forums.silverfrost.com

Kenneth_Smith · Posted: Fri Sep 22, 2017 2:25 pm Post subject:

Not sure where to post this. I have been using the latest DLLs no 13, and playing about with the new native %pl with a view towards using it to update data plots as a simulation progresses. The following code reveals some issues.

Kenneth_Smith · Posted: Fri Sep 22, 2017 2:29 pm Post subject:

/cont

When complied with the 32 bit complier, strange things happen with the x and y scales.

Set Number of points to 5000, length = 1 and observe the y axis at the end of the run.
Set Number of points to 7000, length = 1 and observe the y axis at the end of the run. Different even although the extents of the y axis data has not changed.
Set Number of points to 5000, length = 10 and observe the x axis data at the end of the run. Note the value display at X=100.

These only occur when compiled with the 32 bit compiler - 64 bit is fine, BUT without the

Kenneth_Smith · Posted: Fri Sep 22, 2017 2:33 pm Post subject:

/cont

Some pictures:-

mecej4 · Joined: 31 Oct 2006 Posts: 1899

Kenneth_Smith · Posted: Sat Sep 23, 2017 12:01 am Post subject:

John,

I forgot to mention the flicker, which is slightly irritating.

I guess Paul will read this in despair and remind us that users will always find faults, and we are all trying to plot different data over widely different ranges, but clearly the algorithm for selecting the scales needs improved. I know from my own experience that automatic selection of scales and format codes is difficult - caused me a lot of grief over the years with my own plotting routines - there's always some data that breaks the algorithm and I end up looking at ******!

These loci, which have fascinated me for many years are Euler Spirals.

Ken

PaulLaidler · Posted: Sat Sep 23, 2017 8:22 am Post subject:

By today's standards, the original 32 bit %pl does not produce polished results and my aim was to port it to 64 bits. I am limited by a) starting from an existing interface and b) other demands on my time.

I am pleased that the native %pl is being used and that it often produces results that can be published in professional journals. At the same time I have to avoid the "black hole" of continual improvements.

Initially I did not contemplate log-log scales with data from a random number generator with values out of range but the native %pl now has a reasonable go at it.

Making the tick marks and values always right for various scalings and fonts is unfortunately beyond my capacity in the time at my disposal. So for some data and some purposes the native %pl may not be good enough.

Having said all of that I am grateful to Ken and mecej4 for reporting bugs that should be easy to fix.

Kenneth_Smith · Posted: Sat Sep 23, 2017 1:41 pm Post subject:

Paul, the new %pl is a great addition. I am beginning to see more ways to use it in interactive programs and it removes one on the primary reasons I have a tendency towards procrastination. Hopefully the forum collective will not identify too many bugs but post some good working examples.

Thanks for your efforts. It is appreciated.

Ken

PaulLaidler · Posted: Mon Sep 25, 2017 8:44 am Post subject:

John

The maximum number of plots on one graph is 10, otherwise you should get an error report. There is no programmed limit on the number of points that you can plot, memory is allocated on demand.

I don't have a quick answer to your 2nd question.

PaulLaidler · Posted: Tue Sep 26, 2017 7:05 am Post subject:

In one sense yes, it's just a parameter in the library. But it is a static value that means that the base memory for all calls to %pl would be increased. No doubt the model can be changed but that would require a redesign and off hand I don't know how much work that would entail.

Kenneth_Smith · Posted: Tue Sep 26, 2017 11:30 am Post subject:

I have to agree with John, the maximum I have used so far is N_GRAPHS=9, and I was blissfully unaware of this limitation, until now.

To me the beauty of the new %pl is the ability to display data without having to resort to exporting text files for import to excel (or equivalent) which is all time consuming, particularly where the user needs to view the plots, before the next step of an ongoing calculation. The alternative is to plot the data using %gr - which is equally time consuming to get right first time - and would be a backwards step considering all the functionality that's been built into the native %pl.

One for the wish-list which hopefully Paul will start compiling and work through - time permitting.

PaulLaidler · Posted: Tue Sep 26, 2017 12:02 pm Post subject:

Ken

I have added the option [external_ticks]. This implies [framed] and puts the tick marks externally to the frame.

Kenneth_Smith · Posted: Tue Sep 26, 2017 1:00 pm Post subject:

Thanks Paul, I knew you had a wish list ! Wink

Look forward to the next dll coming out.

John, One way to overcome the N_Graphs=10 limit, might be to use the %pl call back function to overlay additional graphs, in which case the GET_PLOT_POINT@ subroutine would provide all the required geometric data. Just a thought. I will try this when I have time.

Kenneth_Smith · Posted: Wed Sep 27, 2017 10:49 pm Post subject:

Paul, Another one for you to look at I am afraid.

The following code plots two points on the x-y plane. Note that there is a spin wheel on one of the Y values. Running the program and increasing this Y value incrementally reveals some issues.

When compiled under 32 bit, for the following Y values the corresponding point is not plotted 4.5, 9.0, 11.5, 14.5, 15.5, 18.0, 18.5, 20.5, 23.0 etc.

When compiled under 64 bit, this still happens, but it first occurs for a Y value of 48.5, then 80.5, 83.5 etc.

I am using the most recent DLLS (13).

Kenneth_Smith · Posted: Thu Sep 28, 2017 12:49 am Post subject:

John, this might help.

DanRRight · Posted: Wed Oct 04, 2017 7:36 pm Post subject:

I missed this post. You've got a lot of progress!

I am surprised that Mecej4, John Campbell, Eddie and other power users completely missed %pl. The progress to make it much more functional would go faster then. You guys don't plot anything during the run? And don't need debugging sometimes your data? The plotting a graph and look at data would take one single line, almost as short as with the write(*,*) variable (i,j), i=1,n),j=1,m) when you debug with the help of %pl.