Home > Articles

  • Print
  • + Share This
This chapter is from the book

4.4 An Optimized Parallel Version

So far we have marked the parallel regions for acceleration. Now it is time to introduce data regions to optimize data transfers.

4.4.1 Reducing Data Movement

Now that you have identified the problem, you know you must apply some data directives. OpenACC lets you completely control the residency of the data. It has routines to set up data during program initialization, to automatically migrate data going into or out of any region or block of code, and to update at any given point in the code. So don’t worry about what OpenACC can do. Worry about what you want to do.

Pause here and see whether you can come up with a strategy to minimize data movement. What directives does that strategy translate to? Feel free to experiment with the code on your own before reading the answer, which is provided later.

In general, we want the entire simulation to be executed on the device. That is certainly the ideal case and eliminates all the data transfer costs. But most of the time you can’t achieve that objective; the entire problem may not fit in device memory, there may be portions of the code that must execute on the host, or IO may be required at some point.

But let’s start with that objective in mind. If you load your data onto the device at the beginning of the main loop, when do you next need it on the host? Think the first iteration through as a start: there is no reason for the two big arrays to return to the host between the two kernels. They can stay on the device.

What about worst_dt? It is insignificant in size, so you don’t care what it does as long as it is available when needed, as per the default kernels behavior. Once you start to use data regions, you uncouple the execution from the data regions and could prevent unnecessary data movement. Because there is no real performance gain, you won’t override the default by including it in any data directives. It will continue to be set to 0 on the host, get to a maximum in the second nested loop (actually a reduction from all of the “local maximums” found by each processing element (PE) on the device), and get copied back to the host so that it can be checked as the condition to continue the while loop every iteration. Again, this is all default kernels behavior, so we don’t worry about the details.

After that, you run into the output routine. It isn’t an issue for the first 100 iterations, so let’s ignore it for a moment and continue around the loop for the second iteration. At the start of the second iteration, you would like both big arrays to be on the device. That is just where you left them! So it looks as if you can just keep the data on the device between iterations of the while loop. The obvious data directives would be data copy clauses applied to the while loop.

// C
#pragma acc data copy(Temperature_previous, Temperature)
while ( worst_dt > TEMP_TOLERANCE ) {
. . .

! Fortran
!$acc data copy(temperature_previous, temperature)
do while ( worst_dt > temp_tolerance )
. . .

This is indeed the key. It will significantly speed up the code, and you will get the right answer at the end.

However, you do need to address that track_progess() output routine that gets invoked every 100 iterations. You need for the temperature to be back on the host at that point. Otherwise, the host copy of temperature will remain at the initial condition of all zeros until the data copy happens at the termination of the while loop, which is the end of the data region. Many programmers encounter this oversight when they apply the data directives, run the code to a quick completion in the expected 3,372 iterations, and assume victory, only to notice that all of their printed output has been zeros. Make sure you understand exactly how this happens, because it is a good example of what can occur when we decouple the data and execution regions using data directives.

The fix is easy. You just need an update at that point.

// C
. . .
if((iteration % 100) == 0) {
   #pragma acc update host(Temperature)
. . .

! Fortran
. . .
if( mod(iteration,100).eq.0 ) then
   !$acc update host(temperature)
   call track_progress(temperature, iteration)
. . .

It is important to realize that all the tools for convenient data management are already in OpenACC. Once you decide how you want to manage the data conceptually, some combination of data copy, declare, enter/exit, and update clauses should allow you to accomplish that as you wish. If you find yourself fighting the scope or blocking of your code to make the directives match your wishes, take a breath and ask yourself whether the other clauses will allow you to accomplish this more naturally.

4.4.2 Extra Clever Tweaks

There is one more tweak you can apply to the code before you declare victory. If you look a little more carefully at the code, you might notice that you don’t actually need to copy both big arrays into the while loop. It happens that temperature_previous is the array that is initialized in the initialization routine, and temperature uses these values to set itself in the first iteration. So you don’t need to copy it in.

