Why should the results always be rounded to ISO-C specifications? The Amiga RKRM was written primarily for SAS-C, but it also provided a few examples for assembler code.
IIRC the C= mathlibs didn't offer rounding options (for performance reasons). Also, for the OS3.9 mathlibs, H&P wanted to get rid of 68000/68010 + 68881/68882 hardware (hack) support but not add support for any rounding options.
I don't know about OS3.1.4 or OS3.2. But you could check the NDK for these versions.
For the hardware 68K FPU systems, rounding options are available but not always from the mathlibs (e.g. you roll your own FPU code).
Last edited by SpeedGeek; 22 July 2021 at 14:41.
|