On 5/10/25 09:36, Andreas Schwab wrote:
On Mai 07 2025, Richard Henderson wrote:
+ /* Convert the 3 digit decimal exponent to binary. */
+ exp = ((hi >> 24) & 0xf)
+ + ((hi >> 20) & 0xf) * 10
+ + ((hi >> 16) & 0xf) * 100;
This is backwards. An exponent of 123 is stored as 0x01230000, not
0x03210000.
+ for (int i = 24; exp10; i -= 4, exp10 /= 10) {
+ res_hi |= (exp10 % 10) << i;
Again, this is backwards. Also, the decimal exponent of an extended
precision number can be as big as 4951. From the fpsp040:
| A15. Convert the exponent to bcd.
| As in A14 above, the exp is converted to bcd and the
| digits are stored in the final string.
|
| Digits are stored in L_SCR1(a6) on return from BINDEC as:
|
| 32 16 15 0
| -----------------------------------------
| | 0 | e3 | e2 | e1 | e4 | X | X | X |
| -----------------------------------------
That is, the fourth digit of the exponent is stored in the upper 4 bits
of the padding (OPERR is set if e4 is non-zero).
Well that's annoying. There's a mis-print in the m68000pm/ad manual, where it
says
SE SM YY E0 E1 E0 E3 x x D16
I didn't notice E0 listed twice and so thought "0 1 2 3" was the order.
Will fix.
r~