Continuing with that line of thought, you don’t need for both arrays to exit the while loop with the final data; one will suffice. Once again, temperature_previous has the correct values so that you can abandon temperature on the device. This means that temperature is really just a temporary array used on the device, and there is no need to copy it in or out. That is exactly what the data create clause is for.

Note that this last optimization is really not very important. The big win was recognizing that you were copying the large arrays needlessly every iteration. You were copying two large arrays into and out of each of the two kernels each loop:

(2 arrays) × (in and out) × (2 pairs of loops) × (3,372 iterations) = 26,976 copies

Getting rid of all those transfers with a data copy was the big win. Using data create instead of copy for the Temperature array saved one copy in at the beginning of the entire run, and one copy out at the end. It wasn’t significant. So don’t feel bad if you didn’t spot that opportunity.

Likewise, using an update for the track progress routine caused 33 transfers over the course of the run. It was a quick fix for the problem. In comparison to the original 26,876 copies, having 33 remaining is nothing. However now that you are down to one copy in and one copy out for the whole run, it does have an impact on the order of 5 percent of the new and significantly reduced total run time. Given the huge performance improvement you have achieved, you may not care, but for those of you seeking perfection, see Exercise 1 at the end of the chapter.

4.4.3 Final Result

Listing 4.7 shows the final C version of the OpenACC enabled routine, and Listing 4.8 shows the Fortran version.

Listing 4.7. Final C OpenACC Laplace code main loop

#pragma acc data copy(Temperature_previous), create(Temperature)
while ( worst_dt > TEMP_TOLERANCE ) {

   #pragma acc kernels
   for(i = 1; i <= HEIGHT; i++) {
      for(j = 1; j <= WIDTH; j++) {
         Temperature[i][j] = 0.25 * (Temperature_previous[i+1][j]
                                 + Temperature_previous[i-1][j]
                                 + Temperature_previous[i][j+1]
                                 + Temperature_previous[i][j-1]);

   worst_dt = 0.0;

   #pragma acc kernels
   for(i = 1; i <= HEIGHT; i++){
      for(j = 1; j <= WIDTH; j++){
         worst_dt = fmax( fabs(Temperature[i][j]-
         Temperature_previous[i][j] = Temperature[i][j];

   if((iteration % 100) == 0) {
   #pragma acc update host(Temperature)


Listing 4.8. Final Fortran OpenACC Laplace code main loop

!$acc data copy(temperature_previous), create(temperature)
do while ( worst_dt > temp_tolerance )

   !$acc kernels
   do j=1,width
      do i=1,height
         temperature(i,j) =0.25*(temperature_previous(i+1,j)&
                               + temperature_previous(i-1,j)&
                               + temperature_previous(i,j+1)&
                               + temperature_previous(i,j-1))
   !$acc end kernels


   !$acc kernels
   do j=1,width
      do i=1,height
         worst_dt = max( abs(temperature(i,j) – &
                         worst_dt )
         temperature_previous(i,j) = temperature(i,j)
   !$acc end kernels

   if( mod(iteration,100).eq.0 ) then
      !$acc update host(temperature)
      call track_progress(temperature, iteration)

   iteration = iteration+1

!$acc end data

You compile exactly as before. If you again use the compiler verbose information option (-Minfo=acc for PGI), you see that the generated copies are now outside the while loop, as intended. Here is the result.

. . .
. . .
---------- Iteration number: 3200 ------------
. . .[998,998]: 99.18  [999,999]: 99.56  [1000,1000]: 99.86
---------- Iteration number: 3300 ------------
. . .[998,998]: 99.19  [999,999]: 99.56  [1000,1000]: 99.87

Max error at iteration 3372 was 0.009995
Total time was 1.054768 seconds.

