def smoothft(mesh,
R1=None,
R2=None,
quad=False,
shear=False,
pm=None,
Rmore=None):
'''Currently does the following to the mesh-
Deconvolve -> Subtract mean -> Smooth at given scales & GD
'''
pmdc = ft.decic(mesh)
meshdict = {'final': mesh.value, 'decic': pmdc.value}
mean = pmdc[...].mean()
pmdc[...] -= mean
pmdc[...] /= mean
dk = pmdc.r2c()
#Create features and parameters
if R1 is not None:
R1mesh = ft.smooth(dk, R1, 'fingauss')
meshdict['R1'] = R1mesh.value
if R2 is not None:
R2mesh = ft.smooth(dk, R2, 'fingauss')
meshdict['R2'] = R2mesh.value
if R2 is not None and R2 is not None:
R12mesh = R1mesh - R2mesh
meshdict['R12'] = R12mesh.value
if shear:
s2 = ft.shear(pm, mesh)
meshdict['shear'] = s2.value
if Rmore is not None:
for i, R in enumerate(Rmore):
Rmesh = ft.smooth(dk, R, 'fingauss')
meshdict['Rmore%d' % i] = Rmesh.value
if quad:
meshdict['fR1'] = (R1mesh * mesh).value
meshdict['fR2'] = (R2mesh * mesh).value
meshdict['R1R2'] = (R1mesh * R2mesh).value
return meshdict
def getlagfields(pm, basemesh, R=0, smoothd0=True, kernel='gauss'):
'''generate 5 lagrangian fields - 1, \delta, \delta^2, \s^2, \nabla\delta'''
if R > 0:
base = ft.smooth(basemesh, R, kernel)
else:
base = basemesh.copy()
one = (base * 0 + 1).copy()
d2 = 1. * base**2
d2 -= d2.cmean()
s2 = ft.shear(pm, base)
s2 -= s2.cmean()
lap = ft.laplace(pm, base)
lap -= lap.cmean()
#do we smooth the field with b1?
if smoothd0: d0 = base.copy()
else: d0 = basemesh.copy()
return one, d0, d2, s2, lap
doexp=doexp,
mexp=mexp,
cc=cc,
stellar=stellar)
ftt = ntools.createdata(pm, meshdict, pdict['pftname'], plocal)
mftt = ntools.createdata(pm, meshdict, mftname, mlocal)
nnpred = ntools.applynet(ftt, ptup).reshape(nc, nc, nc)
if regression:
nnmass = ntools.applymassreg(mftt, mtup).reshape(nc, nc, nc)
else:
nnmass = ntools.applynet(mftt, mtup).reshape(nc, nc, nc)
if doexp:
nnmass = mf.fmexp(ntools.relu(nnmass), mexp, cc)
predict = pm.create(mode='real')
predict[...] = nnpred * nnmass
predictR = ft.smooth(predict, Rsm, 'fingauss')
#data
hdictf = ntools.gridhalos(pm,
scratch + '/data/L%04d_N%04d_S%04d_40step/' %
(bs, fine * nc, seed),
R1=R1,
R2=R2,
pmesh=True,
abund=abund,
doexp=doexp,
mexp=mexp,
cc=cc,
stellar=stellar)[1]
datap = pm.paint(hdictf['position'][:num], hdictf['mass'][:num])
def gridhalos(pm,
dpath=None,
pos=None,
mass=None,
R1=None,
R2=None,
rank=None,
abund=True,
sigma=None,
seed=123,
pmesh=True,
z=0,
doexp=False,
mexp=None,
cc=None,
stellar=False,
verbose=True):
'''Read halo file or pos/mas arrays and grid and smooth them after scattering or matching abundance or both
The path should have folder FOF in it.
