Databases
StackOverflow2013
Postsdbo
PO
 

Note that there are some explanatory texts on larger screens.

plurals
  1. PO
    primarykey
    data
    text
    <p>The state of the art in GPU sorting isn't particularly encouraging. </p> <p>For sorting 32-bit integers the following paper from 2009 (with 2 authors who are researchers at Nvidia) only claims 23% increase for the best CUDA sort on GTX280 compared to the best CPU sort on a 4 core Yorkfield.</p> <p><a href="http://www.mgarland.org/files/papers/gpusort-ipdps09.pdf" rel="nofollow noreferrer">http://www.mgarland.org/files/papers/gpusort-ipdps09.pdf</a></p> <p>This used a radix sort on the GPU, and merge sort on CPU. You'd need a comparison-based sort in order to construct a suffix array, so instead of GPU radix sort the best of those in the paper would be GPU merge sort, which achieved about half the speed of GPU radix sort (with 1 million keys) - ie about 40% slower than the CPU merge sort.</p> <p>Adding variable length keys seems likely to cause threads in a warp will get out of sync on a GPU, so would reduce the performance on GPU more than CPU.</p> <p>Overall if your purpose is to build an efficient system, I'd recommend that you use a CPU implementation for this problem because it will be faster and easier to write.</p> <p>But, if your purpose is to experiment or just to learn about GPU, then you can find the CUDA implementation of merge sort from the paper in the CUDA SDK:</p> <p><a href="http://developer.download.nvidia.com/compute/cuda/sdk/website/Data-Parallel_Algorithms.html" rel="nofollow noreferrer">http://developer.download.nvidia.com/compute/cuda/sdk/website/Data-Parallel_Algorithms.html</a></p>
    singulars
    1. This table or related slice is empty.
    1. POIs there algorithm for sorting array of strings for GPU?
      singulars
      1. PTQuestion
    1. PTAnswer
    1. This table or related slice is empty.
    1. USRD1
    plurals
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. POIs there algorithm for sorting array of strings for GPU?
      singulars
      1. PTQuestion
    1. This table or related slice is empty.
    1. VO
      singulars
      1. PO
      1. This table or related slice is empty.
      1. VTAcceptedByOriginator
    2. VO
      singulars
      1. PO
      1. This table or related slice is empty.
      1. VTUpMod
    3. VO
      singulars
      1. PO
      1. This table or related slice is empty.
      1. VTUpMod
    1. COIsn't the whole point of CUDA to use a processor that is idle anyway? Even if you got no speed improvement whatsoever on a GPU over a CPU, you would still have a 2X improvement over having a CPU only, provided you can effectively utilize the extra parallelism.
      singulars
      1. PO
      1. USRobert Harvey
    2. CO@Robert Harvey - most uses of CUDA don't keep the CPU busy at the same time. However recently this has become more common, and is usually called hybrid GPU/CPU. The need to copy in between the CPU and GPU memories tends to make it quite tricky to obtain good performance though. In this case, I'd expect at best you'd achieve 150% of the CPU speed, and you'd be better off investing in a system with two CPUs.
      singulars
      1. PO
      1. USRD1
    3. COThanks for your answer. I agree with all your notes about sorting strings on a GPU, I thought in the same way, but I had hoped that there is an algorithm which I had missed.
      singulars
      1. PO
      1. USKentzo
 

Querying!

 
Guidance
A row detail

Detail views are divided into sections. All the information in the data section comes from columns in the selected row. The other sections display data from other, related rows.

Related data can be related in a to-one or a to-many fashion. Captions of data related in a to-many fashion link to a list view showing a filtered view of the table.

Try moving around until you find a non-empty to-many entry and click on the label to get to one. You can move back to the root by clicking on the database name in the header.

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload