<div dir="ltr">Hey Brendan,<div><br></div><div>Could you try running your script using the memory_profiler module on pypi?</div><div><br></div><div>Here's an example script that uses the memory profiler: <a href="http://paste.yt-project.org/show/4748/">http://paste.yt-project.org/show/4748/</a></div>
<div><br></div><div>and the output for that script: <a href="http://paste.yt-project.org/show/4749/">http://paste.yt-project.org/show/4749/</a></div><div><br></div><div>For what it's worth, it does indeed look like matt's suggestion to use load_uniform_grid is an option, and might be more memory efficient in the end since you will go directly to the uniform grid you want without creating an octree. Here's an example: <a href="http://paste.yt-project.org/show/4750/">http://paste.yt-project.org/show/4750/</a></div>
<div><br></div><div>Here's the memory usage information for that example: <a href="http://paste.yt-project.org/show/4751/">http://paste.yt-project.org/show/4751/</a></div><div><br></div><div>I used a 256^3 uniform grid with normally distributed random data - I'm not sure whether it will also be more memory efficient in your case.</div>
<div><br></div></div><div class="gmail_extra"><br><br><div class="gmail_quote">On Sun, Jun 8, 2014 at 9:32 PM, Brendan Griffen <span dir="ltr"><<a href="mailto:brendan.f.griffen@gmail.com" target="_blank">brendan.f.griffen@gmail.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div dir="ltr">This is the full error if it helps at all? It is indeed, loading in all of the quantities.<div><br></div>
<div><div>Loading particles...</div><div> --> Loading particle type: 1</div><div>yt : [INFO ] 2014-06-08 15:12:05,540 Parameters: current_time = 0.0</div>
<div>yt : [INFO ] 2014-06-08 15:12:05,540 Parameters: domain_dimensions = [2 2 2]</div><div>yt : [INFO ] 2014-06-08 15:12:05,541 Parameters: domain_left_edge = [ 0. 0. 0.]</div><div>yt : [INFO ] 2014-06-08 15:12:05,542 Parameters: domain_right_edge = [ 100. 100. 100.]</div>
<div>yt : [INFO ] 2014-06-08 15:12:05,542 Parameters: cosmological_simulation = 0.0</div><div>yt : [INFO ] 2014-06-08 15:12:05,548 Allocating for 1.074e+09 particles</div><div>yt : [INFO ] 2014-06-08 15:16:16,195 Identified 7.584e+07 octs</div>
<div>yt : [INFO ] 2014-06-08 15:16:16,299 Loading field plugins.</div><div>yt : [INFO ] 2014-06-08 15:16:16,299 Loaded angular_momentum (8 new fields)</div><div>yt : [INFO ] 2014-06-08 15:16:16,299 Loaded astro (14 new fields)</div>
<div>yt : [INFO ] 2014-06-08 15:16:16,300 Loaded cosmology (20 new fields)</div><div>yt : [INFO ] 2014-06-08 15:16:16,300 Loaded fluid (56 new fields)</div><div>yt : [INFO ] 2014-06-08 15:16:16,301 Loaded fluid_vector (88 new fields)</div>
<div>yt : [INFO ] 2014-06-08 15:16:16,301 Loaded geometric (103 new fields)</div><div>yt : [INFO ] 2014-06-08 15:16:16,301 Loaded local (103 new fields)</div><div>yt : [INFO ] 2014-06-08 15:16:16,302 Loaded magnetic_field (109 new fields)</div>
<div>yt : [INFO ] 2014-06-08 15:16:16,302 Loaded species (109 new fields)</div><div>---------------------------------------------------------------------------</div><div>MemoryError Traceback (most recent call last)</div>
<div class="">
<div>/nfs/blank/h4231/bgriffen/data/lib/yt-x86_64/lib/python2.7/site-packages/IPython/utils/py3compat.pyc in execfile(fname, *where)</div><div> 202 else:</div><div> 203 filename = fname</div>
<div>--> 204 __builtin__.execfile(filename, *where)</div><div><br></div><div>/nfs/blank/h4231/bgriffen/work/projects/caterpillar/c2ray/cic/ytcic.py in <module>()</div></div><div> 99 slc.set_figure_size(4)</div>
<div> 100 slc.save()</div><div>--> 101 </div><div> 102 for ndim in dimlist:</div><div> 103 print</div><div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc in __getitem__(self, key)</div>
<div> 218 return self.field_data[f]</div><div> 219 else:</div><div>--> 220 self.get_data(f)</div><div> 221 # fi.units is the unit expression string. We depend on the registry</div>
<div> 222 # hanging off the dataset to define this unit object.</div><div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc in get_data(self, fields)</div><div> 627 </div>
<div> 628 fields_to_generate += gen_fluids + gen_particles</div><div>--> 629 self._generate_fields(fields_to_generate)</div><div> 630 </div><div> 631 def _generate_fields(self, fields_to_generate):</div>
<div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc in _generate_fields(self, fields_to_generate)</div><div> 644 fi = self.pf._get_field_info(*field)</div>
<div> 645 try:</div><div>--> 646 fd = self._generate_field(field)</div><div> 647 if type(fd) == np.ndarray:</div><div> 648 fd = self.pf.arr(fd, fi.units)</div>
<div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc in _generate_field(self, field)</div><div> 255 tr = self._generate_particle_field(field)</div><div>
256 else:</div><div>--> 257 tr = self._generate_fluid_field(field)</div><div> 258 if tr is None:</div><div> 259 raise YTCouldNotGenerateField(field, <a href="http://self.pf" target="_blank">self.pf</a>)</div>
<div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc in _generate_fluid_field(self, field)</div><div> 273 finfo.check_available(gen_obj)</div><div> 274 except NeedsGridType as ngt_exception:</div>
<div>--> 275 rv = self._generate_spatial_fluid(field, ngt_exception.ghost_zones)</div><div> 276 else:</div><div> 277 rv = finfo(gen_obj)</div><div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc in _generate_spatial_fluid(self, field, ngz)</div>
<div> 289 o = self._current_chunk.objs[0]</div><div> 290 with o._activate_cache():</div><div>--> 291 ind += o.select(self.selector, self[field], rv, ind)</div>
<div> 292 else:</div><div> 293 chunks = self.index._chunk(self, "spatial", ngz = ngz)</div><div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc in __getitem__(self, key)</div>