'''
#Matter
bs, nc = pm.BoxSize[0], pm.Nmesh[0]
#Halos
if dpath is not None:
from nbodykit.lab import CurrentMPIComm
CurrentMPIComm.set(
pm.comm
) #WORKAROUND because BigFileCatalog ignores comm o/w. Should not need after nbodykit v0.3.4
fofcat = BigFileCatalog(dpath + 'FOF', header='Header', comm=pm.comm)
halopos = fofcat['PeakPosition'].compute()[1:]
halomass = (fofcat['Mass'].compute() * 10**10)[1:]
try:
hvel = fofcat['CMVelocity'].compute()[1:]
except Exception as e:
print(e)
print('Cannot read velocity')
hvel = None
if verbose: print('BigFileCatalog read')
elif pos is not None:
halopos = pos
if mass is not None:
halomass = mass
else:
print(
'No halo masses given, so mass=1. Scatter and abundance not valid'
)
halomass = np.ones(halopos.shape[0])
sigma = None
abund = False
else:
print('Need either path of Bigfile, or catalogs')
return None
if abund:
if verbose: print('Match abundance')
halomass = mf.icdf_sampling(hmass=halomass, bs=bs, z=z)
if stellar:
halomass = mf.stellar_mass(halomass)
if sigma is not None:
print('Scatter catalog with sigma = %0.2f' % sigma)
halomass, halopos = dg.scatter_catalog(halomass, halopos, sigma, seed)
if doexp:
if verbose:
print('Doing doexp with mexp = %0.2f, cc=%0.2f' % (mexp, cc))
halomass = mf.fmexp(halomass, mexp, cc)
#print('In gridhalos')
#print(rank, halomass.size)
#print(dpath)
if rank is not None:
if rank > halomass.size:
print('\nCatalog not big enough to be able to match abundance')
print('\nDesired rank = %d, catalog size = %d' %
(rank, halomass.size))
return None
else:
halopos = halopos[:rank + 1]
halomass = halomass[:rank + 1]
hvel = hvel[:rank + 1]
halos = {'position': halopos, 'mass': halomass, 'velocity': hvel}
if pmesh:
halomesh = pm.paint(halopos, mass=halomass)
hposmesh = pm.paint(halopos)
meshdict = {'hposmesh': hposmesh.value, 'halomesh': halomesh.value}
dkh = halomesh.r2c()
if R1 is not None:
hmR1mesh = ft.smooth(dkh, R1, 'fingauss')
meshdict['hmR1mesh'] = hmR1mesh.value
if R2 is not None:
hmR2mesh = ft.smooth(dkh, R2, 'fingauss')
meshdict['hmR2mesh'] = hmR2mesh.value
else:
meshdict = None
return meshdict, halos
def make_bias_plot():
"""Does the work of making the real-space xi(r) and b(r) figure."""
hmesh = BigFileMesh(dpath+'ZA/opt_s999_h1massA_fourier/datap', 'mapp').paint()
noises = np.loadtxt('/global/u1/c/chmodi/Programs/21cm/21cm_cleaning/data/summaryHI.txt').T
for i in range(noises[0].size):
if noises[0][i] == np.round(1/aa-1, 2): noise = noises[3][i]
print(noise)
zamod = BigFileMesh(dpath+'ZA/opt_s999_h1massA_fourier/fitp/', 'mapp').paint()
pmmod = BigFileMesh(dpath+'T05-B1/opt_s999_h1massA_fourier/fitp/', 'mapp').paint()
fin = BigFileMesh(dpath+'ZA/opt_s999_h1massA_fourier/datap/', 'd').paint()
fin /= fin.cmean()
fin -= 1
finsm = ft.smooth(fin, 3, 'gauss')
grid = pm.mesh_coordinates()*bs/nc
params, finmod = getbias(pm, basemesh=finsm, hmesh=hmesh, pos=grid, grid=grid) #
models = [zamod, pmmod, finmod, fin.copy()]
lbls = ['ZA shift', 'PM shift', 'Eulerian (R=3)', 'Final Matter']
lss = ['-', '-', '-', '--']
fig, ax = plt.subplots(1, 3, figsize=(15, 4))
for ii, mod in enumerate(models):
pmod = FFTPower(mod, mode='1d').power
k, pmod = pmod['k'], pmod['power']
ph = FFTPower(hmesh, mode='1d').power['power']
pxmodh = FFTPower(hmesh, second=mod, mode='1d').power['power']
perr = FFTPower(hmesh -mod, mode='1d').power['power']
ax[0].plot(k, pxmodh/(pmod*ph)**0.5, label=lbls[ii], lw=2, ls=lss[ii])
ax[0].set_ylabel('$r_{cc}$', fontdict=font)
ax[1].plot(k,(pmod/ph)**0.5, lw=2, ls=lss[ii])
#ax[1].set_ylabel(r'$\sqrt{P_{\rm mod}/P_{hh}}$', fontdict=font)
ax[1].set_ylabel(r'$T_f$', fontdict=font)
ax[2].plot(k, perr, lw=2, ls=lss[ii])
ax[2].set_ylabel(r'$P_{\delta_{\rm mod}-\delta_h}$', fontdict=font)
ax[2].set_yscale('log')
ax[2].axhline(noise)
for axis in ax:
axis.set_xscale('log')
axis.grid(which='both')
axis.legend(prop=fontmanage)
axis.set_xlabel(r'$k\quad [h\,{\rm Mpc}^{-1}]$', fontdict=font)
axis.set_xscale('log')
