Sign Hawaii Banking IOU Computer

Industry sign banking hawaii iou computer

hello and welcome to CMG's first webinar in the 2018 series today's topic is efficient simulation leveraging CMG's latest performance enhancement technologies my name is Anjali Kumar and I'm the vice president of engineering solutions at marketing lab CMG I'll be the moderator for today's session I am pleased to introduce Hawaii duong Account Manager at C and D will be presenting today's webinar quick you've been with CMD for a little over 10 years right that's right so the first four at five years how he worked as a reservoir engineer and worked on simulation of advanced recovery processes especially specialized in chemical you are and optimization of field scale chemical your processes subsequently Co he moved into the sales role and he is one of the account managers for the Canada region who holds a dual degree in a BSC in biochemistry and an engineering degree in oil and gas from University of Calgary so at this time I like to inform you that there will be a question and answer session at the end of the presentation however you may submit your questions at any time using the chat window on your screen will also pose the recorded version of the presentation on our website within the next 24 hours you can use your cm GL login using your username and a password to access the material so what do you coin Thank You Angie welcome everyone once again welcome to our very first webinar of 2018 today I will discuss the new performance enhancement technologies that can be found in our latest general release and how you can leverage these to improve the efficiency of your simulation workflow so for the course of the last 40 years CMG has always been the leaders in technologies and innovations and we were able to do this by either pioneering new technologies or were the first to incorporate new technologies into our simulators we were the first to incorporate adaptive implicit methods into our simulator we were the first to use QN SS which is an improved version of the multi-phase flash'd solution that's applied to the matrix over we were one of the first to paralyze our simulators over multiple cores using a parasol solver and this has made us the fastest and still is the fastest commercial simulator out there we are the only one to offer dynamic grading which is the ability to change your good size as it simulation progresses about 12 years ago we have to do CMOS and CMOS AI which was and still is the most comprehensive software to assist you with yours sense analysis history matching uncertainty analysis and optimization workflow a couple of years after that we introduced the unconventional modeling workflow which uses tartan grading to properly model your hydraulic factory this works in conjunction with CMOS to provide an automatic workflow to help you better instead understand your unconventional resources as well to optimize your fracture designs couple of years ago we've automated our numerical tuning workflow once again using CMOS this allows us to use the CMOS optimization engine to help you to obtain the most often numerical parameters for your simulation run the last three here are what we'll talk into detail today are the first one we auto-tune AI which is a one word one-step numerical tuning to give you the best optimal numerical parameters the second one is a new common observer this solver improves the performance of our linear solver unproblematic nonlinear problems and finally it's our MPI and MPI allows us to paralyze our models across multiple CPUs as well as cost multiple nodes so because of our ongoing R&D investment every year including this year we have provided performance enhancement technologies along the entire workflow is for this year we've improved builder we've improved the load time and the interactions of the models with our users for the simulators as mentioned before we improve the runtime by having new linear solver and paralyzing technologies as well as the automatic tuning for faster runtime now all these three work in conjunction to give you a multiplicative speed-up in your run time and on the back end with results we improved the load time as well as the file size so that you can analyze these data much more efficiently now not only that we improve the workflow with the introduction of CMOS AI we now can capture improvements from the beginning to the end of your project okay CMOS AI incorporates machine learning and artificial intelligence and provides you with additional performance and efficiency in all your workflows so let's take a look first at our pre-processing improvement the one trend that we start seeing a lot now is that models are getting bigger and more complex well builder works really great with typical models with these big models you require more memory and we tend to see a slowdown in the loading time so with these large models and with new builder we improve the memory allocations so I said grading algorithm algorithm to significantly reduce the loading time so on this chart he can see the loading time of our latest version thousand seventeen is much much quicker than what is seen from two years ago even from last year we managed to improve in it by at least ten to fifteen percent now with this improvement in loading time we've also improved the 3d visualization performance of our model so these large models now will rotate a lot smoother you can access different sessions are much easier and you can update the data and the