On Mon, Jan 24, 2005 at 05:03:47PM -0700, Seth Kurtzberg wrote: > I recently did a port of linux to ARM, and the floating point issue came up.
Thank you for the detailed reply, Seth. I've been discussing this with Ian Lynagh, who maintains ghc for Debian, and has encountered trouble before. Presently, we do not have a working ARM ghc package, so my first priority is basic floating point support, even if it is slow. [ snip ] > If your issue is floating point functionality rather than floating point > performance, the standard gcc build and linux software emulation of > floating point hardware works fine. You only need to do the more Ian told me that the standard build would die with errors relatnig to decodeFloat. He's also written up a little illustration that ARM's internal floating point representation is different than on other platforms. On Debian, our standard procedure appears to be, from what I can tell, using the kernel trap for floating point. Also, from what I understand from gcc docs, if the software library is to be used instead, the entire system must be recompiled starting with libc. It seems that you have found a way around that... correct? We can't really rebuild the entire system, but it sounds like the library is the way to go. My own efforts are complicated a bit because my only piece of ARM hardware is my Zaurus PDA. It's powerful enough to run binaries that GHC builds, or even GHC itself in some instances, but not powerful enough to build GHC. So I have to rely on various other ARM machines, where I have little control over the kernel or libc environment. In any case, your build instructions and/or code would be most helpful. Also, if you happen to have ghc binaries ready-made on ARM, that would speed up my porting work, of course :-) Thanks again for your help. -- John _______________________________________________ Glasgow-haskell-users mailing list Glasgow-haskell-users@haskell.org http://www.haskell.org/mailman/listinfo/glasgow-haskell-users
