4 ms·
The point about memory bandwidth and general algorithmic factors outweighing raw FPU performance holds true with pretty much all software even today. (And if yo
by taneq 10y ago
The point about memory bandwidth and general algorithmic factors outweighing raw FPU performance holds true with pretty much all software even today. (And if you genuinely need raw FPU performance that badly then you're already running that FPU crunching on a GPU.)
- gpderetta 10y agoNonsense, there is a reason that many floating point kernel as still written in hand optimised assembler. And not all FPU loads are amenable to GPU computation.
- liuliu 10y agoMore about data hazard. You can arrange your access pattern pretty well, but fast-math helps to avoid data dependency therefore better instruction level parallelism and pipelining.