Data Analysis Basics¶
Data analysis in Enzo can be complicated. There are excellent premade packages available for doing Enzo data analysis (see SupportingCodes.). However, it is likely that your data analysis needs will grow beyond these tools.
Enzo reads in initial conditions files and outputs simulation data using the HDF5 structured data format (created and maintained by the NCSA HDF group). Though this format takes a bit more effort to code than pure C/C++ binary output, we find that the advantages are worth it. Unlike raw binary, HDF5 is completely machine-portable and the HDF5 library takes care of error checking. There are many useful standalone utilities included in the HDF5 package that allow a user to examine the contents and structure of a dataset. In addition, there are several visualization and data analysis packages that are HDF5-compatible. See the page on Data Vizualization for more information about this. The NCSA HDF group has an excellent tutorial on working with HDF5.
Note that as of the Enzo 2.0 code release, Enzo still supports reading the HDF4 data format, but not writing to it. We strongly suggest that new users completely avoid this and use the HDF5 version instead. Enzo’s parallel IO only works with HDF5, and we are encouraging users migrate as soon as is feasible.
Using YT to Analyze Data¶
If you have installed YT along with Enzo (as per the build instructions?), you should be able to use it to find halos, examine profiles, prepare plots and handle data directly via physically meaningful objects. Documentation, a wiki and a mailing list are available for support and assistance with installation and usage.
Analysis with VisIt¶
Another tool that has a native reader for Enzo data is VisIt, a parallel VTK-based visualization and analysis tool.
From the VisIt Users website:
VisIt is a free interactive parallel visualization and graphical analysis tool for viewing scientific data on Unix and PC platforms. Users can quickly generate visualizations from their data, animate them through time, manipulate them, and save the resulting images for presentations. VisIt contains a rich set of visualization features so that you can view your data in a variety of ways. It can be used to visualize scalar and vector fields defined on two- and three-dimensional (2D and 3D) structured and unstructured meshes. VisIt was designed to handle very large data set sizes in the tera- to peta-scale range and yet can also handle small data sets in the kilobyte range.
The caveat is that as of version 1.11.2, VisIt only understands the original unpacked AMR format. However, the packed-AMR is in the VisIt development version, and will be included in the next release (1.12). If would like this functionality sooner, it’s not too much work. Here’s how to begin:
- Download the following:
- Untar the source tar file,
- replace the two files named avtEnzo* in visit1.11.2/src/databases/Enzo/ with the ones you’ve just downloaded, and
- retar the file, keeping the same directory structure.
(You can do this without untarring and retarring, but this is a bit clearer for those not familiar with tar.) From this point, you can build and install VisIt using the build_visit script. When you do this, remember to do two things:
- Use the TARBALL option to specify the tar file for the script to unpack. Failing to do this will cause the script to download a new tar file, without the changes that you need.
- Select both HDF5 and HDF4 as optional third-party libraries. This may not strictly be necessary, if you already have HDF5 and HDF4 installed on your system, but the script isn’t clear on how to specify which HDF5 installation to use. (HDF4 needs to be available to satisfy a dependency check for building the Enzo reader. We’ll ask to have this updated in future versions of VisIt.)
Writing your own tools, I - the Enzo Grid Hierarchy¶
Enzo outputs each individual adaptive mesh block as its own grid file. Each of these files is completely self-contained, and has information about all of the grid cells that are within that volume of space. Information on the size and spatial location of a given grid file can be obtained from the hierarchy file, which has the file extension ”.hierarchy”. This ascii file has a listing for each grid that looks something like this:
Grid = 26 GridRank = 3 GridDimension = 34 22 28 GridStartIndex = 3 3 3 GridEndIndex = 30 18 24 GridLeftEdge = 0.5 0.28125 0.078125 GridRightEdge = 0.71875 0.40625 0.25 Time = 101.45392321467 SubgridsAreStatic = 0 NumberOfBaryonFields = 5 FieldType = 0 1 4 5 6 BaryonFileName = RedshiftOutput0011.grid0026 CourantSafetyNumber = 0.600000 PPMFlatteningParameter = 0 PPMDiffusionParameter = 0 PPMSteepeningParameter = 0 NumberOfParticles = 804 ParticleFileName = RedshiftOutput0011.grid0026 GravityBoundaryType = 2 Pointer: Grid->NextGridThisLevel = 27
GridRank gives the dimensionality of the grid (this one is 3D), GridDimension gives the grid size in grid cells, including ghost zones. GridStartIndex and GridEndIndex give the starting and ending indices of the non-ghost zone cells, respectively. The total size of the baryon datasets in each grid along dimension i is (1+ GridEndIndex[i] - GridStartIndex[i]). GridLeftEdge and GridRightEdge give the physical edges of the grids (without ghost zones) in each dimension. NumberOfParticles gives the number of dark matter particles (and/or star particles, for simulations containing star particles) in a given grid. Note that when there are multiple grids covering a given region of space at various levels of resolution, particles are stored in the most highly refined grid. BaryonFileName is the name of the actual grid file, and should be the same as ParticleFileName. Time is the simulation time, and should be the same as InitialTime in the parameter file for the same data dump. The other parameters for each entry are more advanced and probably not relevant for simple data analysis.
Possibly the greatest source of potential confusion in Enzo’s datasets is the overlap of grid cells. In a simulation, when a given grid is further refined, the coarse cells which have not been refined are still kept. The solution to the hydro and gravity equations are still calculated on that level, but are updated with information from more highly refined levels. What this is means is that a volume of space which has been refined beyond the root grid is covered by multiple grid patches at different levels of resolution. Typically, when doing analysis you only want the most highly refined information for a given region of space (or the most highly refined up to a certain level) so that you don’t double-count (or worse) the gas in a given cell. Look at this example analysis code.
Writing your own tools, II - Enzo Physical Units¶
Yet another significant source of confusion is the units that Enzo uses. When doing a cosmology simulation, the code uses a set of units that make most quantities on the order of unity (in principle). The Enzo manual section on the code output format? explains how to convert code units to cgs units. However, there are some subtleties:
- Density fields
- All density fields are in the units described in the AMR guide except electron density. Electron density is only output when MultiSpecies is turned on, and in order to convert the electron density to cgs it must be multiplied by the code density conversion factor and then (m:sub:e/m:sub:p), where m:sub:eand m:sub:pare the electron and proton rest masses (making electron density units different from the other fields by a factor of m:sub:e/m:sub:p). The reason this is done is so that in the code the electron density can be computed directly from the abundances of the ionized species.
- Energy fields
- There are two possible energy fields that appear in the code - Gas energy and total energy. Both are in units of specific energy, ie, energy per unit mass. When Zeus hydro is being used (HydroMethod = 2, there should be only one energy field - “total energy”. This is a misnomer - the Zeus hydro method only follows the specific internal (ie, thermal) energy of the gas explicitly. When the total energy is needed, it is calculated from the velocities. When PPM is used (HydroMethod = 0) the number of energy fields depends on whether or not DualEnergyFormalism is turned on or off. If it is ON (1), there is a “gas energy” field and a “total energy” field, where “gas energy” is the specific internal energy and “total energy” is “gas energy” plus the specific kinetic energy of the gas in that cell. If DualEnergyFormalism is OFF (0), there should only be “total energy”, which is kinetic+internal specific energies. Confused yet?
- Particle mass field
- Particle “masses” are actually stored as densities. This is to facilitate calculation of the gravitational potential. The net result of this is that, in order to calculate the stored particle “mass” to a physical mass, you must first multiply this field by the volume of a cell in which the particle resides. Remember that particle data is only stored in the most refined grid that covers that portion of the simulational volume.
When the simulation is done, Enzo will display the message “Successful run, exiting.” Enzo is a complicated code, with a similarly complicated output format. See the Enzo User Guide page on the Enzo output format? for more information on the data outputs.
Congratulations! If you’ve made it this far, you have now successfully run a simulation using Enzo!