Aquila-Consortium/csiborgtools
0
0
Fork 0
mirror of https://github.com/Richard-Sti/csiborgtools_public.git synced 2025-05-12 13:41:13 +00:00
Code Issues Projects Releases Packages Wiki Activity Actions

Add density field plot and start preparing CSiBORG2 (#94)

Browse source 
* Add RAMSES2HDF5 conversion

* Upload changes

* Clean up

* More clean up

* updates

* Little change

* pep9

* Add basic SPH calculation for a snapshot

* Add submit script

* Remove echo

* Little changes

* Send off changes

* Little formatting

* Little updates

* Add nthreads argument

* Upload chagnes

* Add nthreads arguemnts

* Some local changes..

* Update scripts

* Add submission script

* Update script

* Update params

* Rename CSiBORGBox to CSiBORG1box

* Rename CSiBORG1 reader

* Move files

* Rename folder again

* Add basic plotting here

* Add new skeletons

* Move def

* Update nbs

* Edit directories

* Rename files

* Add units to converted snapshots

* Fix empty dataset bug

* Delete file

* Edits to submission scripts

* Edit paths

* Update .gitignore

* Fix attrs

* Update weighting

* Fix RA/dec bug

* Add FORNAX cluster

* Little edit

* Remove boxes since will no longer need

* Move func back

* Edit to include sort by membership

* Edit paths

* Purge basic things

* Start removing

* Bring file back

* Scratch

* Update the rest

* Improve the entire file

* Remove old things

* Remove old

* Delete old things

* Fully updates

* Rename file

* Edit submit script

* Little things

* Add print statement

* Add here cols_to_structured

* Edit halo cat

* Remove old import

* Add comment

* Update paths manager

* Move file

* Remove file

* Add chains
This commit is contained in:
Richard Stiskalek 2023-12-13 16:08:25 +00:00 • committed by GitHub
parent 6042a87111
commit aaa14fc880
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
30 changed files with 1682 additions and 1728 deletions
Download patch file Download diff file Expand all files Collapse all files

108
scripts_independent/field_obs_vp.py Normal file
View file

@ -0,0 +1,108 @@
# Copyright (C) 2022 Richard Stiskalek
# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 3 of the License, or (at your
# option) any later version.
#
# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
# Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
"""
Script to calculate the peculiar velocity of an observer in the centre of the
CSiBORG box.
"""
from argparse import ArgumentParser
from distutils.util import strtobool
import numpy
from mpi4py import MPI
from taskmaster import work_delegation
from tqdm import tqdm
from utils import get_nsims
try:
import csiborgtools
except ModuleNotFoundError:
import sys
sys.path.append("../")
import csiborgtools
def observer_peculiar_velocity(nsim, parser_args):
"""
Calculate the peculiar velocity of an observer in the centre of the box
for several smoothing scales.
"""
pos = numpy.array([0.5, 0.5, 0.5]).reshape(-1, 3)
boxsize = 677.7
smooth_scales = [0, 0.5, 1.0, 1.5, 2.0, 2.5, 3.0, 3.5, 4.0]
observer_vp = numpy.full((len(smooth_scales), 3), numpy.nan,
dtype=numpy.float32)
paths = csiborgtools.read.Paths(**csiborgtools.paths_glamdring)
field_path = paths.field("velocity", parser_args.MAS, parser_args.grid,
nsim, in_rsp=False)
field0 = numpy.load(field_path)
for j, smooth_scale in enumerate(tqdm(smooth_scales,
desc="Smoothing the fields",
disable=not parser_args.verbose)):
if smooth_scale > 0:
field = [None, None, None]
for k in range(3):
field[k] = csiborgtools.field.smoothen_field(
field0[k], smooth_scale, boxsize)
else:
field = field0
v = csiborgtools.field.evaluate_cartesian(
field[0], field[1], field[2], pos=pos)
observer_vp[j, 0] = v[0][0]
observer_vp[j, 1] = v[1][0]
observer_vp[j, 2] = v[2][0]
fout = paths.observer_peculiar_velocity(parser_args.MAS, parser_args.grid,
nsim)
if parser_args.verbose:
print(f"Saving to ... `{fout}`")
numpy.savez(fout, smooth_scales=smooth_scales, observer_vp=observer_vp)
return observer_vp
###############################################################################
# Command line interface #
###############################################################################
if __name__ == "__main__":
parser = ArgumentParser()
parser.add_argument("--nsims", type=int, nargs="+", default=None,
help="IC realisations. `-1` for all simulations.")
parser.add_argument("--kind", type=str,
choices=["density", "rspdensity", "velocity", "radvel",
"potential", "environment"],
help="What derived field to calculate?")
parser.add_argument("--MAS", type=str,
choices=["NGP", "CIC", "TSC", "PCS"])
parser.add_argument("--grid", type=int, help="Grid resolution.")
parser.add_argument("--verbose", type=lambda x: bool(strtobool(x)),
help="Verbosity flag for reading in particles.")
parser.add_argument("--simname", type=str, default="csiborg",
help="Verbosity flag for reading in particles.")
parser_args = parser.parse_args()
comm = MPI.COMM_WORLD
paths = csiborgtools.read.Paths(**csiborgtools.paths_glamdring)
nsims = get_nsims(parser_args, paths)
def main(nsim):
return observer_peculiar_velocity(nsim, parser_args)
work_delegation(main, nsims, comm, master_verbose=True)