primers much quicker another software we added to help with the pre-processing is C edit now those of you that you see most in the past you might recognize C edit as an editor that you can valium C moles for this year we've taken C at it out as a standalone product C edit now we'll work with both C Mo's as well as your standard text files as well as CMG datasets okay so quick glance at C edit it might look like any other text editors out there but we've added a few features that we think will significantly improve your work with mcedit so first we've changed our layout to include very simple very easy to read icons and with these two icons we introduced a one-click navigation so by clicking on any one of these headings C yet it will take you to the appropriate and corresponding section of the model so for example I want to see the numerical section I click on numerical methods control take you right to that section of your model another improvement we've add is the highlighting of the syntax so when you open a CMC data set in C edit it'll highlight your key words blue leave all the corresponding data or numerical data black right and any comments shown up a screen so with this you can quickly look at your model and you can see what is what what you can change and what you cannot change another very useful feature we've add is the ability to collapse or expand an array so anyone have used a text editor to edit a dataset you know that scrolling through this massive and large array is tedious and very frustrating so we've had the ability to able to collapse and expand this array with this one little button here so with an array collapse you can quickly switch from one section of the model to another we'll having to scroll through thousands of lines of numbers or codes so one last feature we've add our modification markers okay these are the yellow bars show up on the left hand side of your text file and it indicates an area where information has been added or revised since you last saved this model so all these features combined to make it more efficient much better organized but more user friendly so that you are spending less time generating data set and spending more time during the actual simulation itself now speaking of simulation work if you've done a simulation where you've must have heard of numeric or tuning so while the CMG linear solver has been involved to work very well to be very efficient and very fast with the majority of the problems occasionally we do run into models that are numeric complex that might need some numerical tuning and we do the micro tuning on these models for three primary reasons one is to increase the accuracy by maintaining a low material balance area the second is to improve stability so better conversions and lower solver failures and finally and perhaps the most important the reason we numeric or two model is to improve the speed up to make it run faster now anyone that's done numerical tune in the past you know that there is a negative correlation between the speed up or the speed of your run versus the two per the other two parameters for example we look at material balance error you can see that the lower material balance error tends to have a longer run time and vice-versa the shortest one time tend to have a higher material balance Erica so when you do medical tuning and try to pick the optimal solution you will pick something with a compromise in between these two and that solution is shown in red here and typical eyes on the Pareto front on a plot like this historically to hit a pot like this you have done it in two methods one is manually so manually changing your parameters run it get your results back and change grammars again and repeat the process a couple of years ago we've automated process by using CMOS okay so we now can take advantage of the CMOS experimental designs as well as they see most optimizing engine to get you the same result in much quicker time so the CMOS workflows work extremely well and extremely robust but for this year we made it even better and we made it better by the introduction of auto tune AI auto tune AI is an artificial intelligent base adaptive time step in control using only one keyword auto tune a I will apply the optimal numerical settings that we beneficial to your specific models and processes right using this one keyword stars and gem will run anywhere from 2 to 12 times faster in our latest version so how easy is it to use autotune AI I'll show you a quick demo on how we can add or activate Auto Tune AI in either stars or gem so here is the right launcher interface I take my stars model I drag it in to see edit there's a within C edit as mentioned before I can go to my navigation pane and with one click navigation I quickly access my numerical section I type in auto-tune save my data set and now when I drag and drop the star dataset into stars auto-tune is active and now our numerical parameters for this one will always be optimized as the worm progresses in Gem is very similar I take my gem dataset and I open in C edit you once again under navigation I go to my numerical method section so the key word in gem is slightly different the keyword in gem is ad TSC and it's a gem equivalent of auto-tune keyword in stars save my file close it out now once I drag my date my gem data set into gem auto-tune or a DTSC in gem is now activated and it's rerun with the optimal numerical parameters throughout the run so how do these look like in actual results so they're CMG we conservatively estimate that you will see a two to three times speed up by using this feature in thousand seventeen now you can see here some of the testimonials from the numbers we get back from our clients that the speed-up is very case-specific depends on what you're doing it depends on your model these depend on your processes and it depends on your regular default numerical settings right but for gem we see anywhere from were three times up to a 17 times speed-up in stars similarly we see anywhere from the two times up to a 16 times speed up so all these can be found on our website and I encourage you to visit it check out our the testimonials and leave feedbacks if you have any for us as far as what you think auto-tune is and if is helping you some further testimonials comet ridge use a DTSC in the gem simulation and saw a three times speed up on the star sigh oxy Oman used its auto-tune in their steam flood model and they saw a two time increase in their speed up so once again visit our website check out all the testimonials leave us some feedback if you have any thanks Chloe have this time I like to pose a question to the audience the questions will appear on your screen pretty soon and the question is have you used auto tune or a DTSC on your models and you can respond by clicking either yes or no on your screen will wait for you seconds here to be able to see the responses and because it's a relatively new feature we have launched it around and December and we announced it a lot of our customers have started using it since then and have shared with us their test results and their benefits and speed ups that they have obtained as Koya mentioned the speed ups are subject to the models that you are seeing so anywhere between 2 X to 17 X is what people are reporting it also depends on how well your model is originally tuned some of these models were originally tuned using C most are either manually using experience so auto-tuned did help but the speed ups were not that significant but still it's better than spending time manually tuning your complex simulation models and since it's a one keyword solution try it and give us feedback we are continuously improving this feature as it was recently launched so any feedback that you may have will be really appreciated and well received by our development group so it looks like the poles are complete as expected since it's a new feature not a lot of people have tried it 80 / 85 percent of the respondents say no they haven't tried it 15 percent of the people attending today have tried it so a good that people have started to try it we are really really excited about this feature and I hope that it will help improve your productivity and get the maximum benefit from your simulation efforts so over to you quite thank you our journey so once again feel free to leave us some feedback on what you think about it soon AI so the next improvement I want to talk about is our cognitive solver and this can be found in all through your simulator SIMEX gem and stars right so our linear solver has been evolved all over the years to become very efficient and become the fastest simulation linear solver out there in the industry however once in a while we do run into problematic data set that might cause an issue with an inner solver so looking at a very simplified diagram on the right here so only outer edge we have our time step control right within dots hands up control we have a number of new iterations typically 10 to 15 and within those individual new narrations we have a matrix solver iteration or linear solver iteration so in these problematic data set or nonlinear data set we see a block down or a bottleneck within this matrix solver all right so there's a part on this matrix offer it's hardly to converge and if that takes longer to conversion it takes more new iterations to converge and that negatively affects your time step control so the new combinative solver improves a similar performance by improving the efficiency of the linear solver so we're improving the efficiency right at the core so now your linear solver is much more efficient which in turn it's converge much quicker which means your new less new iterations is needed and therefore you have a better time step control for your run so how exactly does this work so our current iou solver or parasol solver once again works extremely well very efficient for the majority of the issue works by taking the magic Oban matrix and under and put it through a series of iterations now in those problematic data sets we see a slowdown of this iteration and a slowdown typically typically is because of our pressure matrix now pressure is always a primary very variable so it's always being solve X implicitly so if pressure becomes an issue this slows down this iteration loop takes longer to converge and in turn takes more newton iterations and in turn negatively affect your time step control so with the new combination solver we have a slightly alter workflow than before so we do is that we extract the pressure matrix out of the Jacobian matrix we precondition this pressure matrix and then we use this new CMC solver as part of the second iteration loop to solve the pressure matrix once the special matrix is solve or very closely resolved it's reinserted ba k to the Jacobian matrix and the Jacobian matrix once again undergoes the same ILU iteration loop the only difference here is that with the pressure matrix almost solve or solve this iteration loop will converge