<div> 218 return self.field_data[f]</div><div> 219 else:</div><div>--> 220 self.get_data(f)</div><div> 221 # fi.units is the unit expression string. We depend on the registry</div>
<div> 222 # hanging off the dataset to define this unit object.</div><div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc in get_data(self, fields)</div><div> 627 </div>
<div> 628 fields_to_generate += gen_fluids + gen_particles</div><div>--> 629 self._generate_fields(fields_to_generate)</div><div> 630 </div><div> 631 def _generate_fields(self, fields_to_generate):</div>
<div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc in _generate_fields(self, fields_to_generate)</div><div> 644 fi = self.pf._get_field_info(*field)</div>
<div> 645 try:</div><div>--> 646 fd = self._generate_field(field)</div><div> 647 if type(fd) == np.ndarray:</div><div> 648 fd = self.pf.arr(fd, fi.units)</div>
<div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc in _generate_field(self, field)</div><div> 255 tr = self._generate_particle_field(field)</div><div>
256 else:</div><div>--> 257 tr = self._generate_fluid_field(field)</div><div> 258 if tr is None:</div><div> 259 raise YTCouldNotGenerateField(field, <a href="http://self.pf" target="_blank">self.pf</a>)</div>
<div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc in _generate_fluid_field(self, field)</div><div> 275 rv = self._generate_spatial_fluid(field, ngt_exception.ghost_zones)</div>
<div> 276 else:</div><div>--> 277 rv = finfo(gen_obj)</div><div> 278 return rv</div><div> 279 </div><div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/fields/derived_field.pyc in __call__(self, data)</div>
<div> 176 "for %s" % (<a href="http://self.name" target="_blank">self.name</a>,))</div><div> 177 with self.unit_registry(data):</div><div>--> 178 dd = self._function(self, data)</div>
<div> 179 for field_name in data.keys():</div><div> 180 if field_name not in original_fields:</div><div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/fields/particle_fields.pyc in particle_cic(field, data)</div>
<div> 113 def particle_cic(field, data):</div><div> 114 pos = data[ptype, coord_name]</div><div>--> 115 d = data.deposit(pos, [data[ptype, mass_name]], method = "cic")</div><div> 116 d = data.apply_units(d, data[ptype, mass_name].units)</div>
<div> 117 d /= data["index", "cell_volume"]</div><div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/octree_subset.pyc in deposit(self, positions, fields, method)</div>
<div> 167 mylog.debug("Depositing %s (%s^3) particles into %s Octs",</div><div> 168 positions.shape[0], positions.shape[0]**0.3333333, nvals[-1])</div><div>--> 169 pos = np.asarray(positions.convert_to_units("code_length"),</div>
<div> 170 dtype="float64")</div><div> 171 # We should not need the following if we know in advance all our fields</div><div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/units/yt_array.pyc in convert_to_units(self, units)</div>
<div> 366 </div><div> 367 self.units = new_units</div><div>--> 368 self *= conversion_factor</div><div> 369 return self</div><div> 370 </div><div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/units/yt_array.pyc in __imul__(self, other)</div>
<div> 667 """ See __mul__. """</div><div> 668 oth = sanitize_units_mul(self, other)</div><div>--> 669 return np.multiply(self, oth, out=self)</div><div> 670 </div>
<div> 671 def __div__(self, right_object):</div><div><br></div><div>/bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/units/yt_array.pyc in __array_wrap__(self, out_arr, context)</div><div> 966 # casting to YTArray avoids creating a YTQuantity with size > 1</div>
<div> 967 return YTArray(np.array(out_arr, unit))</div><div>--> 968 return ret_class(np.array(out_arr), unit)</div><div> 969 </div><div> 970 </div><div><br></div><div>MemoryError:</div>
</div></div><div class="HOEnZb"><div class="h5"><div class="gmail_extra"><br><br><div class="gmail_quote">On Sun, Jun 8, 2014 at 10:50 PM, Matthew Turk <span dir="ltr"><<a href="mailto:matthewturk@gmail.com" target="_blank">matthewturk@gmail.com</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi Brendan,<br>
<div><br>
On Sun, Jun 8, 2014 at 9:21 PM, Brendan Griffen<br>
<<a href="mailto:brendan.f.griffen@gmail.com" target="_blank">brendan.f.griffen@gmail.com</a>> wrote:<br>
> Hi Matt,<br>
><br>
> Thanks for your detailed email. Forgive my naivety but why do you need the<br>
> oct-tree in the first place? I have a my own fortran code for constructing a<br>
> cloud in cell mesh and it uses very little overhead (just the n^3 grid and<br>
> the particle data itself). I then calculate the dx,dy,dzs to the nearest 8<br>
> grid points and distribute accordingly in a omp loop which is done in a<br>
> fraction of a second. Does the situation with yt come about (oct tree etc.)<br>
> necessarily because of the way it handles particle data? Is it essentially<br>
> used to map the particles to domains in the grid or something?<br>
<br>
</div>That's not naive at all. There are two reasons --<br>
<br>
1) The octree is used for indexing for neighbor lookups and<br>
early-termination of region selection for particles<br>
2) The octree is used to estimate the "required resolution" for any<br>
operation that requires a space-filling value. (i.e., any time that a<br>
particle becomes a volume.)<br>
<br>
Projections in yt are adaptive, in that they project down to the<br>
finest appropriate resolution. There's also the "arbitrary_grid"<br>
operation, which does precisely what you're describing, but as it<br>
stands right now the octree gets constructed at time of instantiation<br>
of the indexing system. Thinking it over, you may be able to avoid<br>
that completely by not using load_particles and instead using<br>
load_uniform_grid and supplying your desired dimensions. The field<br>