201
scripts_independent/field_sph.py Normal file
View file

@ -0,0 +1,201 @@
# Copyright (C) 2023 Richard Stiskalek
# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 3 of the License, or (at your
# option) any later version.
#
# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
# Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
"""
Script to construct the density and velocity fields for a simulation snapshot.
The SPH filter is implemented in the cosmotool package.
"""
from argparse import ArgumentParser
from os import environ, remove
from os.path import join, exists
import subprocess
from datetime import datetime
import hdf5plugin # noqa
import numpy as np
from h5py import File
def now():
return datetime.now()
def generate_unique_id(file_path):
"""
Generate a unique ID for a file path.
"""
return file_path.replace('/', '_').replace(':', '_')
def prepare_random(temporary_output_path, npart=100, dtype=np.float32):
"""
Prepare a random dataset for the SPH filter.
"""
print("Preparing random dataset.", flush=True)
arr = np.full((npart, 7), np.nan, dtype=dtype)
arr[:, :3] = np.random.uniform(0, 1, (npart, 3))
arr[:, 3:6] = np.random.normal(0, 1, (npart, 3))
arr[:, 6] = np.ones(npart, dtype=dtype)
dset = np.random.random((npart, 7)).astype(dtype)
dset[:, 6] = np.ones(npart, dtype=dtype)
with File(temporary_output_path, 'w') as target:
target.create_dataset("particles", data=dset, dtype=dtype)
return 1.
def prepare_gadget(snapshot_path, temporary_output_path):
"""
Prepare a GADGET snapshot for the SPH filter. Assumes there is only a
single file per snapshot.
"""
with File(snapshot_path, 'r') as source, File(temporary_output_path, 'w') as target: # noqa
boxsize = source["Header"].attrs["BoxSize"]
npart = sum(source["Header"].attrs["NumPart_Total"])
nhighres = source["Header"].attrs["NumPart_Total"][1]
dset = target.create_dataset("particles", (npart, 7), dtype=np.float32)
# Copy to this dataset the high-resolution particles.
dset[:nhighres, :3] = source["PartType1/Coordinates"][:]
dset[:nhighres, 3:6] = source["PartType1/Velocities"][:]
dset[:nhighres, 6] = np.ones(nhighres, dtype=np.float32) * source["Header"].attrs["MassTable"][1] # noqa
# Now copy the low-resolution particles.
dset[nhighres:, :3] = source["PartType5/Coordinates"][:]
dset[nhighres:, 3:6] = source["PartType5/Velocities"][:]
dset[nhighres:, 6] = source["PartType5/Masses"][:]
return boxsize
def run_sph_filter(particles_path, output_path, boxsize, resolution,
SPH_executable):
"""
Run the SPH filter on a snapshot.
"""
if not exists(particles_path):
raise RuntimeError(f"Particles file `{particles_path}` does not exist.") # noqa
if not isinstance(boxsize, (int, float)):
raise TypeError("`boxsize` must be a number.")
if not isinstance(resolution, int):
raise TypeError("`resolution` must be an integer.")
if not exists(SPH_executable):
raise RuntimeError(f"SPH executable `{SPH_executable}` does not exist.") # noqa
command = [SPH_executable, particles_path, str(1e14), str(boxsize),
str(resolution), str(0), str(0), str(0), output_path, "1"]
print(f"{now()}: executing `simple3DFilter`.", flush=True)
start_time = now()
process = subprocess.Popen(
command, stdout=subprocess.PIPE, stderr=subprocess.STDOUT,
universal_newlines=True)
for line in iter(process.stdout.readline, ""):
print(line, end="", flush=True)
process.wait()
if process.returncode != 0:
raise RuntimeError("`simple3DFilter`failed.")
else:
dt = now() - start_time
print(f"{now()}: `simple3DFilter`completed successfully in {dt}.",
flush=True)
def main(snapshot_path, output_path, resolution, scratch_space, SPH_executable,
snapshot_kind):
"""
Construct the density and velocity fields for a simulation snapshot using
`cosmotool` [1].
Parameters
----------
snapshot_path : str
Path to the simulation snapshot.
output_path : str
Path to the output HDF5 file.
resolution : int
Resolution of the density field.
scratch_space : str
Path to a folder where temporary files can be stored.
SPH_executable : str
Path to the `simple3DFilter` executable [1].
snapshot_kind : str
Kind of the simulation snapshot. Currently only `gadget4` is supported.
Returns
-------
None
References
----------
[1] https://bitbucket.org/glavaux/cosmotool/src/master/sample/simple3DFilter.cpp # noqa
"""
if snapshot_kind != "gadget4":
raise NotImplementedError("Only GADGET HDF5 snapshots are supported.")
print("---------- SPH Density & Velocity Field Job Information ----------")
print(f"Snapshot path: {snapshot_path}")
print(f"Output path: {output_path}")
print(f"Resolution: {resolution}")
print(f"Scratch space: {scratch_space}")
print(f"SPH executable: {SPH_executable}")
print(f"Snapshot kind: {snapshot_kind}")
print("------------------------------------------------------------------")
print(flush=True)
temporary_output_path = join(
scratch_space, generate_unique_id(snapshot_path))
if not temporary_output_path.endswith(".hdf5"):
raise RuntimeError("Temporary output path must end with `.hdf5`.")
print(f"{now()}: preparing snapshot...", flush=True)
boxsize = prepare_gadget(snapshot_path, temporary_output_path)
print(f"{now()}: wrote temporary data to {temporary_output_path}.",
flush=True)
run_sph_filter(temporary_output_path, output_path, boxsize, resolution,
SPH_executable)
print(f"{now()}: removing the temporary snapshot file.", flush=True)
try:
remove(temporary_output_path)
except FileNotFoundError:
pass
if __name__ == "__main__":
parser = ArgumentParser(description="Generate SPH density and velocity field.") # noqa
parser.add_argument("--snapshot_path", type=str, required=True,
help="Path to the simulation snapshot.")
parser.add_argument("--output_path", type=str, required=True,
help="Path to the output HDF5 file.")
parser.add_argument("--resolution", type=int, required=True,
help="Resolution of the density and velocity field.")
parser.add_argument("--scratch_space", type=str, required=True,
help="Path to a folder where temporary files can be stored.") # noqa
parser.add_argument("--SPH_executable", type=str, required=True,
help="Path to the `simple3DFilter` executable.")
parser.add_argument("--snapshot_kind", type=str, required=True,
choices=["gadget4"],
help="Kind of the simulation snapshot.")
args = parser.parse_args()
main(args.snapshot_path, args.output_path, args.resolution,
args.scratch_space, args.SPH_executable, args.snapshot_kind)