# Put on some more labels.
for axis in ax:
axis.set_xscale('log')
for tick in axis.xaxis.get_major_ticks():
tick.label.set_fontproperties(fontmanage)
for tick in axis.yaxis.get_major_ticks():
tick.label.set_fontproperties(fontmanage)
##and finish
plt.tight_layout(rect=[0, 0, 1, 0.95])
plt.savefig(figpath + '/bfit_L%04d_%04d.pdf'%(bs, aa*10000))
def make_bsum_plot():
"""bias rcc, tf, error, bk"""
ik = args.ik
grid = pm.mesh_coordinates()*bs/nc
noises = np.loadtxt('/global/u1/c/chmodi/Programs/21cm/21cm_cleaning/data/summaryHI.txt').T
for i in range(noises[0].size):
if noises[0][i] == np.round(1/aa-1, 2): noise = noises[3][i]
print(noise)
if args.ray:
hmesh = BigFileMesh('/global/cscratch1/sd/chmodi/m3127/H1mass/highres/%d-9100/fastpm_%0.4f/HImesh-N%04d/'%(bs*10, aa, nc), 'ModelA').paint()
lin = BigFileMesh('/global/cscratch1/sd/chmodi/m3127/cm_lowres/5stepT-B1/%d-%d-9100/linear'%(bs, nc), 'LinearDensityK').paint()
dyn = BigFileCatalog('/global/cscratch1/sd/chmodi/m3127/cm_lowres/5stepT-B1/%d-%d-9100/fastpm_%0.4f/1'%(bs, nc, aa), comm=pm.comm, header='Header')
else:
hmesh = BigFileMesh('/global/cscratch1/sd/chmodi/m3127/H1mass/highres/%d-9100-fixed/fastpm_%0.4f/HImesh-N%04d/'%(bs*10, aa, nc), 'ModelA').paint()
lin = BigFileMesh('/global/cscratch1/sd/chmodi/m3127/cm_lowres/5stepT-B1/%d-%d-9100-fixed/linear'%(bs, nc), 'LinearDensityK').paint()
dyn = BigFileCatalog('/global/cscratch1/sd/chmodi/m3127/cm_lowres/5stepT-B1/%d-%d-9100-fixed/fastpm_%0.4f/1'%(bs, nc, aa), comm=pm.comm, header='Header')
print(pm.comm.rank, dyn.Index.compute())
fpos = dyn['Position']
flay = pm.decompose(fpos)
#
dgrow = cosmo.scale_independent_growth_factor(zz)
zapos = za.doza(lin.r2c(), grid, z=zz, dgrow=dgrow)
print(zapos.shape, fpos.shape)
paramsza, zamod = getbias(pm, basemesh=lin, hmesh=hmesh, pos=zapos, grid=grid, ik=ik)
_, pmmod = getbias(pm, basemesh=lin, hmesh=hmesh, pos=fpos, grid=grid, ik=ik)
kk, paramszak, _ = getbiask(pm, basemesh=lin, hmesh=hmesh, pos=zapos, grid=grid)
#
fin = pm.paint(fpos, layout=flay)
fin /= fin.cmean()
fin -= 1
finsm = ft.smooth(fin, 3, 'gauss')
_, finmod = getbias(pm, basemesh=finsm, hmesh=hmesh, pos=grid, grid=grid, ik=ik)
models = [zamod, pmmod, finmod, finsm.copy()]
lbls = ['ZA shift', 'PM shift', 'Eulerian (R=3)', 'Eulerian (R=3), $b_1^E$']
lss = ['-', '-', '-', '--']
print('Setup done')
#####
fig, ax = plt.subplots(2, 2, figsize=(8, 6), sharex=True)
for ii, mod in enumerate(models):
pmod = FFTPower(mod, mode='1d').power
k, pmod = pmod['k'], pmod['power']
ph = FFTPower(hmesh, mode='1d').power['power']
if ii == 3:
mod *= (ph[1]/pmod[1]).real**0.5
pmod *= (ph[1]/pmod[1]).real
pxmodh = FFTPower(hmesh, second=mod, mode='1d').power['power']
perr = FFTPower(hmesh -mod, mode='1d').power['power']
ax[0, 0].plot(k, pxmodh/(pmod*ph)**0.5, label=lbls[ii], lw=2, ls=lss[ii])
ax[0, 0].set_ylabel('$r_{cc}$', fontdict=font)
#if ii == 3: pmod *= ph[1]/pmod[1]
ax[0, 1].plot(k,(pmod/ph)**0.5, lw=2, ls=lss[ii])
ax[0, 1].set_ylabel(r'$T_f$', fontdict=font)
#ax[1].set_ylabel(r'$\sqrt{P_{\rm mod}/P_{hh}}$', fontdict=font)
ax[1, 0].plot(k, perr, lw=2, ls=lss[ii])
ax[1, 0].set_ylabel(r'$P_{\rm err}$', fontdict=font)
ax[1, 0].set_yscale('log')
ax[1, 0].axhline(noise, color='k', ls="--")
ax[1, 0].set_ylim(2, 5e2)
lbls = ['$b_1$', '$b_2$', '$b_s$']
for ii in range(3):
ax[1, 1].plot(kk, paramszak[ii], lw=2, color='C%d'%ii, label=lbls[ii])
ax[1, 1].axhline(paramsza[ii], lw=1, color='C%d'%ii, ls="--")
ax[1, 1].set_ylabel(r'$b(k)$', fontdict=font)
for axis in ax.flatten():
axis.set_xscale('log')
axis.grid(which='both', alpha=0.2, color='gray', lw=0.2)
axis.legend(prop=fontmanage)
axis.set_xlabel(r'$k\quad [h\,{\rm Mpc}^{-1}]$', fontdict=font)
axis.set_xscale('log')
axis.axvline(kk[ik], color='k', ls="--")
