Search
Tuesday, 28th of September 2021, 23:43:50 UTC
6:33:07
ecraven
I have the following small-float format that I want to decode: a 7-bit excess-100 exponent [I think this is octal, so actually 64] and a 17-bit 2's complement normalized mantissa. "normalized" means the left-most bit of the mantissa has to be a 1, right?
6:37:17
pjb
ecraven: yes, but it may be omited. In IEEE it's not stored.
6:37:43
ecraven
that's not the case here, I think
6:44:27
JeSCaLis23-F
ACTION https://o11.me/VK3KV
6:45:30
mfiano
ecraven: What is this for/where did you get those numbers from?
6:45:53
ecraven
the MIT CADR uses this for small flonums
6:46:16
ecraven
the numbers are from uc-arith.lisp, the microcode that implements small flonums, for sys98
6:46:35
mfiano
This seems like the format IBM Mainframes used for floats
6:46:48
ecraven
that was before my time :-/ I'll look into that, thanks
6:49:17
mfiano
Do you have a link to this file?
6:51:06
ecraven
https://tumbleweed.nu/r/sys98/file?name=ucadr/uc-arith.lisp&ci=tip
6:56:30
mfiano
and wow, i can see how that design could be confusing
6:56:32
ecraven
that's sys98, there are other dumps of different versions
6:56:38
mfiano
trying to understand the low-level form
6:56:41
ecraven
ah, different from IBM.. that might be ;)
6:56:49
ecraven
the microcode? that's it's own art :D
Wednesday, 29th of September 2021, 11:43:50 UTC