40
scripts_independent/field_sph.sh Executable file
View file

@ -0,0 +1,40 @@
#!/bin/sh
#SBATCH --ntasks-per-node=1
#SBATCH --nodes=1
#SBATCH --cpus-per-task=16
#SBATCH --mem-per-cpu=7000
#SBATCH -J SPH
#SBATCH -o output_%J.out
#SBATCH -e error_%J.err
#SBATCH -p cosma8-serial
#SBATCH -A dp016
#SBATCH -t 04:00:00
#SBATCH --mail-type=BEGIN,END,FAIL
#SBATCH --mail-user=richard.stiskalek@physics.ox.ac.uk
module purge
module load intel_comp/2019
module load intel_mpi/2019
module load hdf5
module load fftw
module load gsl
module load cmake
module load python/3.10.12
module list
source /cosma/home/dp016/dc-stis1/csiborgtools/venv_csiborgtools/bin/activate
export OMP_NUM_THREADS=16
export OMP_NESTED=true
# ADD CHAINS HERE
snapshot_path="/cosma8/data/dp016/dc-stis1/csiborg2_main/chain_15517/output/snapshot_099_full.hdf5"
output_path="/cosma8/data/dp016/dc-stis1/csiborg2_main/field/chain_15517.hdf5"
resolution=256
scratch_space="/cosma8/data/dp016/dc-stis1/csiborg2_main/field"
SPH_executable="/cosma8/data/dp016/dc-stis1/cosmotool/bld2/sample/simple3DFilter"
snapshot_kind="gadget4"
python3 field_sph.py --snapshot_path $snapshot_path --output_path $output_path --resolution $resolution --scratch_space $scratch_space --SPH_executable $SPH_executable --snapshot_kind $snapshot_kind

