Re: [cc65] Re: floating point support redux

From: Daniel Pralle <dpralle1gmx.de>
Date: 2010-04-29 08:27:53
On 04/28/2010 06:03 PM, silverdr@wfmh.org.pl wrote:
>
> Direct using of ROM routines is not a viable option but what would make more sense could be a "reimplementation based on the concepts" found in the best (definition of "best" to be defined :-) out of the available ROM (or external) implementations. Of course some targets might not be able to use it due to memory constraints, but this is secondary. Many would benefit from it.
>
Exactly my thoughts. One routine works on all 6502-platforms. I don't 
care about the "best", except one is noticeably faster. I would give up 
precision for speed any time. The ATARI-rom is using 6 byte BCD numbers. 
If there really is an interest I can look more into the specifics. 
However, I have (yet) no clue how to support FP in a compiler. Things 
which spring to mind are:

   - constant parsing
   - argument pasing (on stack)
   - type conversions

I guess Uz has a bit more knowledge on that topic :)

Would IEE754 compatibility be satisfied enough, when we use (single) 
floats, convert them to the internal representation and convert the 
result back to single precision float?


salut,
    daniel
----------------------------------------------------------------------
To unsubscribe from the list send mail to majordomo@musoftware.de with
the string "unsubscribe cc65" in the body(!) of the mail.
Received on Thu Apr 29 08:28:01 2010

This archive was generated by hypermail 2.1.8 : 2010-04-29 08:28:03 CEST