Databases
StackOverflow2013
Postsdbo
PO
 

Note that there are some explanatory texts on larger screens.

plurals
  1. PO
    primarykey
    data
    text
    <p>In order to get copy/compute overlap, you must use pinned memory. The reason for this is contained in the paragraph you excerpted. Presumably the whole reason for your multi-streamed approach is for copy/compute overlap, so I don't think the correct answer is to switch to using pageable memory buffers.</p> <p>Regarding your question, assuming <code>h_mem</code> is only used as the source buffer for the pseudo-code you've shown here (i.e. the data in it only participates in that one <code>cudaMemcpyAsync</code> call), then the h_mem buffer is no longer needed once the <em>next</em> cuda operation in that stream begins. So if your <code>kernel_launch</code> were an actual <code>kernel&lt;&lt;&lt;...&gt;&gt;&gt;(...)</code>, then once <code>kernel</code> begins, you can be assured that the previous <code>cudaMemcpyAsync</code> is complete.</p> <p>You could use cudaEvents with <code>cudaEventSynchronize()</code> or <code>cudaStreamWaitEvent()</code>, or you could use <code>cudaStreamSynchronize()</code> directly in the stream. For example, if you have a <code>cudaStreamSynchronize()</code> call somewhere in the stream pseudocode you have shown, and it is after the <code>cudaMemcpyAsync</code> call, then any code after the <code>cudaStreamSynchronize()</code> call is guaranteed to be executing after the <code>cudaMemcpyAsync()</code> call is complete. All of the calls I've referenced are documented in the <a href="http://docs.nvidia.com/cuda/cuda-runtime-api/index.html" rel="nofollow">usual place</a>.</p>
    singulars
    1. This table or related slice is empty.
    1. POWhen is safe to reuse CPU buffer when calling cudaMemcpyAsync?
      singulars
      1. PTQuestion
    1. PTAnswer
    1. This table or related slice is empty.
    1. USRobert Crovella
    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. This table or related slice is empty.
    1. VO
      singulars
      1. PO
      1. This table or related slice is empty.
      1. VTUpMod
    1. COSo I must ensure the memcpyasync has ended before reusing the host memory buffer (h_mem), I could do that synchronizing by puting an event after the cudamemcpyasync call but, then I should make a thread waiting for this event... so I guess I could have a look at this "stream callbacks" pointed on this blog: http://blog.cudahandbook.com/2012/09/16/stream-callbacks.aspx
      singulars
      1. PO
      1. USDredok
    2. COBTW. From the paragraph I excerpted I see that memcpyasync will return when the host memory is copied to a stagging buffer without any device sync. So I guess I could use a memcpyAsync without pinned memory for the first copy and one with pinned memory for the second copy (the results).
      singulars
      1. PO
      1. USDredok
 

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