You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I'm not sure whether anyone is interested in this, but I have successfully ported a number of the OpenLibm math routines to ELKS, an 8086-based Linux kernel and C library, using the ia16-elf-gcc compiler. In this system, int is 16 bits, and long 32 bits. float and double are IEEE 754 32 and 64 bits.
The following patch was required. It may (or may not) be a good idea to consider changing the non-portable unsigned int definitions to something like uint32_t for portability. Not knowing this library well, I am unsure whether unsigned int may need to be 64 bits on some systems, in which case, this would not work. However, it seems the IEEEf2bits structure assumes unsigned int to be 32 bits for its bitfield definitions.
diff --git a/src/fpmath.h b/src/fpmath.h
index 6c37595..5843386 100644
--- a/src/fpmath.h
+++ b/src/fpmath.h
@@ -83,13 +83,13 @@ union IEEEf2bits {
float f;
struct {
#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
- unsigned int man :23;
- unsigned int exp :8;
- unsigned int sign :1;
+ unsigned long man :23;
+ unsigned long exp :8;
+ unsigned long sign :1;
#else /* _BIG_ENDIAN */
- unsigned int sign :1;
- unsigned int exp :8;
- unsigned int man :23;
+ unsigned long sign :1;
+ unsigned long exp :8;
+ unsigned long man :23;
#endif
} bits;
};
@@ -102,19 +102,19 @@ union IEEEd2bits {
struct {
#if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
#if __FLOAT_WORD_ORDER__ == __ORDER_LITTLE_ENDIAN__
- unsigned int manl :32;
+ unsigned long manl :32;
#endif
- unsigned int manh :20;
- unsigned int exp :11;
- unsigned int sign :1;
+ unsigned long manh :20;
+ unsigned long exp :11;
+ unsigned long sign :1;
#if __FLOAT_WORD_ORDER__ == __ORDER_BIG_ENDIAN__
- unsigned int manl :32;
+ unsigned long manl :32;
#endif
#else /* _BIG_ENDIAN */
- unsigned int sign :1;
- unsigned int exp :11;
- unsigned int manh :20;
- unsigned int manl :32;
+ unsigned long sign :1;
+ unsigned long exp :11;
+ unsigned long manh :20;
+ unsigned long manl :32;
#endif
} bits;
};
diff --git a/src/i386_fpmath.h b/src/i386_fpmath.h
index 455631c..8b29e38 100644
--- a/src/i386_fpmath.h
+++ b/src/i386_fpmath.h
@@ -29,16 +29,16 @@
union IEEEl2bits {
long double e;
struct {
- unsigned int manl :32;
- unsigned int manh :32;
- unsigned int exp :15;
- unsigned int sign :1;
- unsigned int junk :16;
+ unsigned long manl :32;
+ unsigned long manh :32;
+ unsigned long exp :15;
+ unsigned long sign :1;
+ unsigned long junk :16;
} bits;
struct {
unsigned long long man :64;
- unsigned int expsign :16;
- unsigned int junk :16;
+ unsigned long expsign :16;
+ unsigned long junk :16;
} xbits;
};
diff --git a/src/math_private.h b/src/math_private.h
index 15a27b2..4f7e1d4 100644
--- a/src/math_private.h
+++ b/src/math_private.h
@@ -162,7 +162,7 @@ typedef union
{
float value;
/* FIXME: Assumes 32 bit int. */
- unsigned int word;
+ unsigned long word;
} ieee_float_shape_type;
/* Get a 32 bit int from a float. */
Thank you!
The text was updated successfully, but these errors were encountered:
Hello all,
I'm not sure whether anyone is interested in this, but I have successfully ported a number of the OpenLibm math routines to ELKS, an 8086-based Linux kernel and C library, using the ia16-elf-gcc compiler. In this system,
int
is 16 bits, andlong
32 bits.float
anddouble
are IEEE 754 32 and 64 bits.The following patch was required. It may (or may not) be a good idea to consider changing the non-portable
unsigned int
definitions to something likeuint32_t
for portability. Not knowing this library well, I am unsure whetherunsigned int
may need to be 64 bits on some systems, in which case, this would not work. However, it seems the IEEEf2bits structure assumes unsigned int to be 32 bits for its bitfield definitions.Thank you!
The text was updated successfully, but these errors were encountered: