Posted by: Anonymous Coward
on September 14, 2006 11:04 PM
Running on 64 bits HAS a definite advantage: getting rid of IA32's Achille's heel: lack of registers who in addition to its direct effect also makes life difficult for compilers (several optimizatiopn algorithms don't work that well on register starved machines like IA32).
On a XEON running RHEL 4 I got 50% improvement on a pure CPU benchmark when I reinstalled with the 64 bit version. Caveats: That was really pure CPU, the entire benchmark fitted in cache. In real world applications who do I/O and who access memory the improvements would have been lesser. Also FP intensive get some improvement (from janitorial operations like saving pointers on the stack) but far less than 50%
However with Gcc progressikng and thus becoming more accurate for the quirks of Xeons/Athlons in 64 bit mode I expect the gap increases.
Learn about Linux
Download Linux
Get Linux help
Registers
Posted by: Anonymous Coward on September 14, 2006 11:04 PMOn a XEON running RHEL 4 I got 50% improvement on a pure CPU benchmark when I reinstalled with the 64 bit version. Caveats: That was really pure CPU, the entire benchmark fitted in cache. In real world applications who do I/O and who access memory the improvements would have been lesser. Also FP intensive get some improvement (from janitorial operations like saving pointers on the stack) but far less than 50%
However with Gcc progressikng and thus becoming more accurate for the quirks of Xeons/Athlons in 64 bit mode I expect the gap increases.
#