, a, z;
long long lrp=-1, dc=0;
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
long long e10=0;
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
int lnz = 0;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
int gotdig = 0;
int rp;
int e2;
long double y;
long double frac=0;
long double bias=0;
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
static const int p10s[] = { 10, 100, 1000, 10000,
100000, 1000000, 10000000, 100000000 };
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
j=0;
k=0;
/* Don't let leading zeros consume buffer space */
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
for (; c=='0'; c = shgetc(f)) gotdig=1;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
x[0] = 0;
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
for (; c-'0'<10U || c=='.'; c = shgetc(f)) {
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (c == '.') {
if (lrp!=-1) break;
lrp = dc;
simplify/debloat radix point alignment code in floatscan now that this is the first operation, it can rely on the circular buffer contents not being wrapped when it begins. we limit the number of digits read slightly in the initial parsing loops too so that this code does not have to consider the case where it might cause the circular buffer to wrap; this is perfectly fine because KMAX is chosen as a power of two for circular-buffer purposes and is much larger than it otherwise needs to be, anyway. these changes should not affect performance at all.
2012-04-11 14:20:45 -04:00
} else if (k < KMAX-2) {
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
dc++;
floatscan: fix incorrect count of leading nonzero digits this off-by-one error was causing values with just one digit past the decimal point to be treated by the integer case. in many cases it would yield the correct result, but if expressions are evaluated in excess precision, double rounding may occur.
2012-04-16 01:53:52 -04:00
if (c!='0') lnz = dc;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (j) x[k] = x[k]*10 + c-'0';
else x[k] = c-'0';
if (++j==9) {
k++;
j=0;
}
gotdig=1;
} else {
dc++;
simplify/debloat radix point alignment code in floatscan now that this is the first operation, it can rely on the circular buffer contents not being wrapped when it begins. we limit the number of digits read slightly in the initial parsing loops too so that this code does not have to consider the case where it might cause the circular buffer to wrap; this is perfectly fine because KMAX is chosen as a power of two for circular-buffer purposes and is much larger than it otherwise needs to be, anyway. these changes should not affect performance at all.
2012-04-11 14:20:45 -04:00
if (c!='0') x[KMAX-3] |= 1;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
}
}
if (lrp==-1) lrp=dc;
if (gotdig && (c|32)=='e') {
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
e10 = scanexp(f, pok);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (e10 == LLONG_MIN) {
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
if (pok) {
shunget(f);
} else {
shlim(f, 0);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
return 0;
}
e10 = 0;
}
lrp += e10;
} else if (c>=0) {
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
shunget(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
}
if (!gotdig) {
set errno properly when parsing floating point
2012-04-10 22:38:21 -04:00
errno = EINVAL;
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
shlim(f, 0);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
return 0;
}
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
/* Handle zero specially to avoid nasty special cases later */
if (!x[0]) return sign * 0.0;
/* Optimize small integers (w/no exponent) and over/under-flow */
if (lrp==dc && dc<10 && (bits>30 || x[0]>>bits==0))
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
return sign * (long double)x[0];
set errno properly when parsing floating point
2012-04-10 22:38:21 -04:00
if (lrp > -emin/2) {
errno = ERANGE;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
return sign * LDBL_MAX * LDBL_MAX;
set errno properly when parsing floating point
2012-04-10 22:38:21 -04:00
}
if (lrp < emin-2*LDBL_MANT_DIG) {
errno = ERANGE;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
return sign * LDBL_MIN * LDBL_MIN;
set errno properly when parsing floating point
2012-04-10 22:38:21 -04:00
}
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
/* Align incomplete final B1B digit */
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (k<KMAX && j) {
for (; j<9; j++) x[k]*=10;
k++;
j=0;
}
a = 0;
z = k;
e2 = 0;
rp = lrp;
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
/* Optimize small to mid-size integers (even in exp. notation) */
if (lnz<9 && lnz<=rp && rp < 18) {
if (rp == 9) return sign * (long double)x[0];
if (rp < 9) return sign * (long double)x[0] / p10s[8-rp];
int bitlim = bits-3*(int)(rp-9);
if (bitlim>30 || x[0]>>bitlim==0)
return sign * (long double)x[0] * p10s[rp-10];
}
/* Align radix point to B1B digit boundary */
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (rp % 9) {
optimize floatscan: avoid excessive upscaling upscaling by even one step too much creates 3-29 extra iterations for the next loop. this is still suboptimal since it always goes by 2^29 rather than using a smaller upscale factor when nearing the target, but performance on common, small-magnitude, few-digit values has already more than doubled with this change. more optimizations on the way...
