CompuServe Thread

Math Trans Help

12 messages in this thread
#117546From: SyndesisAug 22, 1990 7:18 PM
I'd like to use the IEEE math translation libraries in a Manx 3.6 program, to convert some 64 bit IEEE doubles to FFP single precision format. (I know, I'll lose precision.) Any tips on creating and using these functions, if I'm currently compiled with Motorola FFP floats? I'll start studying the Manx source code for these libraries, but any tips would be appreciated. Language warriors please drop dead, I haven't the time to argue "Why?".
#117570From: Mike Spille/ManxAug 22, 1990 9:56 PM
We don't provide any routines for converting from IEEE<->FFP. However, I think there are some routines provided by C= to do such a conversion. I'm at home at the moment and don't have my RKMs handy, but I can check for you tomorrow if you'd like. -Mike
#117584From: SyndesisAug 23, 1990 12:45 AM
Yes, they can be found in the 'mathtrans' libraries. However, double IEEE numbers are 64 bit entities. Commodore's routines require these to be passed in d0/d1 for instance. How could Manx 3.6 handle this? Perhaps I need to code a little assembler glue to get this to work… The 32 bit things could be passed like any long value.
#117619From: Mike Spille/ManxAug 23, 1990 2:20 PM
Yep, some assembler glue code would be in order. -Mike
#117623From: SyndesisAug 23, 1990 3:25 PM
Any ideas on a reference on IEEE numbers, short of ordering some obscure $50 article from IEEE? I'd at least like to confirm that the 8087 32 and 64 bit floats are the same as the ones used in the Amiga IEEE, and that I don't need to do any byte swapping, etc. I'm reading some data in PC IEEE format into the Amiga…
#117628From: Mike Spille/ManxAug 23, 1990 3:55 PM
The Motorola 68881 reference manual describes the IEEE formats (all of 'em). I don't see an ISBN on it or anything so I guess you have to order it directly from Motorola. I'm 95% sure that the 8087's format is identical to the Amiga IEEE (both single and double precision). -Mike
#117733From: SyndesisAug 24, 1990 2:56 PM
Why wasn't any support provided in Manx to use these extended math libraries, aside from choosing which math model to use? What I mean is, shouldn't I be able to call these functions directly, instead of building my own glue for them?
#117741From: Mike Spille/ManxAug 24, 1990 4:24 PM
I think (not sure though) that 3.6 came out before we knew the exact formats of the calls. In any case 5.0 has glue and #pragmas for all the 1.3 functions (and version 5.0d will have all 2.0 stuff). -Mike
#117947From: SyndesisAug 26, 1990 10:44 AM
Formats of the calls? I thought these libraries have been in the Amiga all along. Hmm. I should look at my 5.0 disks…
#117749From: New memberAug 24, 1990 5:55 PM
Language warriors please drop dead? <- may I use this? <grin>
#117945From: SyndesisAug 26, 1990 10:41 AM
Sure. My royalties are low. 🙂 When the Manx vs. Lattice wars get started, I rarely hear debate about the most important issue to me, as exemplified my this situation. I've got a couple megs of code in the InterChange/InterFont products, and can't make the switch to Lattice without a long period of testing. And in other projects at Syndesis, we use Lattice, but would like to interface to other Manx code… Switching is not an issue. Getting things done is the issue.
#118119From: New memberAug 28, 1990 10:16 AM
Well said! I never EVER met a person who was luke warm over his computer choice or him main language. Nuff said!