On Apr 27, 10:43 pm, jason <ja...@njkfrudils.plus.com> wrote: > Hi > > I going to remove some minor optimizations for a fat build on the K6 > and K7 cpu's , again I think that anyone running on these cpu's would > use a normal build anyway , and I think they are not made anymore even > as embedded chips. > > Jason
I've removed the old crud above. In gmp-impl we have these functions which have asm in them popc_limb bswap_limb + varients ulong_parity mpn_incr mpn_decr so I think we should move then to the arch specific longlong , the C version we will leave in gmp-imp.h rather than move them to longlong_post.h for strict compatibility (ie for people who want popc_limb and only include gmp-impl.h not longlong.h). This means including gmp-impl and longlong gives exactly the same code as before , whereas including just gmp-impl.h will give you the C version only. Given the nature of the macros above I think this is best as popc,bswap,parity are not used much and popc and parity have GCC intrinsics. The mpn_incr/decr are important but the asm code is for 32bit x86 only and as the assumption of this macro is ment to be that it loops very few(<2) times , a bit of C code should be faster. So the changes will be completely compatible and only slightly slower(if at all) on these very odd cases and in the case that longlong is not included. I'll go ahead with this unless anyone think this is a bad idea. Jason -- You received this message because you are subscribed to the Google Groups "mpir-devel" group. To post to this group, send email to mpir-devel@googlegroups.com. To unsubscribe from this group, send email to mpir-devel+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/mpir-devel?hl=en.