names should be the same.<br>
<div><br>
><br>
> The machine I max memory on has 128GB and the snapshots are using 1024^3<br>
> particles. Do you have any idea of how much memory the oct-tree uses as a<br>
> function of particle/grid number? I am going to try on a 256GB machine<br>
> (though this is a bit of a hassle). I'll see how I go.<br>
<br>
</div>I am disappointed that it's blowing out your RAM. This week I will<br>
try to get some memory profiling done. Could you file a bug to this<br>
effect, which will help me track it? Peak memory usage during<br>
indexing should only be 64 bits * Nparticles, unless you're using<br>
load_particles, in which case all the fields will *also* have to be in<br>
memory. It's about 8 gigabytes per field. So, I think there's<br>
something going wrong.<br>
<div><div><br>
><br>
> Thanks.<br>
><br>
> Brendan<br>
><br>
><br>
> On Sun, Jun 8, 2014 at 6:25 PM, Matthew Turk <<a href="mailto:matthewturk@gmail.com" target="_blank">matthewturk@gmail.com</a>> wrote:<br>
>><br>
>> Hi all,<br>
>><br>
>> I feel like I owe a brief explanation of why things are tricky right<br>
>> now, what we're planning on doing, and how we're experimenting and<br>
>> developing.<br>
>><br>
>> Presently, the particle geometry handlers build a single mesh from all<br>
>> particles in the dataset, along with a coarse bitmask that correlates<br>
>> files to regions in the domain. This requires the allocation of a<br>
>> single int64 array of size Nparticles, which is sorted in place and<br>
>> then fed into an octree construction algorithm that then spits back<br>
>> out the mesh. Each octree component contains 3 64-bit integers and<br>
>> eitehr a void pointer or a pointer to eight other octs. Increasing<br>
>> n_ref decreases the number of octs in this mesh; when smoothing<br>
>> operaitons are conducted, a second "index" mesh is created for looking<br>
>> up particles near mesh points. Mesh points are used for adaptive<br>
>> resolution smoothing and other "deposit particles on the grid somehow"<br>
>> operations (including SPH kernel).<br>
>><br>
>> Anyway, because right now it requires a global mesh to be constructed,<br>
>> this is expensive and requires holding a 64-bit integer in memory for<br>
>> each particle. I think if you're loading the particles in differently<br>
>> there is some additional overhead as well, but I'm still a bit<br>
>> surprised you OOM on a 1024^3 dataset.<br>
>><br>
>> In general, we don't *need* this global mesh; is can be constructed as<br>
>> required, which would speed up both the initial index phase as well as<br>
>> the final meshing process. I got about 50% of the way to implementing<br>
>> this last fall, but because of various concerns and deadlines I<br>
>> haven't finished it. I intend to get back to it probably in July,<br>
>> right after we put out a 3.0, so that we can have it in time for 3.1.<br>
>> In principle this will make the particle codes much more similar to<br>
>> ARTIO, in that the mesh will be constructed only as required and<br>
>> discarded when no longer required, which will make them much more<br>
>> memory efficient.<br>
>><br>
>> But, getting a single mesh for extremely large data is a very high<br>
>> priority; right now for the 10240^3 run we've been loading up<br>
>> individual sub-chunks, which I want to stop doing.<br>
>><br>
>> From the technical perspective, these are the things that need to<br>
>> happen on the yt side for particle datasets to move to this "lazy"<br>
>> mode of loading; most of this is based on things learned from 2HOT and<br>
>> ARTIO, and will involve converting to a forest-of-octrees.<br>
>><br>
>> * Split into spatially-organized subchunks of ParticleOctreeSubset<br>
>> objects, such that these map 1:Nfiles, and that can be constructed on<br>
>> the fly.<br>
>> * Construct a dual-mesh of the bitmask "ParticleRegion" object that<br>
>> will help with identifying neighbors to a given oct cell, so that if<br>
>> we're inside one octree we know which neighbor octrees to grab if we<br>
>> need particles for smoothing things (fast boundary particle<br>
>> identification is later down the road)<br>
>> * Parallel sort of particles, or using the parallel ring function;<br>
>> may not be necessary after all<br>
>><br>
>> All of this is doable, and I'd be happy to work with people if they'd<br>
>> like to take a shot at implementing it, but I've mostly put it on my<br>
>> list for post-3.0.<br>
>><br>
>> -Matt<br>
>><br>
>> On Sun, Jun 8, 2014 at 2:43 PM, Nathan Goldbaum <<a href="mailto:nathan12343@gmail.com" target="_blank">nathan12343@gmail.com</a>><br>
>> wrote:<br>
>> ><br>
>> ><br>
>> ><br>
>> > On Sun, Jun 8, 2014 at 12:27 PM, Brendan Griffen<br>
>> > <<a href="mailto:brendan.f.griffen@gmail.com" target="_blank">brendan.f.griffen@gmail.com</a>> wrote:<br>
>> >><br>
>> >> Also, how do I construct just a zero filled yt array with dimensions<br>
>> >> (ndim,ndim,ndim)? Thanks<br>
>> ><br>
>> ><br>
>> ><br>
>> > from yt import YTArray<br>
>> > from numpy import np<br>
>> ><br>
>> > arr = YTArray(np.zeros([ndim, ndim, ndim]), input_units=units_string)<br>
>> ><br>
>> > or alternatively:<br>
>> ><br>
>> > from yt.units import kiloparsec<br>
>> ><br>
>> > arr = kiloparsec*np.zeros([ndim, ndim, ndim])<br>
>> ><br>
>> > it doesn't have to be kiloparsec - you can compose the units you want<br>
>> > out of<br>
>> > any of the unit symbols that live in yt.units.<br>
>> ><br>
>> > See this page for a ton more detail about yt's new unit system:<br>
>> > <a href="http://yt-project.org/docs/dev-3.0/analyzing/units/index.html" target="_blank">http://yt-project.org/docs/dev-3.0/analyzing/units/index.html</a><br>
>> ><br>
>> >><br>
>> >><br>
>> >> Brendan<br>
>> >><br>
>> >><br>
>> >> On Sun, Jun 8, 2014 at 3:26 PM, Brendan Griffen<br>
>> >> <<a href="mailto:brendan.f.griffen@gmail.com" target="_blank">brendan.f.griffen@gmail.com</a>> wrote:<br>
>> >>><br>
>> >>> Hi,<br>
>> >>><br>
>> >>> Since I get memory errors. Could I not just read in the blocks of the<br>
>> >>> output individually then basically stack the mesh each time. That way<br>
>> >>> not<br>
>> >>> every single particle of the snapshot has to be loaded at the same<br>
>> >>> time.<br>
>> >>> Would that just be a case of doing<br>
>> >>><br>
>> >>> level = int(math.log(ndim,2))<br>
>> >>> cg = ds.covering_grid(level=level,<br>
>> >>> left_edge=[0,0,0],dims=[ndim,ndim,ndim])<br>
>> >>> arr = cg['deposit', 'all_density']<br>
>> >>> arrall += arr<br>
>> >>><br>
>> >>> in a loop over each HDF5 block?<br>
>> ><br>
>> ><br>
>> > It's likely that the memory use is dominated by the octree rather than<br>
>> > the<br>
>> > covering grid. This is with 1024^3 particles, correct?<br>
>> ><br>
>> > You can probably significantly reduce the memory used by the octree by<br>
>> > increasing n_ref in the call to load_particles.<br>
>> ><br>
>> > See this page for more detail about load_particles:<br>
>> ><br>
>> > <a href="http://yt-project.org/docs/dev-3.0/examining/loading_data.html#generic-particle-data" target="_blank">http://yt-project.org/docs/dev-3.0/examining/loading_data.html#generic-particle-data</a><br>
>> ><br>
>> > Larger n_ref means fewer octree cells (lower resolution), but it also<br>
>> > means<br>
>> > lower poisson noise and lower memory use.<br>
>> ><br>
>> > Alternatively, as Matt suggested, you could break your 1024^3 ensemble<br>
>> > of<br>
>> > particles up into chunks, loop over the chunk, creating a particle<br>
>> > octree<br>
>> > and then a covering grid for each subset of the particles. Your final<br>
>> > covering grid is just the sub of the covering grids for each subset of<br>
>> > particles.<br>
>> ><br>
>> >>><br>
>> >>><br>
>> >>> Thanks.<br>
>> >>> Brendan<br>
>> >>><br>
>> >>><br>
>> >>><br>
>> >>><br>
>> >>> On Fri, Jun 6, 2014 at 7:26 PM, Matthew Turk <<a href="mailto:matthewturk@gmail.com" target="_blank">matthewturk@gmail.com</a>><br>
>> >>> wrote:<br>
>> >>>><br>
>> >>>><br>
>> >>>> On Jun 6, 2014 4:54 PM, "Brendan Griffen"<br>
>> >>>> <<a href="mailto:brendan.f.griffen@gmail.com" target="_blank">brendan.f.griffen@gmail.com</a>><br>
>> >>>> wrote:<br>
>> >>>> ><br>
>> >>>> > OK great. It is very low resolution but it worked. Thanks for all<br>
>> >>>> > your<br>
>> >>>> > help. My higher resolution run 1024^3 in 100 Mpc seems to crash on<br>
>> >>>> > 128GB<br>
>> >>>> > memory machine. I might have to look elsewhere.<br>
>> >>>> ><br>
>> >>>><br>
>> >>>> If you are looking for low resolution extraction you can tune the<br>
>> >>>> memory usage by changing the parameter n_ref to something higher.<br>
>> >>>><br>
>> >>>> Supporting extremely large datasets in a single mesh is on the<br>
>> >>>> roadmap<br>
>> >>>> for the late summer or fall, after a 3.0 release goes out. For now<br>
>> >>>> you can<br>
>> >>>> also extract before you load in; this is sort of how we are<br>
>> >>>> supporting an<br>
>> >>>> INCITE project with very large particle counts.<br>
>> >>>><br>
>> >>>><br>
>> >>>> > Also, I normally use Canopy distribution but I just use an alias to<br>
>> >>>> > loadyt which erases my PYTHONPATH and I can't access scipy and a<br>
>> >>>> > few other<br>
>> >>>> > libraries any more. What is the best practice here? Should I just<br>
>> >>>> > manually<br>
>> >>>> > export PYTHONPATH and point to the libraries need in canopy or can<br>
>> >>>> > they play<br>
>> >>>> > nice together?<br>
>> >>>> ><br>
>> >>>> > Thanks.<br>
>> >>>> ><br>
>> >>>> > BG<br>
>> >>>> ><br>
>> >>>> ><br>
>> >>>> > On Fri, Jun 6, 2014 at 2:54 PM, Nathan Goldbaum<br>
>> >>>> > <<a href="mailto:nathan12343@gmail.com" target="_blank">nathan12343@gmail.com</a>> wrote:<br>
>> >>>> >><br>
>> >>>> >><br>
>> >>>> >><br>
>> >>>> >><br>
>> >>>> >> On Fri, Jun 6, 2014 at 11:48 AM, Brendan Griffen<br>
>> >>>> >> <<a href="mailto:brendan.f.griffen@gmail.com" target="_blank">brendan.f.griffen@gmail.com</a>> wrote:<br>
>> >>>> >>><br>
>> >>>> >>> OK great. Thanks. I just wanted a homogeneous mesh. 512^3 with no<br>
>> >>>> >>> nesting of any kind. Though when I plot the image it looks like<br>
>> >>>> >>> it is<br>
>> >>>> >>> assigning particles incorrectly (low resolution on the outside).<br>
>> >>>> >>> This is<br>
>> >>>> >>> just a test image.<br>
>> >>>> >>><br>
>> >>>> >><br>
>> >>>> >> The SlicePlot is visualizing the octree so there is less<br>
>> >>>> >> resolution<br>
>> >>>> >> where there are fewer particles. If you want to visualize the<br>
>> >>>> >> covering grid<br>
>> >>>> >> you're going to need to visualize that separately.<br>
>> >>>> >><br>
>> >>>> >>><br>
>> >>>> >>> ds = yt.load_particles(data, length_unit=3.08e24,<br>
>> >>>> >>> mass_unit=1.9891e33,bbox=bbox)<br>
>> >>>> >>><br>
>> >>>> >>> ad = ds.all_data()<br>
>> >>>> >>> print ad['deposit', 'all_cic']<br>
>> >>>> >>> slc = yt.SlicePlot(ds, 2, ('deposit', 'all_cic'))<br>
>> >>>> >>> slc.set_figure_size(4)<br>
>> >>>> >>> cg = ds.covering_grid(level=9,<br>
>> >>>> >>> left_edge=[0,0,0],dims=[512,512,512])<br>
>> >>>> >>><br>
>> >>>> >><br>
>> >>>> >> To actually produce the uniform resolution ndarray, you're going<br>
>> >>>> >> to<br>
>> >>>> >> need to do something like:<br>
>> >>>> >><br>
>> >>>> >> array = cg[('deposit', 'all_cic')]<br>
>> >>>> >><br>
>> >>>> >> array will then be a 3D array you can do whatever you want with.<br>
>> >>>> >> By<br>
>> >>>> >> default it has units, but to strip them off you'll just need to<br>
>> >>>> >> cast to<br>
>> >>>> >> ndarray:<br>
>> >>>> >><br>
>> >>>> >> array_without_units = array.v<br>
>> >>>> >><br>
>> >>>> >><br>
>> >>>> >>><br>
