On 02/14/2009 03:43 PM, Kevin Kofler wrote:
I realize, but my recent testing was with x86_64 chips (and IA64 chips). The bottom line is that while x86_64 should be faster, because of their are 8 more registers than the 32-bit version, as well as SSE and SSE2, and linear memory addressing. But, there are still applications that run better in 32-bits. In my benchmarks my companies product ran much slower on the IA64 than on the x86_64 (in 32-bit mode), but after profiling we found it did a lot of jumps. But, comparing, Alpha, x86_64, IA64 is certainly not good comparisons. My benchmarks also indicated that running a 32-bit application on x86_64 hardware with a 32-bit OS was slower than the same application on a very similar box with a 64-bit OS. But, for the most part, 64-bits is a win.x86 is a very different story than the Alpha. x86_64 adds more registers, which is already enough to boost program speed (fewer memory accesses needed). x86_64 also means at least SSE and SSE2 are guaranteed to be there, so you also benefit from those, whereas on 32-bit x86 only a few libs have sse2 versions available.
-- Jerry Feldman <gaf blu org> Boston Linux and Unix PGP key id: 537C5846 PGP Key fingerprint: 3D1B 8377 A3C0 A5F2 ECBB CA3B 4607 4319 537C 5846
Description: OpenPGP digital signature