much quicker and in turn we reduces amount of new iterations that's needed and in turn give you a better time step control over all this give you a very good speed up on your run time so how does this look with actual data so a comment sovereign works extremely well in do continue model so tight oil unconventional oil and gas or in processes with a rapidly moving pressure front such add water flood or seg knee or sieve flood that has mobile water alright so looking at the first three cases on the left here we see a modest increase around 1.3 time in our speed up for the water flood case we saw a really great improvement I and reducing a run time by less than half so once again water flood has a rapidly moving pressure front so this is one example where combinative solver works extremely well to reduce your run time the Esad cases first specifically this one case and the reason we saw we did not see a much any improvement on the run is because the original CS acne cases road was already running extremely well and extremely fast so using Combinator solver in this one particular case saw really no difference in your run time the sekki case here with the Flex well number three so a significant reduction in time and the reason for that is that this safety case has mobile water once again a great candidate for using combinative solver our last moderator tight oil saw 1.5 time increase in speed up and for that case I'll go into a bit more detail is the real Bakken reservoir very large model of 6.2 million active grid blocks do permeability 37 different wells with multiple hydraulic fracturing stages on each of the well so the original run time for this case is 198 minutes using common native solver now reduce your runtime to 135 minutes so we saw a one and a half time increase in your speed without having to use any additional software or additional hardware so once again this is a brand new feature please try it out and send us any feedback any comments that you have our next improvement once again can be found in all three of a similar SIMEX gem and stars it's our new parallelization technology so before I go into detail on that let's take a look at a layout of a typical compute node so the very bottom you have your compute which is your self-sustaining standalone unit that has its own power supply his own input and output is an operating system right now within that computer note you have multiple CPUs or central processing units these are your brains behind the actual computer each CPU has his own memory bank and each CPU can communicate with each other all right now within is each CPU we have cores corsa individual threads and these core can either act independently or in conjunction with other cores within the same chief CPU and within the same compute node so in paralyzation in the simplest form is dividing a large model into smaller sub models that can be solved simultaneously across multiple cores so here we take a very large model we break them into smaller sub models that can be solved on the individual cores and you can see where the advantage lie the eventual line to kind of things one is that these smaller models are much more simpler to solve so they can be solve a lot quicker and secondly these sub models are solved concurrently and simultaneously adding additional improvement in your runtime so over the course of the years seem Chi continue to evolve our paralyzes and technologies to take advantage of current as well as future hardware configurations so there's two primary methods to paralyze our models over multiple cores or CPUs one for shared memory we use OpenMP is anything that occurs within a CPU that's sharing a memory bank and the second which is brand-new is hybrid OpenMP which is used for the student memory which is across multiple CPUs or multiple compute node so let's take a look into detail onto the shared memory method so OpenMP here improves your performance within a CPU or memory bank so we look at this model right this model sits on the one CPU within one compute node so all these cores while separated I still share a memory bank so by sharing memory bank open MP allows better communications quick communication between these sub models so these are models can be run and recombine very efficiently to show very significant one time in distribute memory we're not looking at only sharing memory banks baloon get sharing across multiple CPU which would have separate memory banks and on top of that we have multiple computer nodes which have their own memory and their own input and output so in this case what we see our memory barriers so now the bottleneck is not the actual calculation or the simulation of the sub models the barik becomes the communication between these summer rolls across multiple CPUs as well as multiple compute nodes so we solve that by using a hybrid OpenMP MPI technique now this improves performance with thinner core using OpenMP as well as between different CPUs and different nodes so MPI allows for better communication between CPUs which have multiple of it has different memory banks as well as better communication between compute node which are two separate computer connected via network so looking at some of the runtime comparison for MPI and OpenMP so for the first two now we saw no difference in one time and the reason for that is because these are across the same CPU so we simply say memory bank so when you share a memory bank openmp and I wrote hybrid model gives you no difference in runtime when we start seeing a difference in run time is when we're spreading it across multiple memory banks okay and this can be seen in a sixteen and thirty core example so while he sits on the same note they're using two different CPUs so their requirement there's a requirement for communication between two different memory banks so this where the benefit of MPI or the hybrid openmp Empire models can be seen in multiple nodes in two more nodes there there is really no reason to run only openmp since you now have to communicate not only cross multiple CPUs or you're communicating cost multiple nodes so for those cases we've ran strictly on openmp and MPI and with these we saw really good speed-up going up to 256 cores and we'll continue testing and testing and we will present you with even higher core counts down the road so some benefits of paralyzation okay so paralysis allows you to improve your time and your memory limits to run a certain size model you now not limit to what one computer can do you can spread over multiple computers with OpenMP and MPI hybrid model we can simulate a lot faster on those with higher socket count so these are the one with multiple CPUs ours our implementation does not sacrifi cracy so your data remains or your results remain consistent whether you run on one node or three different nodes I one big benefit is that now you can combine multiple existing computers so you have a bunch of HCPs lying around you combine to make 32 or 64 and not having to buy brand new computers another reason paralyzation is beneficial is that now we can spread this across the cloud whether it's cg cloud or take advantage of Microsoft Azure or AWS we now can spread our run over hundreds or thousands of cores further improving your work and your efficiency the next feature I want to discuss is Dinah great so as mentioned Dinah great is unique to CMG and is unique for frontal processes what Dinah grid does is a dynamically change your grip box size through all your simulation one to ensure that you have speed up or maintain consistent and accurate results so this here is a very typical cross-section of a safety model we have a steam chamber and anyone that's run simulation we know that we want very small grid blocks so that we can perfectly capture all the mechanisms and all the behaviors of our models the downside of having small grid box is that you have a lot of great blocks when your model and a large hundred number of grid blocks ultimately leads to a very long runtime so how do we solve this we solve this by changing our model from this to something like this so Dyna grid will allow you to amalgamate or combine the small grid box into larger group blocks or vice versa allow you to the amalgamate or refine larger grid box into smaller groupings so we know that in a simulation like this I'm a lot of grit box on the outer edge are not changing they'll have the same temperature the same pressure same property so Dyna grid will use that to combine them into larger grid blocks to reduce the number of great blocks you need similarly inside the steam chamber after has been developed the inside a steam chamber can be fairly homogeneous and since the all these great box are very similar in properties we could combine them into larger grid blocks what's important is that we kept the grid box at the interface where all the cool stuff happened this is where we had to calculate a mass transfer or heat transfer in those areas our grid box which means small so that we can perfectly and more accurately capture those mechanisms so this this here will give you an optimal combination of speed without sacrificing any accuracy so how this diner grid perform in actual field models so in combustion we see about a five to seven time increase in speed up in sag knee and vein packs we see about a two or three times increase in speed up in cyclic steam once again about a seven times speed up by using dynamic grid alright so here it is a pull from one of our client here in cm in Calgary so next in CNOOC saw a reduction in runtime by 95 percent for the hybrid steam and in-situ combustion by using Dyna quick so a significant crease or sorry only slip can decrease in runtime by using this one feature without sacrificing any of your results so the next thing I want to talk about is a post-processing improvement so for this we've improved it in two ways first we've reduced the file size with a new SS reformat and with this reduction in file size you now can improve the loading time and your interaction with the new models so anyone have used our products in the past we know that we used to have two separate software for post-processing we used to have graph for your 2d or graph plotting and we have the 3d version for looking at the models in three dimensional and both of these use our old method are all format sr2 in thousand 17 we combine these into one software so both 2d and 3d is now in one tool and this is now used the new sr3 format so this is much more intuitive much more user-friendly and much more efficient in analyzing your data so speaking of the reduce file size how do they actually how much reduction do we actually see between the SR 2 and SR 3 so here's a chart of some of the examples that we we've run and you can see that the reduction is fairly significant on a more modest case we saw around a 50% reduction in size however in some cases we saw up to a 90% reduction in time so our our srt file some time is only about 10% of what osr2 was of course with this reduction