Deneidez on 10 January 2009
| haxxiy said: I assume you're saying that double precision are slower than single precision, which has no base at all. They provide more exponent width and significant precision than common single flops. Thus it provides numerical stability to the cpu and make it capable of handling more tasks. |
Wtf are you talking about? :D
Each PPU can complete two double precision operations per clock cycle using a scalar-fused multiply-add instruction, which translates to 6.4 GFLOPS at 3.2 GHz; or eight single precision operations per clock cycle with a vector fused-multiply-add instruction, which translates to 25.6 GFLOPS at 3.2 GHz.
Source: Wikipedia
( If you would have any linux and I wouldn't be sick, I could modify my 3D engine to use double instead of floats and you would see that it would make it run slower, much slower. The same would happen with oh so great CELL. )
