Slashdot Mirror
Impressive Benchmarks: Sorting with a GPU
An anonymous reader writes "The Graphics research group at the University of North Carolina at Chapel Hill has posted some interesting benchmarks for a sorting implementation which is done entirely on a GPU. There have been efforts on doing general purpose computation on GPUs before (previous Slashdot article). However, most of them had generally utilized the fragment processing pipeline of the GPUs which is slower then the default high speed rendering pipeline. Apparently, the above implementation is done using "simple texture mapping operations" and "cache efficient memory accesses" only. There also seems to an option to download the distribution for non-commercial use, though the requirements seem pretty hefty (a very decent nVidia graphics card and the latest nVidia drivers)."
"Who the fuck has enough data to sort on a regular basis that they'd need hardware-assisted sorting?"
Database admins?
Who the fuck has enough data to sort on a regular basis that they'd need hardware-assisted sorting?
Perhaps you've heard of, I dunno.. Google, or Oracle?
I've heard about doing GPU calculations (as per the implementation being non-IEEE)? Doubly, would someone in the know help explain what the aforementioned "weirdness" means?
There are just some functions missing, like different roundings and the missing double precision. GPUs are simply not optimized for scientific calculations, but that doesn't mean that they can't be programmed to be useful for those workloads.
Well, to be fair GPUs are only "so much more powerful than CPUs" if your task is suitable for running on a GPU. If not, then you're better off using the CPU.
Kind of like how a bulldozer is much more powerful than a hammer, but totally unsuitable to banging a nail in a bit of wood. If you want something torn down or moved about, though...
It's official. Most of you are morons.
It is important because sorting is an important algorithmic building block for solving many problems. If you want to solve important problems entirely on the GPU (which has many more GFLOPS available than the CPU, if you can manage to use them efficiently), then you need to be able to do the sorting component of those larger algorithms.
In high flow operations like sorting, memory latency and bandwidth matters.
Woh.
What seems unfair to me is comparing 24bit accuracy with 32bit. Yea, you can do it faster, but you're sacrificing a lot of the usefulness of doing it!