>> >>>> >>> Also, is there a way to load multiple particle types?<br>
>> >>>> >>><br>
>> >>>> >>> Do I just need to stack the particles into the array here?<br>
>> >>>> >>><br>
>> >>>> >>> data = {'particle_position_x': pos[:,0],<br>
>> >>>> >>> 'particle_position_y': pos[:,1],<br>
>> >>>> >>> 'particle_position_z': pos[:,2],<br>
>> >>>> >>> 'particle_mass': np.array([mpart]*npart)}<br>
>> >>>> >>><br>
>> >>>> >>> Then feed it in as usual?<br>
>> >>>> >><br>
>> >>>> >><br>
>> >>>> >> That's right, although if the particle masses are different for<br>
>> >>>> >> the<br>
>> >>>> >> different particle types that code snippet will need to be<br>
>> >>>> >> generalized to<br>
>> >>>> >> handle that.<br>
>> >>>> >><br>
>> >>>> >> I think in principle it should be possible to make load_particles<br>
>> >>>> >> handle different particle types just like an SPH dataset that<br>
>> >>>> >> contains<br>
>> >>>> >> multiple particle types, but right now that hasn't been<br>
>> >>>> >> implemented yet.<br>
>> >>>> >><br>
>> >>>> >>><br>
>> >>>> >>><br>
>> >>>> >>> Brendan<br>
>> >>>> >>><br>
>> >>>> >>><br>
>> >>>> >>> On Thu, Jun 5, 2014 at 9:44 PM, Nathan Goldbaum<br>
>> >>>> >>> <<a href="mailto:nathan12343@gmail.com" target="_blank">nathan12343@gmail.com</a>> wrote:<br>
>> >>>> >>>><br>
>> >>>> >>>> That's right, you can set that via the bbox keyword parameter<br>
>> >>>> >>>> for<br>
>> >>>> >>>> load_particles. I'd urge you to take a look at the docstrings<br>
>> >>>> >>>> and source<br>
>> >>>> >>>> code for load_particles.<br>
>> >>>> >>>><br>
>> >>>> >>>><br>
>> >>>> >>>> On Thu, Jun 5, 2014 at 6:34 PM, Brendan Griffen<br>
>> >>>> >>>> <<a href="mailto:brendan.f.griffen@gmail.com" target="_blank">brendan.f.griffen@gmail.com</a>> wrote:<br>
>> >>>> >>>>><br>
>> >>>> >>>>> Thanks very much Nathan. I tried to load in my own data but I<br>
>> >>>> >>>>> think there are too many particles or I have to specifically<br>
>> >>>> >>>>> set the domain<br>
>> >>>> >>>>> size.<br>
>> >>>> >>>>><br>
>> >>>> >>>>> In this area:<br>
>> >>>> >>>>><br>
>> >>>> >>>>> data = {'particle_position_x': pos[:,0],<br>
>> >>>> >>>>> 'particle_position_y': pos[:,1],<br>
>> >>>> >>>>> 'particle_position_z': pos[:,2],<br>
>> >>>> >>>>> 'particle_mass': np.array([mpart]*npart)}<br>
>> >>>> >>>>><br>
>> >>>> >>>>> ds = yt.load_particles(data, length_unit=3.08e24,<br>
>> >>>> >>>>> mass_unit=1.9891e36)<br>
>> >>>> >>>>> ad = ds.all_data()<br>
>> >>>> >>>>> print ad['deposit', 'all_cic']<br>
>> >>>> >>>>><br>
>> >>>> >>>>> In [3]: run ytcic.py<br>
>> >>>> >>>>> yt : [INFO ] 2014-06-05 21:29:06,183 Parameters:<br>
>> >>>> >>>>> current_time<br>
>> >>>> >>>>> = 0.0<br>
>> >>>> >>>>> yt : [INFO ] 2014-06-05 21:29:06,183 Parameters:<br>
>> >>>> >>>>> domain_dimensions = [2 2 2]<br>
>> >>>> >>>>> yt : [INFO ] 2014-06-05 21:29:06,184 Parameters:<br>
>> >>>> >>>>> domain_left_edge = [ 0. 0. 0.]<br>
>> >>>> >>>>> yt : [INFO ] 2014-06-05 21:29:06,185 Parameters:<br>
>> >>>> >>>>> domain_right_edge = [ 1. 1. 1.]<br>
>> >>>> >>>>> yt : [INFO ] 2014-06-05 21:29:06,185 Parameters:<br>
>> >>>> >>>>> cosmological_simulation = 0.0<br>
>> >>>> >>>>> yt : [INFO ] 2014-06-05 21:29:06,188 Allocating for<br>
>> >>>> >>>>> 1.342e+08<br>
>> >>>> >>>>> particles<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> ---------------------------------------------------------------------------<br>
>> >>>> >>>>> YTDomainOverflow Traceback (most<br>
>> >>>> >>>>> recent<br>
>> >>>> >>>>> call last)<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /nfs/blank/h4231/bgriffen/data/lib/yt-x86_64/lib/python2.7/site-packages/IPython/utils/py3compat.pyc<br>
>> >>>> >>>>> in execfile(fname, *where)<br>
>> >>>> >>>>> 202 else:<br>
>> >>>> >>>>> 203 filename = fname<br>
>> >>>> >>>>> --> 204 __builtin__.execfile(filename, *where)<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /nfs/blank/h4231/bgriffen/work/projects/caterpillar/c2ray/cic/ytcic.py in<br>
>> >>>> >>>>> <module>()<br>
>> >>>> >>>>> 52<br>
>> >>>> >>>>> 53 ad = ds.all_data()<br>
>> >>>> >>>>> ---> 54 print ad['deposit', 'all_cic']<br>
>> >>>> >>>>> 55 slc = yt.SlicePlot(ds, 2, ('deposit', 'all_cic'))<br>
>> >>>> >>>>> 56 slc.set_figure_size(4)<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc<br>
>> >>>> >>>>> in __getitem__(self, key)<br>
>> >>>> >>>>> 205 Returns a single field. Will add if necessary.<br>
>> >>>> >>>>> 206 """<br>
>> >>>> >>>>> --> 207 f = self._determine_fields([key])[0]<br>
>> >>>> >>>>> 208 if f not in self.field_data and key not in<br>
>> >>>> >>>>> self.field_data:<br>
>> >>>> >>>>> 209 if f in self._container_fields:<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/data_containers.pyc<br>
>> >>>> >>>>> in _determine_fields(self, fields)<br>
>> >>>> >>>>> 453 raise YTFieldNotParseable(field)<br>
>> >>>> >>>>> 454 ftype, fname = field<br>
>> >>>> >>>>> --> 455 finfo = self.pf._get_field_info(ftype,<br>
>> >>>> >>>>> fname)<br>
>> >>>> >>>>> 456 else:<br>
>> >>>> >>>>> 457 fname = field<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/static_output.pyc<br>
>> >>>> >>>>> in _get_field_info(self, ftype, fname)<br>
>> >>>> >>>>> 445 _last_finfo = None<br>
>> >>>> >>>>> 446 def _get_field_info(self, ftype, fname = None):<br>
>> >>>> >>>>> --> 447 self.index<br>
>> >>>> >>>>> 448 if fname is None:<br>
>> >>>> >>>>> 449 ftype, fname = "unknown", ftype<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/data_objects/static_output.pyc<br>
>> >>>> >>>>> in index(self)<br>
>> >>>> >>>>> 277 raise RuntimeError("You should not<br>
>> >>>> >>>>> instantiate Dataset.")<br>
>> >>>> >>>>> 278 self._instantiated_index =<br>
>> >>>> >>>>> self._index_class(<br>
>> >>>> >>>>> --> 279 self, dataset_type=self.dataset_type)<br>
>> >>>> >>>>> 280 # Now we do things that we need an<br>
>> >>>> >>>>> instantiated index for<br>
