CompuServe Thread

Moto FFP Details?

6 messages in this thread
#40870From: SyndesisMay 5, 1994 10:16 AM
I still haven't found any paper docs on the bit-level details of the Motorola fast-floating-point FFP numeric format. I need to write some portable C code that maps as best as possible between IEEE and FFP. (Yes, I know there will be some truncation involved because they're not the same.) My usual sources of info have turned up flat…
#40871From: Dale LarsonMay 5, 1994 1:39 PM
I would think Moto would send you something. Why won't they? — Dale L. Larson, Intangible Assets Manufacturing
#40880From: ShraddhanMay 6, 1994 2:49 AM
The ROM Kernel Manual – Libraries and Devices states the format of the Motorola FFP format. The current edition says that the format is 24 bits of mantissa in the range [0.5, 1) followed by 8 bits of excess-64 exponent. This cannot be correct. The previous edition (for WB 1.3) states that it is 24 bits of mantissa in the range [0.5, 1), then 1 sign bit and 7 bits of excess-64 exponent. This latter seems to be the (more) correct specification. The value is always normalised, except for 0.0 which is all zero bits. The old RKM says that the range of values is 5.42101070e-20 to 9.22337717e18, -9.22337717e18 to -2.71050535e-20, and 0.0 I hope this makes sense. Regards, Shraddhan (via AP from Hertfordshire, England)
#40882From: SyndesisMay 6, 1994 10:24 AM
I guess I shouldn'tve overlooked the original source. 🙂
#40883From: ShraddhanMay 6, 1994 11:16 AM
> I guess I shouldn'tve overlooked the original source. I didn't know the RKM was the original source. I thought that the FFP format already existed and Commodore merely used it. Or am I mistaken? Regards, Shraddhan (via AP from Hertfordshire, England)
#40893From: SyndesisMay 7, 1994 11:19 AM
No, I meant that I should've checked the RKMs myself. I rarely use them, they're hidden behind stacks of junk on the bottom shelf of a bookcase.