Databases
StackOverflow2013
Postsdbo
POmoving elements between arrays in a CUDA kernel
 

Note that there are some explanatory texts on larger screens.

plurals
  1. POmoving elements between arrays in a CUDA kernel
    primarykey
    data
    text
    <p>I am stuck in a very simple thing and I need an opinion. I have a very simple kernel in CUDA that copies the elements between two arrays (there is a reason I want to do it in this way) and </p> <pre><code>__global__ void kernelExample( float* A, float* B, float* C, int rows, int cols ) { int r = blockIdx.y * blockDim.y + threadIdx.y; // vertical dim in block int c = blockIdx.x * blockDim.x + threadIdx.x; // horizontal dim in block if ( r &lt; rows &amp;&amp; c &lt; cols) { // row-major order C[ c + r*cols ] = A[ c + r*cols ]; } //__syncthreads(); } </code></pre> <p>I am taking unsatisfying results. Any suggestions please?</p> <p>The kernel is called like this: </p> <pre><code>int numElements = rows * cols; int threadsPerBlock = 256; int blocksPerGrid = ceil( (double) numElements / threadsPerBlock); kernelExample&lt;&lt;&lt;blocksPerGrid , threadsPerBlock &gt;&gt;&gt;( d_A, d_B, d_C, rows, cols ); </code></pre> <p><strong>Updated</strong>(After Eric's help):</p> <pre><code>int numElements = rows * cols; int threadsPerBlock = 32; //talonmies comment int blocksPerGrid = ceil( (double) numElements / threadsPerBlock); dim3 dimBlock( threadsPerBlock,threadsPerBlock ); dim3 dimGrid( blocksPerGrid,blocksPerGrid ); kernelExample&lt;&lt;&lt;dimBlock, dimBlock&gt;&gt;&gt;( d_A, d_B, d_C, rows, cols ); </code></pre> <p>For example having the matrix A </p> <pre><code>A =[ 0 1 2 1 0 2 0 0 2 0 0 1 2 1 2 2 2 2 0 0 2 1 2 2 3 1 2 2 2 2 ] </code></pre> <p>the returned matrix C is</p> <pre><code>C = [ 0 1 2 1 0 2 0 0 2 0 0 1 2 1 2 2 2 2 0 0 2 1 2 2 3 1 2 2 2 2 ] </code></pre>
    singulars
    1. PO
      singulars
      1. PTAnswer
    1. This table or related slice is empty.
    1. PTQuestion
    1. USDarkmoor
    1. USDarkmoor
    plurals
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. PO
      singulars
      1. PTAnswer
    1. VO
      singulars
      1. POmoving elements between arrays in a CUDA kernel
      1. This table or related slice is empty.
      1. VTDownMod
    1. COWhat does "unsatisfied results" mean? Please describe the problem you are having in more detail. It is pretty hard to help you if you don't tell us what the problem is....
      singulars
      1. POmoving elements between arrays in a CUDA kernel
      1. UStalonmies
    2. COYour "update" can't be correct - a 256x256 blocksize is illegal on every CUDA GPU ever made, so the kernel won't launch.
      singulars
      1. POmoving elements between arrays in a CUDA kernel
      1. UStalonmies
    3. COYes you are right talonmies sorry for the wrong info but the kernel it worked, an interesting thing happened. I debugged the code and I noticed that the grid was split in blocks of 1x1 dim. That means that blockIdx.y = 0,...,14 blockIdx.x = 0,...,14 ,blockDim.y = 1 ,blockDim.x = 1 and threadIdx.x = 0 , threadIdx.y = 0. Do you thing that I must update the post? Thanks again.
      singulars
      1. POmoving elements between arrays in a CUDA kernel
      1. USDarkmoor
 

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