mirror of
https://github.com/netwide-assembler/nasm.git
synced 2025-03-13 17:57:12 +08:00
Add ieee_ prefix to filenames with otherwise highly ambiguous names;
esp round() which is a standard C function.
This commit is contained in:
parent
c650b87792
commit
aa19cc2372
52
float.c
52
float.c
@ -25,7 +25,7 @@
|
||||
* => we only have to worry about _one_ bit shift to the left
|
||||
*/
|
||||
|
||||
static int multiply(unsigned short *to, unsigned short *from)
|
||||
static int ieee_multiply(unsigned short *to, unsigned short *from)
|
||||
{
|
||||
unsigned long temp[MANT_WORDS*2];
|
||||
int i, j;
|
||||
@ -56,7 +56,7 @@ static int multiply(unsigned short *to, unsigned short *from)
|
||||
}
|
||||
}
|
||||
|
||||
static void flconvert(char *string, unsigned short *mant, long *exponent,
|
||||
static void ieee_flconvert(char *string, unsigned short *mant, long *exponent,
|
||||
efunc error)
|
||||
{
|
||||
char digits[MANT_DIGITS];
|
||||
@ -166,8 +166,8 @@ static void flconvert(char *string, unsigned short *mant, long *exponent,
|
||||
extratwos = 0;
|
||||
while (tenpwr) {
|
||||
if (tenpwr & 1)
|
||||
twopwr += extratwos + multiply (mant, mult);
|
||||
extratwos = extratwos * 2 + multiply (mult, mult);
|
||||
twopwr += extratwos + ieee_multiply (mant, mult);
|
||||
extratwos = extratwos * 2 + ieee_multiply (mult, mult);
|
||||
tenpwr >>= 1;
|
||||
}
|
||||
|
||||
@ -183,7 +183,7 @@ static void flconvert(char *string, unsigned short *mant, long *exponent,
|
||||
/*
|
||||
* Shift a mantissa to the right by i (i < 16) bits.
|
||||
*/
|
||||
static void shr(unsigned short *mant, int i)
|
||||
static void ieee_shr(unsigned short *mant, int i)
|
||||
{
|
||||
unsigned short n = 0, m;
|
||||
int j;
|
||||
@ -198,7 +198,7 @@ static void shr(unsigned short *mant, int i)
|
||||
/*
|
||||
* Round a mantissa off after i words.
|
||||
*/
|
||||
static int round(unsigned short *mant, int i)
|
||||
static int ieee_round(unsigned short *mant, int i)
|
||||
{
|
||||
if (mant[i] & 0x8000) {
|
||||
do {
|
||||
@ -220,7 +220,7 @@ static int to_double(char *str, long sign, unsigned char *result,
|
||||
|
||||
sign = (sign < 0 ? 0x8000L : 0L);
|
||||
|
||||
flconvert (str, mant, &exponent, error);
|
||||
ieee_flconvert (str, mant, &exponent, error);
|
||||
if (mant[0] & 0x8000) {
|
||||
/*
|
||||
* Non-zero.
|
||||
@ -231,10 +231,10 @@ static int to_double(char *str, long sign, unsigned char *result,
|
||||
* Normalised.
|
||||
*/
|
||||
exponent += 1023;
|
||||
shr(mant, 11);
|
||||
round(mant, 4);
|
||||
ieee_shr(mant, 11);
|
||||
ieee_round(mant, 4);
|
||||
if (mant[0] & 0x20) /* did we scale up by one? */
|
||||
shr(mant, 1), exponent++;
|
||||
ieee_shr(mant, 1), exponent++;
|
||||
mant[0] &= 0xF; /* remove leading one */
|
||||
put(result+6,(exponent << 4) | mant[0] | sign);
|
||||
put(result+4,mant[1]);
|
||||
@ -246,9 +246,9 @@ static int to_double(char *str, long sign, unsigned char *result,
|
||||
*/
|
||||
int shift = -(exponent+1011);
|
||||
int sh = shift % 16, wds = shift / 16;
|
||||
shr(mant, sh);
|
||||
if (round(mant, 4-wds) || (sh>0 && (mant[0]&(0x8000>>(sh-1))))) {
|
||||
shr(mant, 1);
|
||||
ieee_shr(mant, sh);
|
||||
if (ieee_round(mant, 4-wds) || (sh>0 && (mant[0]&(0x8000>>(sh-1))))) {
|
||||
ieee_shr(mant, 1);
|
||||
if (sh==0)
|
||||
mant[0] |= 0x8000;
|
||||
exponent++;
|
||||
@ -281,7 +281,7 @@ static int to_float(char *str, long sign, unsigned char *result,
|
||||
|
||||
sign = (sign < 0 ? 0x8000L : 0L);
|
||||
|
||||
flconvert (str, mant, &exponent, error);
|
||||
ieee_flconvert (str, mant, &exponent, error);
|
||||
if (mant[0] & 0x8000) {
|
||||
/*
|
||||
* Non-zero.
|
||||
@ -292,10 +292,10 @@ static int to_float(char *str, long sign, unsigned char *result,
|
||||
* Normalised.
|
||||
*/
|
||||
exponent += 127;
|
||||
shr(mant, 8);
|
||||
round(mant, 2);
|
||||
ieee_shr(mant, 8);
|
||||
ieee_round(mant, 2);
|
||||
if (mant[0] & 0x100) /* did we scale up by one? */
|
||||
shr(mant, 1), exponent++;
|
||||
ieee_shr(mant, 1), exponent++;
|
||||
mant[0] &= 0x7F; /* remove leading one */
|
||||
put(result+2,(exponent << 7) | mant[0] | sign);
|
||||
put(result+0,mant[1]);
|
||||
@ -305,9 +305,9 @@ static int to_float(char *str, long sign, unsigned char *result,
|
||||
*/
|
||||
int shift = -(exponent+118);
|
||||
int sh = shift % 16, wds = shift / 16;
|
||||
shr(mant, sh);
|
||||
if (round(mant, 2-wds) || (sh>0 && (mant[0]&(0x8000>>(sh-1))))) {
|
||||
shr(mant, 1);
|
||||
ieee_shr(mant, sh);
|
||||
if (ieee_round(mant, 2-wds) || (sh>0 && (mant[0]&(0x8000>>(sh-1))))) {
|
||||
ieee_shr(mant, 1);
|
||||
if (sh==0)
|
||||
mant[0] |= 0x8000;
|
||||
exponent++;
|
||||
@ -335,7 +335,7 @@ static int to_ldoub(char *str, long sign, unsigned char *result,
|
||||
|
||||
sign = (sign < 0 ? 0x8000L : 0L);
|
||||
|
||||
flconvert (str, mant, &exponent, error);
|
||||
ieee_flconvert (str, mant, &exponent, error);
|
||||
if (mant[0] & 0x8000) {
|
||||
/*
|
||||
* Non-zero.
|
||||
@ -346,8 +346,8 @@ static int to_ldoub(char *str, long sign, unsigned char *result,
|
||||
* Normalised.
|
||||
*/
|
||||
exponent += 16383;
|
||||
if (round(mant, 4)) /* did we scale up by one? */
|
||||
shr(mant, 1), mant[0] |= 0x8000, exponent++;
|
||||
if (ieee_round(mant, 4)) /* did we scale up by one? */
|
||||
ieee_shr(mant, 1), mant[0] |= 0x8000, exponent++;
|
||||
put(result+8,exponent | sign);
|
||||
put(result+6,mant[0]);
|
||||
put(result+4,mant[1]);
|
||||
@ -359,9 +359,9 @@ static int to_ldoub(char *str, long sign, unsigned char *result,
|
||||
*/
|
||||
int shift = -(exponent+16383);
|
||||
int sh = shift % 16, wds = shift / 16;
|
||||
shr(mant, sh);
|
||||
if (round(mant, 4-wds) || (sh>0 && (mant[0]&(0x8000>>(sh-1))))) {
|
||||
shr(mant, 1);
|
||||
ieee_shr(mant, sh);
|
||||
if (ieee_round(mant, 4-wds) || (sh>0 && (mant[0]&(0x8000>>(sh-1))))) {
|
||||
ieee_shr(mant, 1);
|
||||
if (sh==0)
|
||||
mant[0] |= 0x8000;
|
||||
exponent++;
|
||||
|
Loading…
x
Reference in New Issue
Block a user