2012-04-11 14:11:47 -04:00
int rpm9 = rp>=0 ? rp%9 : rp%9+9;
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
int p10 = p10s[8-rpm9];
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
uint32_t carry = 0;
simplify/debloat radix point alignment code in floatscan now that this is the first operation, it can rely on the circular buffer contents not being wrapped when it begins. we limit the number of digits read slightly in the initial parsing loops too so that this code does not have to consider the case where it might cause the circular buffer to wrap; this is perfectly fine because KMAX is chosen as a power of two for circular-buffer purposes and is much larger than it otherwise needs to be, anyway. these changes should not affect performance at all.
2012-04-11 14:20:45 -04:00
for (k=a; k!=z; k++) {
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
uint32_t tmp = x[k] % p10;
x[k] = x[k]/p10 + carry;
carry = 1000000000/p10 * tmp;
if (k==a && !x[k]) {
a = (a+1 & MASK);
rp -= 9;
}
}
simplify/debloat radix point alignment code in floatscan now that this is the first operation, it can rely on the circular buffer contents not being wrapped when it begins. we limit the number of digits read slightly in the initial parsing loops too so that this code does not have to consider the case where it might cause the circular buffer to wrap; this is perfectly fine because KMAX is chosen as a power of two for circular-buffer purposes and is much larger than it otherwise needs to be, anyway. these changes should not affect performance at all.
2012-04-11 14:20:45 -04:00
if (carry) x[z++] = carry;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
rp += 9-rpm9;
}
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
/* Upscale until desired number of bits are left of radix point */
fix stupid typo in floatscan that caused excess rounding of some values
2012-04-11 21:29:12 -04:00
while (rp < 9*LD_B1B_DIG || (rp == 9*LD_B1B_DIG && x[a]<th[0])) {
optimize floatscan: avoid excessive upscaling upscaling by even one step too much creates 3-29 extra iterations for the next loop. this is still suboptimal since it always goes by 2^29 rather than using a smaller upscale factor when nearing the target, but performance on common, small-magnitude, few-digit values has already more than doubled with this change. more optimizations on the way...
2012-04-11 14:11:47 -04:00
uint32_t carry = 0;
e2 -= 29;
for (k=(z-1 & MASK); ; k=(k-1 & MASK)) {
uint64_t tmp = ((uint64_t)x[k] << 29) + carry;
if (tmp > 1000000000) {
carry = tmp / 1000000000;
x[k] = tmp % 1000000000;
} else {
carry = 0;
x[k] = tmp;
}
if (k==(z-1 & MASK) && k!=a && !x[k]) z = k;
if (k==a) break;
}
if (carry) {
rp += 9;
if (a == z) {
z = (z-1 & MASK);
x[z-1 & MASK] |= x[z];
}
a = (a-1 & MASK);
x[a] = carry;
}
}
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
/* Downscale until exactly number of bits are left of radix point */
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
for (;;) {
uint32_t carry = 0;
int sh = 1;
for (i=0; i<LD_B1B_DIG; i++) {
k = (a+i & MASK);
if (k == z || x[k] < th[i]) {
i=LD_B1B_DIG;
break;
}
if (x[a+i & MASK] > th[i]) break;
}
if (i==LD_B1B_DIG && rp==9*LD_B1B_DIG) break;
/* FIXME: find a way to compute optimal sh */
if (rp > 9+9*LD_B1B_DIG) sh = 9;
e2 += sh;
revert invalid optimization in floatscan
2012-04-11 23:08:50 -04:00
for (k=a; k!=z; k=(k+1 & MASK)) {
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
uint32_t tmp = x[k] & (1<<sh)-1;
x[k] = (x[k]>>sh) + carry;
carry = (1000000000>>sh) * tmp;
if (k==a && !x[k]) {
a = (a+1 & MASK);
optimize floatscan downscaler to skip results that won't be needed when upscaling, even the very last digit is needed in cases where the input is exact; no digits can be discarded. but when downscaling, any digits less significant than the mantissa bits are destined for the great bitbucket; the only influence they can have is their presence (being nonzero). thus, we simply throw them away early. the result is nearly a 4x performance improvement for processing huge values. the particular threshold LD_B1B_DIG+3 is not chosen sharply; it's simply a "safe" distance past the significant bits. it would be nice to replace it with a sharp bound, but i suspect performance will be comparable (within a few percent) anyway.