>> >>>> >>>>> 281 # ...first off, we create our field_info<br>
>> >>>> >>>>> now.<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/frontends/stream/data_structures.pyc<br>
>> >>>> >>>>> in __init__(self, pf, dataset_type)<br>
>> >>>> >>>>> 942 def __init__(self, pf, dataset_type = None):<br>
>> >>>> >>>>> 943 self.stream_handler = pf.stream_handler<br>
>> >>>> >>>>> --> 944 super(StreamParticleIndex, self).__init__(pf,<br>
>> >>>> >>>>> dataset_type)<br>
>> >>>> >>>>> 945<br>
>> >>>> >>>>> 946 def _setup_data_io(self):<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/geometry/particle_geometry_handler.pyc<br>
>> >>>> >>>>> in __init__(self, pf, dataset_type)<br>
>> >>>> >>>>> 48 self.directory =<br>
>> >>>> >>>>> os.path.dirname(self.index_filename)<br>
>> >>>> >>>>> 49 self.float_type = np.float64<br>
>> >>>> >>>>> ---> 50 super(ParticleIndex, self).__init__(pf,<br>
>> >>>> >>>>> dataset_type)<br>
>> >>>> >>>>> 51<br>
>> >>>> >>>>> 52 def _setup_geometry(self):<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/geometry/geometry_handler.pyc<br>
>> >>>> >>>>> in __init__(self, pf, dataset_type)<br>
>> >>>> >>>>> 54<br>
>> >>>> >>>>> 55 mylog.debug("Setting up domain geometry.")<br>
>> >>>> >>>>> ---> 56 self._setup_geometry()<br>
>> >>>> >>>>> 57<br>
>> >>>> >>>>> 58 mylog.debug("Initializing data grid data IO")<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/geometry/particle_geometry_handler.pyc<br>
>> >>>> >>>>> in _setup_geometry(self)<br>
>> >>>> >>>>> 52 def _setup_geometry(self):<br>
>> >>>> >>>>> 53 mylog.debug("Initializing Particle Geometry<br>
>> >>>> >>>>> Handler.")<br>
>> >>>> >>>>> ---> 54 self._initialize_particle_handler()<br>
>> >>>> >>>>> 55<br>
>> >>>> >>>>> 56<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/geometry/particle_geometry_handler.pyc<br>
>> >>>> >>>>> in _initialize_particle_handler(self)<br>
>> >>>> >>>>> 87 pf.domain_left_edge,<br>
>> >>>> >>>>> pf.domain_right_edge,<br>
>> >>>> >>>>> 88 [N, N, N], len(self.data_files))<br>
>> >>>> >>>>> ---> 89 self._initialize_indices()<br>
>> >>>> >>>>> 90 self.oct_handler.finalize()<br>
>> >>>> >>>>> 91 self.max_level = self.oct_handler.max_level<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/geometry/particle_geometry_handler.pyc<br>
>> >>>> >>>>> in _initialize_indices(self)<br>
>> >>>> >>>>> 109 npart =<br>
>> >>>> >>>>> sum(data_file.total_particles.values())<br>
>> >>>> >>>>> 110 morton[ind:ind + npart] = \<br>
>> >>>> >>>>> --> 111 self.io._initialize_index(data_file,<br>
>> >>>> >>>>> self.regions)<br>
>> >>>> >>>>> 112 ind += npart<br>
>> >>>> >>>>> 113 morton.sort()<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> /bigbang/data/bgriffen/lib/yt-x86_64/src/yt-hg/yt/frontends/stream/io.pyc in<br>
>> >>>> >>>>> _initialize_index(self, data_file, regions)<br>
>> >>>> >>>>> 144 raise YTDomainOverflow(pos.min(axis=0),<br>
>> >>>> >>>>> pos.max(axis=0),<br>
>> >>>> >>>>> 145<br>
>> >>>> >>>>> data_file.pf.domain_left_edge,<br>
>> >>>> >>>>> --> 146<br>
>> >>>> >>>>> data_file.pf.domain_right_edge)<br>
>> >>>> >>>>> 147 regions.add_data_file(pos,<br>
>> >>>> >>>>> data_file.file_id)<br>
>> >>>> >>>>> 148 morton.append(compute_morton(<br>
>> >>>> >>>>><br>
>> >>>> >>>>> YTDomainOverflow: Particle bounds [ 0. 0. 0.] and [<br>
>> >>>> >>>>> 99.99999237<br>
>> >>>> >>>>> 99.99999237 99.99999237] exceed domain bounds [ 0. 0. 0.]<br>
>> >>>> >>>>> code_length and<br>
>> >>>> >>>>> [ 1. 1. 1.] code_length<br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> On Thu, Jun 5, 2014 at 8:22 PM, Nathan Goldbaum<br>
>> >>>> >>>>> <<a href="mailto:nathan12343@gmail.com" target="_blank">nathan12343@gmail.com</a>> wrote:<br>
>> >>>> >>>>>><br>
>> >>>> >>>>>> Here's a worked out example that does what you're looking for<br>
>> >>>> >>>>>> using a fake 1 million particle dataset:<br>
>> >>>> >>>>>><br>
>> >>>> >>>>>><br>
>> >>>> >>>>>> <a href="http://nbviewer.ipython.org/gist/ngoldbaum/546d37869aafe71cfe38" target="_blank">http://nbviewer.ipython.org/gist/ngoldbaum/546d37869aafe71cfe38</a><br>
>> >>>> >>>>>><br>
>> >>>> >>>>>> In this notebook I make use of two key yt features:<br>
>> >>>> >>>>>> `load_particles`, and `covering_grid`.<br>
>> >>>> >>>>>><br>
>> >>>> >>>>>> load_particles creates a "stream" dataset based on in-memory<br>
>> >>>> >>>>>> data<br>
>> >>>> >>>>>> fed in as a numpy array. This dataset acts just like an<br>
>> >>>> >>>>>> on-disk simulation<br>
>> >>>> >>>>>> dataset, but doesn't come with the baggage of needing to write<br>
>> >>>> >>>>>> a custom<br>
>> >>>> >>>>>> frontend to read a specific data format off disk.<br>
>> >>>> >>>>>><br>
>> >>>> >>>>>> covering_grid is a way to generate uniform resolution data<br>
>> >>>> >>>>>> from<br>
>> >>>> >>>>>> an AMR dataset. It acts like a python dictionary where the<br>
>> >>>> >>>>>> keys are field<br>
>> >>>> >>>>>> names and returns 3D numpy arrays of whatever uniform<br>
>> >>>> >>>>>> resolution you specify<br>
>> >>>> >>>>>> when you create the covering_grid.<br>
>> >>>> >>>>>><br>
>> >>>> >>>>>> Note that if you're using load_particles all of your data<br>
>> >>>> >>>>>> needs<br>
>> >>>> >>>>>> to live in memory. If your data is too big for that you'll<br>
>> >>>> >>>>>> need to write a<br>
>> >>>> >>>>>> frontend for your data format or use a memmap to an on-disk<br>
>> >>>> >>>>>> file somehow.<br>
>> >>>> >>>>>> I'm not an expert on that but others on the list should be<br>
>> >>>> >>>>>> able to help out.<br>
>> >>>> >>>>>><br>
>> >>>> >>>>>> Hope that gets you well on your way :)<br>
>> >>>> >>>>>><br>
>> >>>> >>>>>> -Nathan<br>
>> >>>> >>>>>><br>
>> >>>> >>>>>><br>
>> >>>> >>>>>> On Thu, Jun 5, 2014 at 5:04 PM, Desika Narayanan<br>