737
scripts_independent/process_snapshot.py Normal file
View file

@ -0,0 +1,737 @@
# Copyright (C) 2023 Richard Stiskalek
# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 3 of the License, or (at your
# option) any later version.
#
# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
# Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
"""
Script to process simulation snapshots to sorted HDF5 files. Be careful
because reading the HDF5 file may require `hdf5plugin` package to be installed.
The snapshot particles are sorted by their halo ID, so that particles of a halo
can be accessed by slicing the array.
CSiBORG1 reader will complain unless it can find the halomaker FOF files
where it expects them:
fdir = f"/mnt/extraspace/rstiskalek/csiborg1/chain_{self.nsim}/FOF"
"""
from abc import ABC, abstractmethod
from argparse import ArgumentParser
from datetime import datetime
from gc import collect
from glob import glob, iglob
from os import makedirs
from os.path import basename, dirname, exists, join
from warnings import catch_warnings, filterwarnings, warn
import hdf5plugin
import numpy
import pynbody
import readgadget
from astropy import constants, units
from h5py import File
from numba import jit
from readfof import FoF_catalog
from tqdm import tqdm, trange
MSUNCGS = constants.M_sun.cgs.value
BLOSC_KWARGS = {"cname": "blosclz",
"clevel": 9,
"shuffle": hdf5plugin.Blosc.SHUFFLE,
}
###############################################################################
# Utility functions #
###############################################################################
def now():
"""
Return current time.
"""
return datetime.now()
def flip_cols(arr, col1, col2):
"""
Flip values in columns `col1` and `col2` of a structured array `arr`.
"""
if col1 not in arr.dtype.names or col2 not in arr.dtype.names:
raise ValueError(f"Both `{col1}` and `{col2}` must exist in `arr`.")
arr[col1], arr[col2] = numpy.copy(arr[col2]), numpy.copy(arr[col1])
def convert_str_to_num(s):
"""
Convert a string representation of a number to its appropriate numeric type
(int or float).
Parameters
----------
s : str
The string representation of the number.
Returns
-------
num : int or float
"""
try:
return int(s)
except ValueError:
try:
return float(s)
except ValueError:
warn(f"Cannot convert string '{s}' to number", UserWarning)
return s
def cols_to_structured(N, cols):
"""
Allocate a structured array from `cols`, a list of (name, dtype) tuples.
"""
if not (isinstance(cols, list)
and all(isinstance(c, tuple) and len(c) == 2 for c in cols)):
raise TypeError("`cols` must be a list of (name, dtype) tuples.")
names, formats = zip(*cols)
dtype = {"names": names, "formats": formats}
return numpy.full(N, numpy.nan, dtype=dtype)
###############################################################################
# Base reader of snapshots #
###############################################################################
class BaseReader(ABC):
"""Base reader layout that every subsequent reader should follow."""
@abstractmethod
def read_info(self):
pass
@abstractmethod
def read_snapshot(self, kind):
pass
@abstractmethod
def read_halo_id(self, pids):
pass
@abstractmethod
def read_halos(self):
pass
###############################################################################
# CSiBORG particle reader #
###############################################################################
class CSiBORG1Reader:
"""
Object to read in CSiBORG snapshots from the binary files and halo
catalogues.
Parameters
----------
nsim : int
IC realisation index.
which_snapshot : str
Which snapshot to read. Options are `initial` or `final`.
"""
def __init__(self, nsim, which_snapshot):
self.nsim = nsim
base_dir = "/mnt/extraspace/hdesmond/"
if which_snapshot == "initial":
self.nsnap = 1
raise RuntimeError("TODO not implemented")
self.source_dir = None
elif which_snapshot == "final":
sourcedir = join(base_dir, f"ramses_out_{nsim}")
self.nsnap = max([int(basename(f).replace("output_", ""))
for f in glob(join(sourcedir, "output_*"))])
self.source_dir = join(sourcedir,
f"output_{str(self.nsnap).zfill(5)}")
else:
raise ValueError(f"Unknown snapshot option `{which_snapshot}`.")
self.output_dir = f"/mnt/extraspace/rstiskalek/csiborg1/chain_{self.nsim}" # noqa
self.output_snap = join(self.output_dir,
f"snapshot_{str(self.nsnap).zfill(5)}.hdf5")
self.output_cat = join(self.output_dir,
f"fof_{str(self.nsnap).zfill(5)}.hdf5")
self.halomaker_dir = join(self.output_dir, "FOF")
def read_info(self):
filename = glob(join(self.source_dir, "info_*"))
if len(filename) > 1:
raise ValueError("Found too many `info` files.")
filename = filename[0]
with open(filename, "r") as f:
info = f.read().split()
# Throw anything below ordering line out
info = numpy.asarray(info[:info.index("ordering")])
# Get indexes of lines with `=`. Indxs before/after be keys/vals
eqs = numpy.asarray([i for i in range(info.size) if info[i] == '='])
keys = info[eqs - 1]
vals = info[eqs + 1]
return {key: convert_str_to_num(val) for key, val in zip(keys, vals)}
def read_snapshot(self, kind):
with catch_warnings():
filterwarnings("ignore", category=UserWarning)
sim = pynbody.load(self.source_dir)
info = self.read_info()
if kind == "pid":
x = numpy.array(sim["iord"], dtype=numpy.uint32)
elif kind == "pos":
x = numpy.array(sim[kind], dtype=numpy.float32)
# Convert box units to Mpc / h
box2mpc = (info["unit_l"] / units.kpc.to(units.cm) / info["aexp"]
* 1e-3 * info["H0"] / 100)
x *= box2mpc