2012-04-11 14:51:08 -04:00
i--;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
rp -= 9;
}
}
revert invalid optimization in floatscan
2012-04-11 23:08:50 -04:00
if (carry) {
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if ((z+1 & MASK) != a) {
x[z] = carry;
z = (z+1 & MASK);
} else x[z-1 & MASK] |= 1;
}
}
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
/* Assemble desired bits into floating point variable */
fix float scanning of certain values ending in zeros for example, "1000000000" was being read as "1" due to this loop exiting early. it's necessary to actually update z and zero the entries so that the subsequent rounding code does not get confused; before i did that, spurious inexact exceptions were being raised.
2012-04-10 23:41:54 -04:00
for (y=i=0; i<LD_B1B_DIG; i++) {
if ((a+i & MASK)==z) x[z=(z+1 & MASK)] = 0;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
y = 1000000000.0L * y + x[a+i & MASK];
fix float scanning of certain values ending in zeros for example, "1000000000" was being read as "1" due to this loop exiting early. it's necessary to actually update z and zero the entries so that the subsequent rounding code does not get confused; before i did that, spurious inexact exceptions were being raised.
2012-04-10 23:41:54 -04:00
}
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
y *= sign;
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
/* Limit precision for denormal results */
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (bits > LDBL_MANT_DIG+e2-emin) {
bits = LDBL_MANT_DIG+e2-emin;
if (bits<0) bits=0;
}
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
/* Calculate bias term to force rounding, move out lower bits */
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (bits < LDBL_MANT_DIG) {
bias = copysignl(scalbn(1, 2*LDBL_MANT_DIG-bits-1), y);
frac = fmodl(y, scalbn(1, LDBL_MANT_DIG-bits));
y -= frac;
y += bias;
}
optimize more integer cases in floatscan; comment the whole procedure
2012-04-12 00:16:01 -04:00
/* Process tail of decimal input so it can affect rounding */
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if ((a+i & MASK) != z) {
uint32_t t = x[a+i & MASK];
if (t < 500000000 && (t || (a+i+1 & MASK) != z))
frac += 0.25*sign;
else if (t > 500000000)
frac += 0.75*sign;
else if (t == 500000000) {
if ((a+i+1 & MASK) == z)
frac += 0.5*sign;
else
frac += 0.75*sign;
}
if (LDBL_MANT_DIG-bits >= 2 && !fmodl(frac, 1))
frac++;
}
y += frac;
y -= bias;
y = scalbnl(y, e2);
set errno properly when parsing floating point
2012-04-10 22:38:21 -04:00
if (!y) errno = ERANGE;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
return y;
}
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
static long double hexfloat(FILE *f, int bits, int emin, int sign, int pok)
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
{
uint32_t x = 0;
long double y = 0;
long double scale = 1;
long double bias = 0;
int gottail = 0, gotrad = 0, gotdig = 0;
long long rp = 0;
long long dc = 0;
long long e2 = 0;
int d;
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
int c;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
c = shgetc(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
/* Skip leading zeros */
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
for (; c=='0'; c = shgetc(f)) gotdig = 1;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (c=='.') {
gotrad = 1;
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
c = shgetc(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
/* Count zeros after the radix point before significand */
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
for (rp=0; c=='0'; c = shgetc(f), rp--) gotdig = 1;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
}
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
for (; c-'0'<10U || (c|32)-'a'<6U || c=='.'; c = shgetc(f)) {
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (c=='.') {
if (gotrad) break;
rp = dc;
gotrad = 1;
} else {
gotdig = 1;
if (c > '9') d = (c|32)+10-'a';
else d = c-'0';
if (dc<8) {
x = x*16 + d;
} else if (dc < LDBL_MANT_DIG/4+1) {
y += d*(scale/=16);
} else if (d && !gottail) {
y += 0.5*scale;
gottail = 1;
}
dc++;
}
}
if (!gotdig) {
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
shunget(f);
if (pok) {
shunget(f);
if (gotrad) shunget(f);
} else {
shlim(f, 0);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
}
return 0;
}
if (!gotrad) rp = dc;
while (dc<8) x *= 16, dc++;
if ((c|32)=='p') {
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
e2 = scanexp(f, pok);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (e2 == LLONG_MIN) {
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
if (pok) {
shunget(f);
} else {
shlim(f, 0);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
return 0;
}
e2 = 0;
}
fix bug parsing lone zero followed by junk, and hex float over-reading
2012-04-11 00:18:57 -04:00
} else {
shunget(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
}
e2 += 4*rp - 32;
if (!x) return sign * 0.0;