This is much better. Table 4.1 sums it up. With only a handful of directives, you have managed to speed up the serial code more than 20 times. But you had to think about your data migration in order to get there. This is typical of accelerator development.

Table 4.1. Laplace code performance

Serial 21.3  
kernels directive 35.2 0.60
data directives 1.05 20.3

To review, you looked for the large loops and placed kernels directives there. Then (prompted by terrible performance) you thought about how the data should really flow between the host and the device. Then you used the appropriate data directives to make that happen. Further performance improvements are possible (see the exercises), but you have achieved the lion’s share of what can be done.

  • + Share This
  • 🔖 Save To Your Account

InformIT Promotional Mailings & Special Offers

I would like to receive exclusive offers and hear about products from InformIT and its family of brands. I can unsubscribe at any time.


Pearson Education, Inc., 221 River Street, Hoboken, New Jersey 07030, (Pearson) presents this site to provide information about products and services that can be purchased through this site.

This privacy notice provides an overview of our commitment to privacy and describes how we collect, protect, use and share personal information collected through this site. Please note that other Pearson websites and online products and services have their own separate privacy policies.

Collection and Use of Information

To conduct business and deliver products and services, Pearson collects and uses personal information in several ways in connection with this site, including:

Questions and Inquiries

For inquiries and questions, we collect the inquiry or question, together with name, contact details (email address, phone number and mailing address) and any other additional information voluntarily submitted to us through a Contact Us form or an email. We use this information to address the inquiry and respond to the question.

Online Store

For orders and purchases placed through our online store on this site, we collect order details, name, institution name and address (if applicable), email address, phone number, shipping and billing addresses, credit/debit card information, shipping options and any instructions. We use this information to complete transactions, fulfill orders, communicate with individuals placing orders or visiting the online store, and for related purposes.


Pearson may offer opportunities to provide feedback or participate in surveys, including surveys evaluating Pearson products, services or sites. Participation is voluntary. Pearson collects information requested in the survey questions and uses the information to evaluate, support, maintain and improve products, services or sites, develop new products and services, conduct educational research and for other purposes specified in the survey.

Contests and Drawings

Occasionally, we may sponsor a contest or drawing. Participation is optional. Pearson collects name, contact information and other information specified on the entry form for the contest or drawing to conduct the contest or drawing. Pearson may collect additional personal information from the winners of a contest or drawing in order to award the prize and for tax reporting purposes, as required by law.


If you have elected to receive email newsletters or promotional mailings and special offers but want to unsubscribe, simply email information@informit.com.

Service Announcements

On rare occasions it is necessary to send out a strictly service related announcement. For instance, if our service is temporarily suspended for maintenance we might send users an email. Generally, users may not opt-out of these communications, though they can deactivate their account information. However, these communications are not promotional in nature.

Customer Service

We communicate with users on a regular basis to provide requested services and in regard to issues relating to their account we reply via email or phone in accordance with the users' wishes when a user submits their information through our Contact Us form.

Other Collection and Use of Information

Application and System Logs

Pearson automatically collects log data to help ensure the delivery, availability and security of this site. Log data may include technical information about how a user or visitor connected to this site, such as browser type, type of computer/device, operating system, internet service provider and IP address. We use this information for support purposes and to monitor the health of the site, identify problems, improve service, detect unauthorized access and fraudulent activity, prevent and respond to security incidents and appropriately scale computing resources.

Web Analytics

Pearson may use third party web trend analytical services, including Google Analytics, to collect visitor information, such as IP addresses, browser types, referring pages, pages visited and time spent on a particular site. While these analytical services collect and report information on an anonymous basis, they may use cookies to gather web trend information. The information gathered may enable Pearson (but not the third party web trend services) to link information with application and system log data. Pearson uses this information for system administration and to identify problems, improve service, detect unauthorized access and fraudulent activity, prevent and respond to security incidents, appropriately scale computing resources and otherwise support and deliver this site and its services.

