diff options
Diffstat (limited to 'libyasm/floatnum.c')
| -rw-r--r-- | libyasm/floatnum.c | 16 |
1 files changed, 15 insertions, 1 deletions
diff --git a/libyasm/floatnum.c b/libyasm/floatnum.c index 951e17ca..a7cfcb7e 100644 --- a/libyasm/floatnum.c +++ b/libyasm/floatnum.c @@ -631,7 +631,17 @@ floatnum_get_common(const yasm_floatnum *flt, /*@out@*/ unsigned char *ptr, return retval; } -/* IEEE-754 (Intel) "single precision" format: +/* IEEE-754r "half precision" format: + * 16 bits: + * 15 9 Bit 0 + * | | | + * seee eemm mmmm mmmm + * + * e = bias 15 exponent + * s = sign bit + * m = mantissa bits, bit 10 is an implied one bit. + * + * IEEE-754 (Intel) "single precision" format: * 32 bits: * Bit 31 Bit 22 Bit 0 * | | | @@ -672,6 +682,9 @@ yasm_floatnum_get_sized(const yasm_floatnum *flt, unsigned char *ptr, yasm_internal_error(N_("unsupported floatnum functionality")); } switch (destsize) { + case 2: + retval = floatnum_get_common(flt, ptr, 2, 10, 1, 5); + break; case 4: retval = floatnum_get_common(flt, ptr, 4, 23, 1, 8); break; @@ -702,6 +715,7 @@ int yasm_floatnum_check_size(/*@unused@*/ const yasm_floatnum *flt, size_t size) { switch (size) { + case 16: case 32: case 64: case 80: |