elif kind == "mass":
x = numpy.array(sim[kind], dtype=numpy.float32)
# Convert box units to Msun / h
box2msun = (info["unit_d"] * info["unit_l"]**3 / MSUNCGS
* info["H0"] / 100)
x *= box2msun
elif kind == "vel":
x = numpy.array(sim[kind], dtype=numpy.float16)
# Convert box units to km / s
box2kms = (1e-2 * info["unit_l"] / info["unit_t"] / info["aexp"]
* 1e-3)
x *= box2kms
else:
raise ValueError(f"Unknown kind `{kind}`. "
"Options are: `pid`, `pos`, `vel` or `mass`.")
# Because of a RAMSES bug x and z are flipped.
if kind in ["pos", "vel"]:
print(f"For kind `{kind}` flipping x and z.")
x[:, [0, 2]] = x[:, [2, 0]]
del sim
collect()
return x
def read_halo_id(self, pids):
fpath = join(self.halomaker_dir, "*particle_membership*")
fpath = next(iglob(fpath, recursive=True), None)
if fpath is None:
raise FileNotFoundError(f"Found no Halomaker files in `{self.halomaker_dir}`.") # noqa
print(f"{now()}: mapping particle IDs to their indices.")
pids_idx = {pid: i for i, pid in enumerate(pids)}
# Unassigned particle IDs are assigned a halo ID of 0.
print(f"{now()}: mapping HIDs to their array indices.")
hids = numpy.zeros(pids.size, dtype=numpy.int32)
# Read line-by-line to avoid loading the whole file into memory.
with open(fpath, 'r') as file:
for line in tqdm(file, desc="Reading membership"):
hid, pid = map(int, line.split())
hids[pids_idx[pid]] = hid
del pids_idx
collect()
return hids
def read_halos(self):
info = self.read_info()
h = info["H0"] / 100
fpath = join(self.halomaker_dir, "fort.132")
hid = numpy.genfromtxt(fpath, usecols=0, dtype=numpy.int32)
pos = numpy.genfromtxt(fpath, usecols=(1, 2, 3), dtype=numpy.float32)
totmass = numpy.genfromtxt(fpath, usecols=4, dtype=numpy.float32)
m200c = numpy.genfromtxt(fpath, usecols=5, dtype=numpy.float32)
dtype = {"names": ["index", "x", "y", "z", "totpartmass", "m200c"],
"formats": [numpy.int32] + [numpy.float32] * 5}
out = numpy.full(hid.size, numpy.nan, dtype=dtype)
out["index"] = hid
out["x"] = pos[:, 0] * h + 677.7 / 2
out["y"] = pos[:, 1] * h + 677.7 / 2
out["z"] = pos[:, 2] * h + 677.7 / 2
# Because of a RAMSES bug x and z are flipped.
flip_cols(out, "x", "z")
out["totpartmass"] = totmass * 1e11 * h
out["m200c"] = m200c * 1e11 * h
return out
###############################################################################
# CSiBORG2 particle reader #
###############################################################################
class CSiBORG2Reader(BaseReader):
"""
Object to read in CSiBORG2 snapshots. Because this is Gadget4 the final
snapshot is already sorted, however we still have to sort the initial
snapshot.
Parameters
----------
nsim : int
IC realisation index.
which_snapshot : str
Which snapshot to read. Options are `initial` or `final`.
"""
def __init__(self, nsim, which_snapshot, kind):
self.nsim = nsim
if kind not in ["main", "random", "varysmall"]:
raise ValueError(f"Unknown kind `{kind}`.")
base_dir = f"/mnt/extraspace/rstiskalek/csiborg2_{kind}"
if which_snapshot == "initial":
self.nsnap = 0
elif which_snapshot == "final":
self.nsnap = 99
else:
raise ValueError(f"Unknown snapshot option `{which_snapshot}`.")
self.source_dir = join(
base_dir, f"chain_{nsim}", "output",
f"snapshot_{str(self.nsnap).zfill(3)}_full.hdf5")
self.output_dir = join(base_dir, f"chain_{nsim}", "output")
self.output_snap = join(
self.output_dir,
f"snapshot_{str(self.nsnap).zfill(3)}_sorted.hdf5")
self.output_cat = None
def read_info(self):
fpath = join(dirname(self.source_dir), "snapshot_99_full.hdf5")
with File(fpath, 'r') as f:
header = f["Header"]
params = f["Parameters"]
out = {"BoxSize": header.attrs["BoxSize"],
"MassTable": header.attrs["MassTable"],
"NumPart_Total": header.attrs["NumPart_Total"],
"Omega_m": params.attrs["Omega0"],
"Omega_l": params.attrs["OmegaLambda"],
"Omega_b": params.attrs["OmegaBaryon"],
"h": params.attrs["HubbleParam"],
"redshift": header.attrs["Redshift"],
}
return out
def read_snapshot(self, kind):
raise RuntimeError("TODO Not implemented.")
def read_halo_id(self, pids):
raise RuntimeError("TODO Not implemented.")
def read_halos(self):
raise RuntimeError("TODO Not implemented.")
###############################################################################
# Quijote particle reader #
###############################################################################
class QuijoteReader:
"""
Object to read in Quijote snapshots from the binary files.
Parameters
----------
nsim : int
IC realisation index.
which_snapshot : str
Which snapshot to read. Options are `initial` or `final`.
"""
def __init__(self, nsim, which_snapshot):
self.nsim = nsim
quijote_dir = "/mnt/extraspace/rstiskalek/quijote"
if which_snapshot == "initial":
self.nsnap = -1
snap_str = "ICs"
self.source_dir = join(quijote_dir, "Snapshots_fiducial",
str(nsim), "ICs", "ics")
elif which_snapshot == "final":
self.nsnap = 4
snap_str = str(self.nsnap).zfill(3)
self.source_dir = join(
quijote_dir, "Snapshots_fiducial",
str(nsim), f"snapdir_{snap_str}", f"snap_{snap_str}")
else:
raise ValueError(f"Unknown snapshot option `{which_snapshot}`.")
self.fof_dir = join(quijote_dir, "Halos_fiducial", str(nsim))
self.output_dir = f"/mnt/extraspace/rstiskalek/quijote/fiducial_processed/chain_{self.nsim}" # noqa
self.output_snap = join(self.output_dir, f"snapshot_{snap_str}.hdf5")
self.output_cat = join(self.output_dir, f"fof_{snap_str}.hdf5")
def read_info(self):
header = readgadget.header(self.source_dir)
out = {"BoxSize": header.boxsize / 1e3, # Mpc/h