in file size we see a corresponding reduction in load time so for this chart here we've normalized the load time of the SR 2 to 1 and as you can see the load time in SL 3 in all cases have been reduced significantly or in some case down to about one sixth of it was before so once again with this reduction you spend less time loading spend less time worrying about it loading up and allow more time for you to do the actual analysis to look at your data and see what you can learn out of so one last thing I do want to talk about is CMOS III now I can talk a lot about Somoza I do too time for this one but we do have our own standalone CMOS I webinar in March so keep your eyes out for invitation and encourage you to attend that session as you will learn a lot of new features that seem else AI will have to help you with your workflow but for CMOS AI c will say i will leverage artificial intelligence and machine learning along with human expertise to identify the optimal solutions so with AI and with machine learning we now can create accurate data-driven proxies that allow you for quick and interactive what-if analysis on top of that we improve the work fill out the workflow to be much more intuitive and new to help you improve your decision making abilities so here's an example of some a new CMOS AI feature this here is our dashboard interface or - us or our proxy dashboard or proxy interface so hiss for this example is unconventional model so you can see some of the parameters were changing including your fracture parameters your half-length your processor your perm endpoints of your realm curves so on the plots what we see are the blue lines or your historical your field data the black line is your base case and you might not see it here but you can see a bit of a reddish tinge and that red is actually your simulation run your optimal case so when I play this video you can see that as you change some of these parameters your predicted value or your printed results change along with it in real time interactively so this gives you the ability to quickly predict new results by just changing some of these parameters right and this prediction come from the proxy model and this proxy model is comes constantly changing constantly getting updated as new data being found so once you've picked a set of primers you like you can do two things you can add it to the experiment right to continue updating your proxy model or you can actually run in our standalone and compare the results you see to the models predicted okay so new CMOS also included a bunch of new automated results that will help look quickly and I said data one of the more useful one is a stack bar plot and in this case we plotted it global history matching versus all different runs that we've run so with each global history matching error we can see how much of that is made up of our individual global or local objective function right from there you can quickly isolate and see where the real issue is and focus your simulation and your stay on that particular section another thing we do have is a proxy analysis section model here you can look at different models different analysis such as the solo and morris analysis and see how your proxy models behaves or how they are in a certain way you can look at different proxy settings and you can even export your proxy model out to excel for you to work with your own software or to do your own analysis so in conclusions are any investments that provide us and allow us to continually add performance enhancement technologies along the entire simulation workflow we improve our pre-processing with builder and see edit faster data set creation much better interactions with the users for the simulation itself we have improved auto tune with the auto tune AI faster runtime with one simple keyword a new comedy solver improve runtime for problematic nonlinear problems our hybrid OpenMP marathon methodology allow you to paralyze your models across multiple cores multiple nodes and allows you access to hundreds and thousands of roads that are out there in the cloud so these three work in conjunction once again and give you a combined speed-up of up to 12 times in some cases on the post-processing end we improve load time and usability with our new results and finally CMOS III and compress the entire project improve run time improve productivity and improve your disturbia to make correct and the right decision thank you coin we will now take up questions if you have any questions and have already not submitted it you can do so now using the chat window on your screen I see there are already several questions on the screen that I can start taking it one by one and the first question is from rompere cha and the question is can combinative solver be used with a black oil simulator that is IMAX yes definitely combinative solver is available in all the three simulators IMX gem and stars so you can definitely use it in with IMAX the next question is from young one and the question i cannot a team be used with the do mechanics for do mechanics speed-up definitely auto-tune is is a simulator keyword so any of the functionalities that come with a standard simulator with whether it's geomechanics whether it's any other feature it's all can be used together as well as combinative and auto tune can be used together to get a combined or a compounded benefit from each of those two features so take the next question again the next question is from RAM and the question is can 2017 results