Databases
StackOverflow2013
Postsdbo
POvectorize a loop which accesses non-consecutive memory locations
 

Note that there are some explanatory texts on larger screens.

plurals
  1. POvectorize a loop which accesses non-consecutive memory locations
    primarykey
    data
    text
    <p>I have a loop of this structure</p> <p><strong>Reference : <a href="http://sc.tamu.edu/help/power/powerlearn/html/ScalarOptnw/sld018.htm" rel="nofollow">Maxwell Code Example</a></strong></p> <pre><code>do z=1,zend do y=1,yend do x=1,xend k=arr(x,y,z) do while(k.ne.0) ix=fooX(k) iy=fooY(k) iz=fooZ(k) x1=x(ix ,iy ,iz) x2=x(ix+1,iy ,iz) x3=x(ix ,iy+1,iz) x4=x(ix+1,iy+1,iz) x5=x(ix ,iy ,iz+1) x6=x(ix+1,iy ,iz+1) x7=x(ix ,iy+1,iz+1) x8=x(ix+1,iy+1,iz+1) y1=y(ix ,iy ,iz) y2=y(ix+1,iy ,iz) y3=y(ix ,iy+1,iz) y4=y(ix+1,iy+1,iz) y5=y(ix ,iy ,iz+1) y6=y(ix+1,iy ,iz+1) y7=y(ix ,iy+1,iz+1) y8=y(ix+1,iy+1,iz+1) z1=z(ix ,iy ,iz) z2=z(ix+1,iy ,iz) z3=z(ix ,iy+1,iz) z4=z(ix+1,iy+1,iz) z5=z(ix ,iy ,iz+1) z6=z(ix+1,iy ,iz+1) z7=z(ix ,iy+1,iz+1) z8=z(ix+1,iy+1,iz+1) sumX+=x1+x2+..x8 sumY+=y1+y2+..y8 sumZ+=z1+z2+..z8 k=linkArr(k) enddo enddo enddo enddo </code></pre> <p>x1 through x8 are the 8 corners of a rectangular cuboid. There are three challenges to vectorize this code. One is that the 8 array elements are not contiguous in memory. Second is the inherent while loop structure along with linked List access. Third the values of ix, iy, iz returned from from fooX, fooY, fooZ are not not contiguous. So each iteration of the loop has a completely different set of ix, iy, iz. So the even across the iterations the memory access is scattered. I tried the following approaches: 1. unrolled the 3-level DO loops as :</p> <pre><code>do z=1,zend do y=1,yend do x=1,xend if(arr(x,y,z).NE.0) then kArr(indx)=arr(x,y,z) DO WHILE (kArr(indx).NE.0) indx = indx + 1 kArr(indx)=linkArr(kArr(indx-1)) ENDDO endif enddo enddo enddo </code></pre> <p>With this i have got rid of the while loop structure and now I'm able to run one big loop on kArr inside which i group 8 elements (say my VPU can accomodate 8 sets of data at a time). It did not give a performance improvement. I can post the details of these if anyone is interested. I need suggestions on how to optimize this code. Another option i tried was to combine x,y,z data in a single array so that when i compute x1, y1 &amp; z1 also will be in adjacent memory locations.</p>
    singulars
    1. PO
      singulars
      1. PTAnswer
    1. This table or related slice is empty.
    1. PTQuestion
    1. USarunmoezhi
    1. USarunmoezhi
    plurals
    1. This table or related slice is empty.
    1. PL
      singulars
      1. LTLinked
    1. This table or related slice is empty.
    1. PO
      singulars
      1. PTAnswer
    1. VO
      singulars
      1. POvectorize a loop which accesses non-consecutive memory locations
      1. USarunmoezhi
      1. VTBountyStart
    2. VO
      singulars
      1. POvectorize a loop which accesses non-consecutive memory locations
      1. This table or related slice is empty.
      1. VTUpMod
    3. VO
      singulars
      1. POvectorize a loop which accesses non-consecutive memory locations
      1. This table or related slice is empty.
      1. VTUpMod
    1. COYou didn't tell us enough. What happens to the value x1...x8, y1..y8, z1..z8? The way the code is presented now they can all be eliminated and you don't have a computational problem because your loop body ie empty.
      singulars
      1. POvectorize a loop which accesses non-consecutive memory locations
      1. USIra Baxter
    2. CO@IraBaxter: edited the code snippet. Thanks for pointing that out or my loop execution would be pointless.
      singulars
      1. POvectorize a loop which accesses non-consecutive memory locations
      1. USarunmoezhi
    3. COIf you really want to improve the performance, then you should consider using SSE intrinsics. "Hoping" that the compiler will produce fast/vectorized code will certainly lead to disappointment. I always assume that the compiler generates lousy code and that time critical sections need to be written in ASM or intrinsics (this assumption is usually correct).
      singulars
      1. POvectorize a loop which accesses non-consecutive memory locations
      1. USBitBank
 

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