"Nall": header.nall[1], # Tot num of particles
"PartMass": header.massarr[1] * 1e10, # Part mass in Msun/h
"Omega_m": header.omega_m,
"Omega_l": header.omega_l,
"h": header.hubble,
"redshift": header.redshift,
}
out["TotMass"] = out["Nall"] * out["PartMass"]
out["Hubble"] = (100.0 * numpy.sqrt(
header.omega_m * (1.0 + header.redshift)**3 + header.omega_l))
return out
def read_snapshot(self, kind):
info = self.read_info()
ptype = [1] # DM
if kind == "pid":
return readgadget.read_block(self.source_dir, "ID ", ptype)
elif kind == "pos":
pos = readgadget.read_block(self.source_dir, "POS ", ptype) / 1e3
return pos.astype(numpy.float32)
elif kind == "vel":
vel = readgadget.read_block(self.source_dir, "VEL ", ptype)
vel = vel.astype(numpy.float16)
vel *= (1 + info["redshift"]) # km / s
return vel
elif kind == "mass":
return numpy.full(info["Nall"], info["PartMass"],
dtype=numpy.float32)
else:
raise ValueError(f"Unknown kind `{kind}`. "
"Options are: `pid`, `pos`, `vel` or `mass`.")
def read_halo_id(self, pids):
cat = FoF_catalog(self.fof_dir, self.nsnap)
group_pids = cat.GroupIDs
group_len = cat.GroupLen
# Create a mapping from particle ID to FoF group ID.
print(f"{now()}: mapping particle IDs to their indices.")
ks = numpy.insert(numpy.cumsum(group_len), 0, 0)
with catch_warnings():
# Ignore because we are casting NaN as integer.
filterwarnings("ignore", category=RuntimeWarning)
pid2hid = numpy.full((group_pids.size, 2), numpy.nan,
dtype=numpy.uint64)
for i, (k0, kf) in enumerate(zip(ks[:-1], ks[1:])):
pid2hid[k0:kf, 0] = i + 1
pid2hid[k0:kf, 1] = group_pids[k0:kf]
pid2hid = {pid: hid for hid, pid in pid2hid}
# Create the final array of hids matchign the snapshot array.
# Unassigned particles have hid 0.
print(f"{now()}: mapping HIDs to their array indices.")
hids = numpy.full(pids.size, 0, dtype=numpy.uint32)
for i in trange(pids.size):
hids[i] = pid2hid.get(pids[i], 0)
return hids
def read_halos(self):
fof = FoF_catalog(self.fof_dir, self.nsnap, long_ids=False, swap=False,
SFR=False, read_IDs=False)
cols = [("x", numpy.float32),
("y", numpy.float32),
("z", numpy.float32),
("vx", numpy.float32),
("vy", numpy.float32),
("vz", numpy.float32),
("GroupMass", numpy.float32),
("npart", numpy.int32),
("index", numpy.int32)
]
data = cols_to_structured(fof.GroupLen.size, cols)
pos = fof.GroupPos / 1e3
vel = fof.GroupVel * (1 + self.read_info()["redshift"])
for i, p in enumerate(["x", "y", "z"]):
data[p] = pos[:, i]
data[f"v{p}"] = vel[:, i]
data["GroupMass"] = fof.GroupMass * 1e10
data["npart"] = fof.GroupLen
# We want to start indexing from 1. Index 0 is reserved for
# particles unassigned to any FoF group.
data["index"] = 1 + numpy.arange(data.size, dtype=numpy.uint32)
return data
###############################################################################
# Group Offsets #
###############################################################################
@jit(nopython=True, boundscheck=False)
def minmax_halo(hid, halo_ids, start_loop=0):
"""
Find the start and end index of a halo in a sorted array of halo IDs.
This is much faster than using `numpy.where` and then `numpy.min` and
`numpy.max`.
"""
start = None
end = None
for i in range(start_loop, halo_ids.size):
n = halo_ids[i]
if n == hid:
if start is None:
start = i
end = i
elif n > hid:
break
return start, end
def make_offset_map(part_hids):
"""
Make group offsets for a list of particles' halo IDs. This is a
2-dimensional array, where the first column is the halo ID, the second
column is the start index of the halo in the particle list, and the third
index is the end index of the halo in the particle list. The start index is
inclusive, while the end index is exclusive.
"""
unique_halo_ids = numpy.unique(part_hids)
unique_halo_ids = unique_halo_ids[unique_halo_ids != 0]
with catch_warnings():
filterwarnings("ignore", category=RuntimeWarning)
halo_map = numpy.full((unique_halo_ids.size, 3), numpy.nan,
dtype=numpy.uint32)
start_loop, niters = 0, unique_halo_ids.size
for i in trange(niters):
hid = unique_halo_ids[i]
k0, kf = minmax_halo(hid, part_hids, start_loop=start_loop)
halo_map[i, :] = hid, k0, kf
start_loop = kf
return halo_map, unique_halo_ids
###############################################################################
# Process the final snapshot and sort it by groups #
###############################################################################
def process_final_snapshot(nsim, simname):
"""
Read in the snapshot particles, sort them by their halo ID and dump
into a HDF5 file. Stores the first and last index of each halo in the
particle array for fast slicing of the array to acces particles of a single
halo.
"""
if simname == "csiborg1":
reader = CSiBORG1Reader(nsim, "final")
elif simname == "quijote":
reader = QuijoteReader(nsim, "final")
else:
raise RuntimeError(f"Simulation `{simname}` is not supported.")
if not exists(reader.output_dir):
makedirs(reader.output_dir)
print("---- Processing Final Snapshot Information ----")
print(f"Simulation index: {nsim}")
print(f"Simulation name: {simname}")
print(f"Output snapshot: {reader.output_snap}")
print(f"Output catalogue: {reader.output_cat}")
print("-----------------------------------------------")
print(flush=True)
# First off load the particle IDs from the raw data.
pids = reader.read_snapshot("pid")
# Then, load the halo ids and make sure their ordering is the same as the
# particle IDs ordering.
print(f"{now()}: loading HIDs.")
halo_ids = reader.read_halo_id(pids)