>> >>>> >>>>>> <<a href="mailto:dnarayan@haverford.edu" target="_blank">dnarayan@haverford.edu</a>> wrote:<br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>>> Hey Brendan,<br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>>> A couple of extra tools you might find helpful in conjunction<br>
>> >>>> >>>>>>> with Nathan's example of depositing the particles onto an<br>
>> >>>> >>>>>>> octree are at:<br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>>> <a href="http://paste.yt-project.org/show/4737/" target="_blank">http://paste.yt-project.org/show/4737/</a><br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>>> Where I load a gadget snapshot, and then recover the<br>
>> >>>> >>>>>>> coordinates<br>
>> >>>> >>>>>>> and width of each cell.<br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>>> In response to your last question - the particles are<br>
>> >>>> >>>>>>> deposited<br>
>> >>>> >>>>>>> into an octree grid (so, you'll see that the cell sizes<br>
>> >>>> >>>>>>> aren't all the same<br>
>> >>>> >>>>>>> size). I don't know if depositing onto a regular NxNxN mesh<br>
>> >>>> >>>>>>> is possible,<br>
>> >>>> >>>>>>> though would be interested to hear if so.<br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>>> -d<br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>>> On Thu, Jun 5, 2014 at 7:58 PM, Brendan Griffen<br>
>> >>>> >>>>>>> <<a href="mailto:brendan.f.griffen@gmail.com" target="_blank">brendan.f.griffen@gmail.com</a>> wrote:<br>
>> >>>> >>>>>>>><br>
>> >>>> >>>>>>>> Thanks. I'll get the "bleeding edge" version first then try<br>
>> >>>> >>>>>>>> your suggestions. Though I want to return the NxNxN array<br>
>> >>>> >>>>>>>> and be able to<br>
>> >>>> >>>>>>>> write this mesh to a file. It is *only* using the cic part<br>
>> >>>> >>>>>>>> of yt and it<br>
>> >>>> >>>>>>>> should return the mesh to be written? Just wanted to<br>
>> >>>> >>>>>>>> clarify?<br>
>> >>>> >>>>>>>><br>
>> >>>> >>>>>>>> Thanks.<br>
>> >>>> >>>>>>>> Brendan<br>
>> >>>> >>>>>>>><br>
>> >>>> >>>>>>>><br>
>> >>>> >>>>>>>> On Thu, Jun 5, 2014 at 6:49 PM, Nathan Goldbaum<br>
>> >>>> >>>>>>>> <<a href="mailto:nathan12343@gmail.com" target="_blank">nathan12343@gmail.com</a>> wrote:<br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>> On Thu, Jun 5, 2014 at 3:36 PM, John ZuHone<br>
>> >>>> >>>>>>>>> <<a href="mailto:jzuhone@gmail.com" target="_blank">jzuhone@gmail.com</a>> wrote:<br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> Hi Brendan,<br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> Which version of yt are you using?<br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> If you're using 3.0, this is actually fairly easy. If you<br>
>> >>>> >>>>>>>>>> look in <a href="http://yt.fields.particle_fields.py" target="_blank">yt.fields.particle_fields.py</a>, around line 85, you<br>
>> >>>> >>>>>>>>>> can see how this<br>
>> >>>> >>>>>>>>>> is done for the "particle_density" and "particle_mass"<br>
>> >>>> >>>>>>>>>> fields. Basically you<br>
>> >>>> >>>>>>>>>> can call a "deposit" method which takes the particle field<br>
>> >>>> >>>>>>>>>> quantity you want<br>
>> >>>> >>>>>>>>>> deposited and deposits it into cells. The underlying<br>
>> >>>> >>>>>>>>>> calculation is done<br>
>> >>>> >>>>>>>>>> using Cython, so it's fast.<br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>> And you shouldn't ever actually need to call these<br>
>> >>>> >>>>>>>>> "deposit"<br>
>> >>>> >>>>>>>>> functions, since "deposit" is exposed as a field type for<br>
>> >>>> >>>>>>>>> all datasets that<br>
>> >>>> >>>>>>>>> contain particles.<br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>> Here is a notebook that does this for Enzo AMR data:<br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>> <a href="http://nbviewer.ipython.org/gist/ngoldbaum/5e19e4e6cc2bf330149c" target="_blank">http://nbviewer.ipython.org/gist/ngoldbaum/5e19e4e6cc2bf330149c</a><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>> This dataset contains about a million particles and<br>
>> >>>> >>>>>>>>> generates<br>
>> >>>> >>>>>>>>> a CIC deposition for the whole domain in about 6 seconds<br>
>> >>>> >>>>>>>>> from a cold start.<br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> If you're using 2.x, then you can do the same thing, but<br>
>> >>>> >>>>>>>>>> it's<br>
>> >>>> >>>>>>>>>> not as straightforward. You can see how this works in<br>
>> >>>> >>>>>>>>>> <a href="http://yt.data_objects.universal_fields.py" target="_blank">yt.data_objects.universal_fields.py</a>, around line 986,<br>
>> >>>> >>>>>>>>>> where the<br>
>> >>>> >>>>>>>>>> "particle_density" field is defined. Basically, it calls<br>
>> >>>> >>>>>>>>>> CICDeposit_3, which<br>
>> >>>> >>>>>>>>>> is in yt.utilities.lib.CICDeposit.pyx.<br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> Let me know if you need any more clarification.<br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> Best,<br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> John Z<br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> On Jun 5, 2014, at 6:07 PM, Brendan Griffen<br>
>> >>>> >>>>>>>>>> <<a href="mailto:brendan.f.griffen@gmail.com" target="_blank">brendan.f.griffen@gmail.com</a>> wrote:<br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> > Hi,<br>
>> >>>> >>>>>>>>>> ><br>
>> >>>> >>>>>>>>>> > I was wondering if there were any Cython routines within<br>
>> >>>> >>>>>>>>>> > yt<br>
>> >>>> >>>>>>>>>> > which takes particle data and converts it into a<br>
>> >>>> >>>>>>>>>> > cloud-in-cell based mesh<br>
>> >>>> >>>>>>>>>> > which can be written to a file of my choosing.<br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>> What sort of mesh were you looking for? yt will internally<br>
>> >>>> >>>>>>>>> construct an octree if it is fed particle data. I'm not<br>
>> >>>> >>>>>>>>> sure whether this<br>