so vide old SR two files yes in the 2017 version we actually allowed results to be both sa-2 and sa-3 version of the of the results from 2018 version results will only read the SRT format and CMG will be making that 2017 and prior version is a two format obsolete the next question is from Arthur and the is speed up is always good but could you comment on the influence of auto-tune and community solar on the results of the simulator should be expected results to change at all if yes how much the results change generally the one thing that I'd like to highlight is CMD doesn't like to compromise on results or accuracy for speed all of these runs that high showed to you in terms of speed ups the results were absolutely identical to without using the auto-tune or the combinative solver so there is no compromise in the accuracy and results when you use any of these features in in our simulators the next question is from Hashem and the question is do I need to use all pre-processing performance improvement on all the I scrolled on me I'll reread the question here is do I need to use all pre-processing performance improvement for all types of models or any one is suitable for special cases again as koa mentioned these performance improvement tools are very very case-specific it depends on how complex your reservoir grill is how complex your geology or property distribution is what kind of you are or any other recovery process you're trying to to model how well the original model runs and again as some of the cases that people have reported and it's on our website so do check it out there's a wide variety of processes and wide variety of complex to simple reservoir models that people are using and seeing the benefit I must caution that there are a certain number of cases which do not see improvement from otta tune the reason being is that the base model itself run pretty fast the other is that the process itself is not suited for the auto tune algorithm to give you a speed-up but fortunately those kind of cases are relative small I will say less than 5% so do give it a try on your model and if you see speed ups that is really really great if it's not giving you the desired speed else then do get back to us and we will appreciate your feedback and work with you on improving this feature so that you can get the maximum benefit out of of these features so again auto-tuned cognitive all of these features or performance enhancement technologies can be used in combination to get the compounded benefit from your simulation runs ok the next question is from Neha bunsall and the question is any plans to extend auto tune to IMAX actually we are actively working on implementing auto tune into a mix there is a prototype already available if you are interested we can have it shipped to you to give it a try it will be available in the 2018 general release that is expected to come out sometime in some of this year so yes auto tune will be available in IMAX very soon the next question is from Hashem again and the question is are all of these new modifications available in 2016 or 2017 version all of these are features that koi talked about is available in 2017 version of our simulators except a few actually Dyna grid has been available in our simulators for over 10 years now or maybe 15 and over the years we have improved its capabilities quite a bit so give it a try and and see how it performs on your on your models rest of the features like auto tune combinative and mpi our 2017 these features I think I have time to take on one more questions there are already several questions here so do not feel left out we will individually approach each one of you with answer to your questions and so the last question again and I think I already took that one is can auto tune and community be used together and the answer is definitely yes and you can expect a compounded benefit from each of these features so there's few more questions that we will not be able to take it for the sake of time but we will definitely approach each one of you individually with our response and the next 24 hours so I hope you found the latest performance enhancement technologies from CMG exciting we are really excited to bring it to you so I hope you will derive the maximum benefit from some of these simulation software capabilities as well as if you haven't tried it do give it a try and provide us feedback we really really appreciate your feedback check out the customer testimonials and customer results on our website we showed you the link quite a few slides back we will also post the today's presentation on our website within the next 24 hours so you can use your cm GL username and a password to access the material in addition if you have any other questions that wasn't answered or if you would like to know more about our product offerings on any of the features that was discussed today do not hesitate to email us at sales at CMC o dot CA we also have several training classes that we offer on a variety of subjects if you want to attend any of your any of these courses and refresh your skillsets on some of these modeling capabilities and use of our simulators then do check it out on our website at CL Garcia slash training and the registration is online it's fairly straightforward again thank you all for joining us for the webinar today on efficient simulation leveraging CMG's performance enhancement technologies and we do hope to hear on your experience and feedback on these technologies thank you again