print(f"{now()}: sorting HIDs.")
# Get a mask that sorts the halo ids and then write the information to
# the data files sorted by it.
sort_indxs = numpy.argsort(halo_ids)
halo_ids = halo_ids[sort_indxs]
with File(reader.output_snap, 'w') as f:
print(f"{now()}: creating dataset `ParticleIDs`...",
flush=True)
f.create_dataset("ParticleIDs", data=pids[sort_indxs],
**hdf5plugin.Blosc(**BLOSC_KWARGS))
del pids
collect()
print(f"{now()}: creating dataset `Coordinates`...",
flush=True)
f.create_dataset(
"Coordinates", data=reader.read_snapshot("pos")[sort_indxs],
**hdf5plugin.Blosc(**BLOSC_KWARGS))
print(f"{now()}: creating dataset `Velocities`...",
flush=True)
f.create_dataset(
"Velocities", data=reader.read_snapshot("vel")[sort_indxs],
**hdf5plugin.Blosc(**BLOSC_KWARGS))
print(f"{now()}: creating dataset `Masses`...",
flush=True)
f.create_dataset(
"Masses", data=reader.read_snapshot("mass")[sort_indxs],
**hdf5plugin.Blosc(**BLOSC_KWARGS))
if simname == "csiborg1":
header = f.create_dataset("Header", (0,))
header.attrs["BoxSize"] = 677.7 # Mpc/h
header.attrs["Omega0"] = 0.307
header.attrs["OmegaBaryon"] = 0.0
header.attrs["OmegaLambda"] = 0.693
header.attrs["HubleParam"] = 0.6777
header.attrs["Redshift"] = 0.0
elif simname == "quijote":
info = reader.read_info()
header = f.create_dataset("Header", (0,))
header.attrs["BoxSize"] = info["BoxSize"]
header.attrs["Omega0"] = info["Omega_m"]
header.attrs["OmegaLambda"] = info["Omega_l"]
header.attrs["OmegaBaryon"] = 0.0
header.attrs["HubleParam"] = info["h"]
header.attrs["Redshift"] = info["redshift"]
else:
raise ValueError(f"Unknown simname `{simname}`.")
print(f"{now()}: done with `{reader.output_snap}`.",
flush=True)
# Lastly, create the halo mapping and default catalogue.
print(f"{datetime.now()}: creating `GroupOffset`...")
halo_map, unique_halo_ids = make_offset_map(halo_ids)
# Dump the halo mapping.
with File(reader.output_cat, "w") as f:
f.create_dataset("GroupOffset", data=halo_map)
# Add the halo finder catalogue
print(f"{now()}: adding the halo finder catalogue.")
with File(reader.output_cat, "r+") as f:
cat = reader.read_halos()
hid2pos = {hid: i for i, hid in enumerate(unique_halo_ids)}
for key in cat.dtype.names:
x = numpy.full(unique_halo_ids.size, numpy.nan,
dtype=cat[key].dtype)
for i in range(len(cat)):
j = hid2pos[cat["index"][i]]
x[j] = cat[key][i]
f.create_dataset(key, data=x)
def process_initial_snapshot(nsim, simname):
"""
Sort the initial snapshot particles according to their final snapshot and
add them to the final snapshot's HDF5 file.
"""
if simname == "csiborg1":
reader = CSiBORG1Reader(nsim, "initial")
output_snap_final = CSiBORG1Reader(nsim, "final").output_snap
elif simname == "quijote":
reader = QuijoteReader(nsim, "initial")
output_snap_final = QuijoteReader(nsim, "final").output_snap
elif "csiborg2" in simname:
reader = CSiBORG2Reader(nsim, "initial", simname.split("_")[1])
output_snap_final = CSiBORG2Reader(nsim, "final", simname.split("_")[1]).output_snap # noqa
raise RuntimeError("TODO Not implemented.")
else:
raise RuntimeError(f"Simulation `{simname}` is not supported.")
print("---- Processing Initial Snapshot Information ----")
print(f"Simulation index: {nsim}")
print(f"Simulation name: {simname}")
print(f"Output snapshot: {reader.output_snap}")
print(f"Output catalogue: {reader.output_cat}")
print("-----------------------------------------------")
print(flush=True)
print(f"{now()}: loading and sorting the initial PID.")
sort_indxs = numpy.argsort(reader.read_snapshot("pid"))
print(f"{now()}: loading the final particles.")
with File(output_snap_final, "r") as f:
sort_indxs_final = f["ParticleIDs"][:]
f.close()
print(f"{now()}: sorting the particles according to the final snapshot.")
sort_indxs_final = numpy.argsort(numpy.argsort(sort_indxs_final))
sort_indxs = sort_indxs[sort_indxs_final]
del sort_indxs_final
collect()
print(f"{now()}: loading and sorting the initial particle position.")
pos = reader.read_snapshot("pos")[sort_indxs]
del sort_indxs
collect()
# In Quijote some particles are positioned precisely at the edge of the
# box. Move them to be just inside.
if simname == "quijote":
boxsize = reader.read_info()["BoxSize"]
mask = pos >= boxsize
if numpy.any(mask):
spacing = numpy.spacing(pos[mask])
assert numpy.max(spacing) <= 1e-3
pos[mask] -= spacing
print(f"{now()}: dumping particles `{reader.output_snap}`.")
with File(reader.output_snap, 'w') as f:
f.create_dataset("Coordinates", data=pos,
**hdf5plugin.Blosc(**BLOSC_KWARGS))
###############################################################################
# Process the initial snapshot and sort it like the final snapshot #
###############################################################################
if __name__ == "__main__":
parser = ArgumentParser(description="Tool to manage the `raw` simulation data.") # noqa
parser.add_argument("--nsim", type=int, required=True,
help="Simulation index.")
parser.add_argument("--simname", type=str, required=True,
choices=["csiborg1", "quijote"],
help="Simulation name.")
parser.add_argument("--mode", type=int, required=True, choices=[0, 1, 2],
help="0: process final snapshot, 1: process initial snapshot, 2: process both.") # noqa
args = parser.parse_args()
if args.mode == 0:
process_final_snapshot(args.nsim, args.simname)
elif args.mode == 1:
process_initial_snapshot(args.nsim, args.simname)
else:
process_final_snapshot(args.nsim, args.simname)
process_initial_snapshot(args.nsim, args.simname)

