Ouch, Thanks for the link. Very deep and broad subject I stumbled on. But still, there must be some computational primitives that could be used to build a cross platform 128 bit exact floating point library. With all the SIMD sub-processors, and GPGPU tricks being developed, the performance hit couldn't be too bad. Even a software / integer only layer for brain dead platforms would be useful. I'm getting from this paper that even the 64 and 32 bit floating point results on different platforms could be different. Ouch again.
Looks like the GPGPU programmers have even worse problems... http://www.cs.unc.edu/~ibr/projects/paranoia Thanks again. This message posted from opensolaris.org _______________________________________________ opensolaris-code mailing list [email protected] https://opensolaris.org:444/mailman/listinfo/opensolaris-code
