> > "And on other hand,32-bit code on the AMD64 CPU family is not > > native, and that is why 32-bit code is significant slower on this CPU > > family." > > If that's in the Postgres FAQ, then I'm surprised at the FUD they've got > masquerading as fact. Try the MySQL FAQ instead :)
Taking the article which is closed to what you'd like to expect is not always sience. I presume the postgresql group has done some testing with Solaris and 32/64 bit mode on amd. As also the MySQL team on a linux 32/64bit AMD. > > There is NO speed difference between 32 and 64 bit on AMD processors. > It runs natively either way. If you run on an Itanium, then it's a > different story - they do some fancy code conversion if given 32 bit > binaries. > > The main difference for database use is that 64 bit processors can > address MUCH more memory, so your DB software doesn't have to do nasty > tricks ( that slow things down ) to deal with large databases. So essentially you say here, I believe MySQL is right, there is no speed difference between 32 and 64 bit AMD, but the 32 bit is slower? Native does NOT mean no speed difference, it just means there is no fancy stuff scaling 32 bit operations to 64 bit. If a pointer is 64 bit on the other hand, it has to be scaled to 32 bit in 32 bit mode, which is done automatically / native (as has been done in the whole 32bit era) , but still results in 32 bits of the adres generater being unused. This will result in more calculation which can be done in 1 when it was in 64 bit mode. More calculation is more time. However, dbmail uses a lott of (IMHO unnecessary) BIG INT variables. Which is by default more efficient if calculated in 64bit mode. Also indexing/indices with on BIG INT columns will work better. Marc _______________________________________________ DBmail mailing list [email protected] https://mailman.fastxs.nl/mailman/listinfo/dbmail