set errno properly when parsing floating point
2012-04-10 22:38:21 -04:00
if (e2 > -emin) {
errno = ERANGE;
return sign * LDBL_MAX * LDBL_MAX;
}
if (e2 < emin-2*LDBL_MANT_DIG) {
errno = ERANGE;
return sign * LDBL_MIN * LDBL_MIN;
}
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
while (x < 0x80000000) {
if (y>=0.5) {
x += x + 1;
y += y - 1;
} else {
x += x;
y += y;
}
e2--;
}
if (bits > 32+e2-emin) {
bits = 32+e2-emin;
if (bits<0) bits=0;
}
if (bits < LDBL_MANT_DIG)
bias = copysignl(scalbn(1, 32+LDBL_MANT_DIG-bits-1), sign);
if (bits<32 && y && !(x&1)) x++, y=0;
y = bias + sign*(long double)x + sign*y;
y -= bias;
set errno properly when parsing floating point
2012-04-10 22:38:21 -04:00
if (!y) errno = ERANGE;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
return scalbnl(y, e2);
}
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
long double __floatscan(FILE *f, int c, int prec, int pok)
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
{
int sign = 1;
int i;
int bits;
int emin;
switch (prec) {
case 0:
remove magic numbers from floatscan
2012-04-12 00:25:52 -04:00
bits = FLT_MANT_DIG;
emin = FLT_MIN_EXP-bits;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
break;
case 1:
remove magic numbers from floatscan
2012-04-12 00:25:52 -04:00
bits = DBL_MANT_DIG;
emin = DBL_MIN_EXP-bits;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
break;
case 2:
bits = LDBL_MANT_DIG;
remove magic numbers from floatscan
2012-04-12 00:25:52 -04:00
emin = LDBL_MIN_EXP-bits;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
break;
default:
return 0;
}
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
if (c<0) c = shgetc(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (c=='+' || c=='-') {
sign -= 2*(c=='-');
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
c = shgetc(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
}
for (i=0; i<8 && (c|32)=="infinity"[i]; i++)
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
if (i<7) c = shgetc(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (i==3 || i==8 || (i>3 && pok)) {
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
if (i==3) shunget(f);
if (pok) for (; i>3; i--) shunget(f);
else shlim(f, 0);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
return sign * INFINITY;
}
if (!i) for (i=0; i<3 && (c|32)=="nan"[i]; i++)
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
if (i<3) c = shgetc(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if (i==3) {
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
return NAN;
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
}
if (i) {
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
shunget(f);
set errno properly when parsing floating point
2012-04-10 22:38:21 -04:00
errno = EINVAL;
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
shlim(f, 0);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
return 0;
}
if (c=='0') {
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
c = shgetc(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
if ((c|32) == 'x')
add "scan helper getc" and rework strtod, etc. to use it the immediate benefit is a significant debloating of the float parsing code by moving the responsibility for keeping track of the number of characters read to a different module. by linking shgetc with the stdio buffer logic, counting logic is defered to buffer refill time, keeping the calls to shgetc fast and light. in the future, shgetc will also be useful for integrating the new float code with scanf, which needs to not only count the characters consumed, but also limit the number of characters read based on field width specifiers. shgetc may also become a useful tool for simplifying the integer parsing code.
2012-04-10 21:47:37 -04:00
return hexfloat(f, bits, emin, sign, pok);
fix bug parsing lone zero followed by junk, and hex float over-reading
2012-04-11 00:18:57 -04:00
shunget(f);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
c = '0';
}
fix bug parsing lone zero followed by junk, and hex float over-reading
2012-04-11 00:18:57 -04:00
return decfloat(f, c, bits, emin, sign, pok);
new floating point parser/converter this version is intended to be fully conformant to the ISO C, POSIX, and IEEE standards for conversion of decimal/hex floating point strings to float, double, and long double (ld64 or ld80 only at present) values. in particular, all results are intended to be rounded correctly according to the current rounding mode. further, this implementation aims to set the floating point underflow, overflow, and inexact flags to reflect the conversion performed. a moderate amount of testing has been performed (by nsz and myself) prior to integration of the code in musl, but it still may have bugs. so far, only strto(d|ld|f) use the new code. scanf integration will be done as a separate commit, and i will add implementations of the wide character functions later.
2012-04-10 11:52:55 -04:00
}
Reference in New Issue Copy Permalink