Cookies and Related Technologies

This site uses cookies and similar technologies to personalize content, measure traffic patterns, control security, track use and access of information on this site, and provide interest-based messages and advertising. Users can manage and block the use of cookies through their browser. Disabling or blocking certain cookies may limit the functionality of this site.

Do Not Track

This site currently does not respond to Do Not Track signals.


Pearson uses appropriate physical, administrative and technical security measures to protect personal information from unauthorized access, use and disclosure.


This site is not directed to children under the age of 13.


Pearson may send or direct marketing communications to users, provided that

  • Pearson will not use personal information collected or processed as a K-12 school service provider for the purpose of directed or targeted advertising.
  • Such marketing is consistent with applicable law and Pearson's legal obligations.
  • Pearson will not knowingly direct or send marketing communications to an individual who has expressed a preference not to receive marketing.
  • Where required by applicable law, express or implied consent to marketing exists and has not been withdrawn.

Pearson may provide personal information to a third party service provider on a restricted basis to provide marketing solely on behalf of Pearson or an affiliate or customer for whom Pearson is a service provider. Marketing preferences may be changed at any time.

Correcting/Updating Personal Information

If a user's personally identifiable information changes (such as your postal address or email address), we provide a way to correct or update that user's personal data provided to us. This can be done on the Account page. If a user no longer desires our service and desires to delete his or her account, please contact us at customer-service@informit.com and we will process the deletion of a user's account.


Users can always make an informed choice as to whether they should proceed with certain services offered by InformIT. If you choose to remove yourself from our mailing list(s) simply visit the following page and uncheck any communication you no longer want to receive: www.informit.com/u.aspx.

Sale of Personal Information

Pearson does not rent or sell personal information in exchange for any payment of money.

While Pearson does not sell personal information, as defined in Nevada law, Nevada residents may email a request for no sale of their personal information to NevadaDesignatedRequest@pearson.com.

Supplemental Privacy Statement for California Residents

California residents should read our Supplemental privacy statement for California residents in conjunction with this Privacy Notice. The Supplemental privacy statement for California residents explains Pearson's commitment to comply with California law and applies to personal information of California residents collected in connection with this site and the Services.

Sharing and Disclosure

Pearson may disclose personal information, as follows:

  • As required by law.
  • With the consent of the individual (or their parent, if the individual is a minor)
  • In response to a subpoena, court order or legal process, to the extent permitted or required by law
  • To protect the security and safety of individuals, data, assets and systems, consistent with applicable law
  • In connection the sale, joint venture or other transfer of some or all of its company or assets, subject to the provisions of this Privacy Notice
  • To investigate or address actual or suspected fraud or other illegal activities
  • To exercise its legal rights, including enforcement of the Terms of Use for this site or another contract
  • To affiliated Pearson companies and other companies and organizations who perform work for Pearson and are obligated to protect the privacy of personal information consistent with this Privacy Notice
  • To a school, organization, company or government agency, where Pearson collects or processes the personal information in a school setting or on behalf of such organization, company or government agency.


This web site contains links to other sites. Please be aware that we are not responsible for the privacy practices of such other sites. We encourage our users to be aware when they leave our site and to read the privacy statements of each and every web site that collects Personal Information. This privacy statement applies solely to information collected by this web site.

Requests and Contact

Please contact us about this Privacy Notice or if you have any requests or questions relating to the privacy of your personal information.

Changes to this Privacy Notice

We may revise this Privacy Notice through an updated posting. We will identify the effective date of the revision in the posting. Often, updates are made to provide greater clarity or to comply with changes in regulatory requirements. If the updates involve material changes to the collection, protection, use or disclosure of Personal Information, Pearson will provide notice of the change through a conspicuous notice on this site or other appropriate way. Continued use of the site after the effective date of a posted revision evidences acceptance. Please contact us if you have questions or concerns about the Privacy Notice or any objection to any revisions.

Last Update: November 17, 2020