On Sep 12, 2:52 am, jason <ja...@njkfrudils.plus.com> wrote: > Hi > > I've removed all the defines HAVE_HOST_CPU_* , except those left > behind in the msvc build directory. > Some sacrifices were made , our system would of handled them , but as > they were only minor optimizations for very old cpus's (ie 386,486), I > didn't bother. > There was only one thing for a modern chip , powerpc doesn't have a > time stamp counter for tuning , I will turn it on temporary when we > tune for our release , but to do it properly I'll have to rationalize > all the timing stuff , which will have to wait until the next release. > I also cleared out non-ieee float stuff and other old bits. > > Jason
I'm going to put the time stamp counter macro into longlong.h , this will remove the need for all the extra existing logic that we currently use. Another simplification can also be done , if we restrict speed.exe to only use rdtsc or a generic fallback then another lot of horrible code can be got rid of , however this does reduce some of our options for speed.exe ie the timing results will only display in cycles or the generic fallback form , it will no longer display in seconds . The existing speed.exe has generic scaling method anyway so all this change means is that for users who want seconds will have to know the frequency of their cpu. I believe the current code is wrong anyway as cpu's with turbo boost which is all current chips , will report incorrect results anyway. 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.