>> >>>> >>>>>>>>> octree can be saved to disk for later analysis.<br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>> It's also possible to create a uniform resolution covering<br>
>> >>>> >>>>>>>>> grid containing field data for a deposited quantity, which<br>
>> >>>> >>>>>>>>> can be quite<br>
>> >>>> >>>>>>>>> easily saved to disk in a number of ways.<br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> I heard a while ago there was some such functionality but<br>
>> >>>> >>>>>>>>>> it<br>
>> >>>> >>>>>>>>>> could be too far down the yt rabbit hole to be used as a<br>
>> >>>> >>>>>>>>>> standalone? Is this<br>
>> >>>> >>>>>>>>>> true? I have my own Python code for doing it but it just<br>
>> >>>> >>>>>>>>>> isn't fast enough<br>
>> >>>> >>>>>>>>>> and thought I'd ask the yt community if there were any<br>
>> >>>> >>>>>>>>>> wrapper tools<br>
>> >>>> >>>>>>>>>> available to boost the speed.<br>
>> >>>> >>>>>>>>>> ><br>
>> >>>> >>>>>>>>>> > Thanks.<br>
>> >>>> >>>>>>>>>> > Brendan<br>
>> >>>> >>>>>>>>>> > _______________________________________________<br>
>> >>>> >>>>>>>>>> > yt-users mailing list<br>
>> >>>> >>>>>>>>>> > <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> >>>>>>>>>> ><br>
>> >>>> >>>>>>>>>> ><br>
>> >>>> >>>>>>>>>> > <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> _______________________________________________<br>
>> >>>> >>>>>>>>>> yt-users mailing list<br>
>> >>>> >>>>>>>>>> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>><br>
>> >>>> >>>>>>>>>> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>> _______________________________________________<br>
>> >>>> >>>>>>>>> yt-users mailing list<br>
>> >>>> >>>>>>>>> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>>> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>> >>>>>>>>><br>
>> >>>> >>>>>>>><br>
>> >>>> >>>>>>>><br>
>> >>>> >>>>>>>> _______________________________________________<br>
>> >>>> >>>>>>>> yt-users mailing list<br>
>> >>>> >>>>>>>> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> >>>>>>>><br>
>> >>>> >>>>>>>> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>> >>>>>>>><br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>>> _______________________________________________<br>
>> >>>> >>>>>>> yt-users mailing list<br>
>> >>>> >>>>>>> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>>> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>> >>>>>>><br>
>> >>>> >>>>>><br>
>> >>>> >>>>>><br>
>> >>>> >>>>>> _______________________________________________<br>
>> >>>> >>>>>> yt-users mailing list<br>
>> >>>> >>>>>> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> >>>>>> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>> >>>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>><br>
>> >>>> >>>>> _______________________________________________<br>
>> >>>> >>>>> yt-users mailing list<br>
>> >>>> >>>>> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> >>>>> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>> >>>>><br>
>> >>>> >>>><br>
>> >>>> >>>><br>
>> >>>> >>>> _______________________________________________<br>
>> >>>> >>>> yt-users mailing list<br>
>> >>>> >>>> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> >>>> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>> >>>><br>
>> >>>> >>><br>
>> >>>> >>><br>
>> >>>> >>> _______________________________________________<br>
>> >>>> >>> yt-users mailing list<br>
>> >>>> >>> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> >>> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>> >>><br>
>> >>>> >><br>
>> >>>> >><br>
>> >>>> >> _______________________________________________<br>
>> >>>> >> yt-users mailing list<br>
>> >>>> >> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> >> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>> >><br>
>> >>>> ><br>
>> >>>> ><br>
>> >>>> > _______________________________________________<br>
>> >>>> > yt-users mailing list<br>
>> >>>> > <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> > <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>> ><br>
>> >>>><br>
>> >>>><br>
>> >>>> _______________________________________________<br>
>> >>>> yt-users mailing list<br>
>> >>>> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >>>> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >>>><br>
>> >>><br>
>> >><br>
>> >><br>
>> >> _______________________________________________<br>
>> >> yt-users mailing list<br>
>> >> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> >> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> >><br>
>> ><br>
>> ><br>
>> > _______________________________________________<br>
>> > yt-users mailing list<br>
>> > <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> > <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
>> ><br>
>> _______________________________________________<br>
>> yt-users mailing list<br>
>> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
>> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
><br>
><br>
><br>
> _______________________________________________<br>
> yt-users mailing list<br>
> <a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
> <a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
><br>
_______________________________________________<br>
yt-users mailing list<br>
<a href="mailto:yt-users@lists.spacepope.org" target="_blank">yt-users@lists.spacepope.org</a><br>
<a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
</div></div></blockquote></div><br></div>
</div></div><br>_______________________________________________<br>
yt-users mailing list<br>
<a href="mailto:yt-users@lists.spacepope.org">yt-users@lists.spacepope.org</a><br>
<a href="http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org" target="_blank">http://lists.spacepope.org/listinfo.cgi/yt-users-spacepope.org</a><br>
<br></blockquote></div><br></div>