89
scripts_independent/run_field_sph.py Normal file
View file

@ -0,0 +1,89 @@
# Copyright (C) 2023 Richard Stiskalek
# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 3 of the License, or (at your
# option) any later version.
#
# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
# Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
"""
Script to write the SLURM submission script and submit it to the queue to
calculate the SPH density & velocity field.
"""
from os import system
def write_submit(chain_index, kind, resolution, nthreads):
if kind not in ["main", "random", "varysmall"]:
raise RuntimeError(f"Unknown kind `{kind}`.")
txt = f"""#!/bin/sh
#SBATCH --ntasks-per-node=1
#SBATCH --nodes=1
#SBATCH --cpus-per-task={nthreads}
#SBATCH --mem-per-cpu=7000
#SBATCH -J SPH_{chain_index}
#SBATCH -o output_{chain_index}_%J.out
#SBATCH -e error_{chain_index}_%J.err
#SBATCH -p cosma8-serial
#SBATCH -A dp016
#SBATCH -t 16:00:00
#SBATCH --mail-type=BEGIN,END,FAIL
#SBATCH --mail-user=richard.stiskalek@physics.ox.ac.uk
module purge
module load intel_comp/2019
module load intel_mpi/2019
module load hdf5
module load fftw
module load gsl
module load cmake
module load python/3.10.12
module list
source /cosma/home/dp016/dc-stis1/csiborgtools/venv_csiborgtools/bin/activate
export OMP_NUM_THREADS={nthreads}
export OMP_NESTED=true
snapshot_path="/cosma8/data/dp016/dc-stis1/csiborg2_{kind}/chain_{chain_index}/output/snapshot_099_full.hdf5"
output_path="/cosma8/data/dp016/dc-stis1/csiborg2_{kind}/field/chain_{chain_index}_{resolution}.hdf5"
resolution={resolution}
scratch_space="/snap8/scratch/dp016/dc-stis1/"
SPH_executable="/cosma8/data/dp016/dc-stis1/cosmotool/bld2/sample/simple3DFilter"
snapshot_kind="gadget4"
python3 field_sph.py --snapshot_path $snapshot_path --output_path $output_path --resolution $resolution --scratch_space $scratch_space --SPH_executable $SPH_executable --snapshot_kind $snapshot_kind
"""
fname = f"submit_SPH_{kind}_{chain_index}.sh"
print(f"Writing file: `{fname}`.")
with open(fname, "w") as txtfile:
txtfile.write(txt)
# Make the file executable
system(f"chmod +x {fname}")
return fname
if __name__ == "__main__":
# kind = "main"
# chains = [15617, 15717, 15817, 15917, 16017, 16117, 16217, 16317, 16417, 16517, 16617, 16717, 16817, 16917, 17017, 17117, 17217, 17317, 17417]
# kind = "varysmall"
# chains = ["16417_001", "16417_025", "16417_050", "16417_075", "16417_100", "16417_125", "16417_150", "16417_175", "16417_200", "16417_225", "16417_250", "16417_275", "16417_300", "16417_325", "16417_350", "16417_375", "16417_400", "16417_425", "16417_450", "16417_475"]
kind = "random"
chains = [1, 25, 50, 75, 100, 125, 150, 175, 200, 225, 250, 275, 300, 325, 350, 375, 400, 425, 450, 475]
resolution = 1024
nthreads = 32
for chain_index in chains:
fname = write_submit(chain_index, kind, resolution, nthreads)
system(f"sbatch {fname}")

31
scripts_independent/run_process_snapshot.py Normal file
View file

@ -0,0 +1,31 @@
# Copyright (C) 2023 Richard Stiskalek
# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 3 of the License, or (at your
# option) any later version.
#
# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
# Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
from os import system
if __name__ == "__main__":
# Quijote chains
chains = [1]
simname = "quijote"
mode = 2
env = "/mnt/zfsusers/rstiskalek/csiborgtools/venv_csiborg/bin/python"
memory = 64
for chain in chains:
out = f"output_{simname}_{chain}_%j.out"
cmd = f"addqueue -q berg -o {out} -n 1x1 -m {memory} {env} process_snapshot.py --nsim {chain} --simname {simname} --mode {mode}" # noqa
print(cmd)
system(cmd)
print()

77
scripts_independent/sort_ramseshdf5.py Normal file
View file

@ -0,0 +1,77 @@
# Copyright (C) 2023 Richard Stiskalek
# This program is free software; you can redistribute it and/or modify it
# under the terms of the GNU General Public License as published by the
# Free Software Foundation; either version 3 of the License, or (at your
# option) any later version.
#
# This program is distributed in the hope that it will be useful, but
# WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General
# Public License for more details.
#
# You should have received a copy of the GNU General Public License along
# with this program; if not, write to the Free Software Foundation, Inc.,
# 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
def add_initial_snapshot(nsim, simname, halo_finder, verbose):
"""
Sort the initial snapshot particles according to their final snapshot and
add them to the final snapshot's HDF5 file.
"""
paths = csiborgtools.read.Paths(**csiborgtools.paths_glamdring)
fname = paths.processed_output(nsim, simname, halo_finder)
if simname == "csiborg":
partreader = csiborgtools.read.CSiBORGReader(paths)
else:
partreader = csiborgtools.read.QuijoteReader(paths)
fprint(f"processing simulation `{nsim}`.", verbose)
if simname == "csiborg":
nsnap0 = 1
elif simname == "quijote":
nsnap0 = -1
else:
raise ValueError(f"Unknown simulation `{simname}`.")
fprint("loading and sorting the initial PID.", verbose)
sort_indxs = numpy.argsort(partreader.read_snapshot(nsnap0, nsim, "pid"))
fprint("loading the final particles.", verbose)
with h5py.File(fname, "r") as f:
sort_indxs_final = f["snapshot_final/pid"][:]
f.close()
fprint("sorting the particles according to the final snapshot.", verbose)
sort_indxs_final = numpy.argsort(numpy.argsort(sort_indxs_final))
sort_indxs = sort_indxs[sort_indxs_final]
del sort_indxs_final
collect()
fprint("loading and sorting the initial particle position.", verbose)
pos = partreader.read_snapshot(nsnap0, nsim, "pos")[sort_indxs]
del sort_indxs
collect()
# In Quijote some particles are position precisely at the edge of the
# box. Move them to be just inside.
if simname == "quijote":
mask = pos >= 1
if numpy.any(mask):
spacing = numpy.spacing(pos[mask])
assert numpy.max(spacing) <= 1e-5
pos[mask] -= spacing
fprint(f"dumping particles for `{nsim}` to `{fname}`.", verbose)
with h5py.File(fname, "r+") as f:
if "snapshot_initial" in f.keys():
del f["snapshot_initial"]
group = f.create_group("snapshot_initial")
group.attrs["header"] = "Initial snapshot data."
dset = group.create_dataset("pos", data=pos)
dset.attrs["header"] = "DM particle positions in box units."
f.close()