# Put on some more labels.
for axis in ax.flatten():
axis.set_xscale('log')
for tick in axis.xaxis.get_major_ticks():
tick.label.set_fontproperties(fontmanage)
for tick in axis.yaxis.get_major_ticks():
tick.label.set_fontproperties(fontmanage)
##and finish
plt.tight_layout(rect=[0, 0, 1, 0.95])
if args.ray:
plt.savefig(figpath + '/bsum_L%04d_N%04d_%04d_ik%d.pdf'%(bs, nc, aa*10000, ik))
else:
plt.savefig(figpath + '/bsumfix_L%04d_N%04d_%04d_ik%d.pdf'%(bs, nc, aa*10000, ik))
def make_biask_plot():
"""Does the work of making the real-space xi(r) and b(r) figure."""
hmesh = BigFileMesh(dpath+'ZA/opt_s999_h1massA_fourier/datap', 'mapp').paint()
noises = np.loadtxt('/global/u1/c/chmodi/Programs/21cm/21cm_cleaning/data/summaryHI.txt').T
for i in range(noises[0].size):
if noises[0][i] == np.round(1/aa-1, 2): noise = noises[3][i]
print(noise)
zamod = BigFileMesh(dpath+'ZA/opt_s999_h1massA_fourier/fitp/', 'mapp').paint()
pmmod = BigFileMesh(dpath+'T05-B1/opt_s999_h1massA_fourier/fitp/', 'mapp').paint()
fin = BigFileMesh(dpath+'ZA/opt_s999_h1massA_fourier/datap/', 'd').paint()
fin /= fin.cmean()
fin -= 1
finsm = ft.smooth(fin, 3, 'gauss')
grid = pm.mesh_coordinates()*bs/nc
params, finmod = getbias(pm, basemesh=finsm, hmesh=hmesh, pos=grid, grid=grid)
models = [zamod, pmmod, finmod, finsm.copy()]
lbls = ['ZA shift', 'PM shift', 'Eulerian (R=3)', 'Eulerian (R=3), $b_1$']
lss = ['-', '-', '-', '--']
lin = BigFileMesh(dpath+'ZA/opt_s999_h1massA_fourier/datap/', 's').paint()
dyn = BigFileCatalog('/global/cscratch1/sd/chmodi/m3127/cm_lowres/5stepT-B1/%d-%d-9100-fixed/fastpm_%0.4f/1'%(bs, nc, aa))
dgrow = cosmo.scale_independent_growth_factor(zz)
zapos = za.doza(lin.r2c(), grid, z=zz, dgrow=dgrow)
fpos = dyn['Position']
params, _ = getbias(pm, basemesh=lin, hmesh=hmesh, pos=zapos, grid=grid)
kk, paramsk, _ = getbiask(pm, basemesh=lin, hmesh=hmesh, pos=zapos, grid=grid)
fig, ax = plt.subplots(1, 2, figsize=(9, 4))
for ii, mod in enumerate(models):
pmod = FFTPower(mod, mode='1d').power
k, pmod = pmod['k'], pmod['power']
ph = FFTPower(hmesh, mode='1d').power['power']
if ii == 3:
mod *= (ph[1]/pmod[1]).real **0.5
pmod *= (ph[1]/pmod[1]).real
pxmodh = FFTPower(hmesh, second=mod, mode='1d').power['power']
perr = FFTPower(hmesh -mod, mode='1d').power['power']
ax[0].plot(k, perr, lw=2, ls=lss[ii], label=lbls[ii])
ax[0].set_ylabel(r'$P_{\rm err}$', fontdict=font)
ax[0].set_yscale('log')
ax[0].axhline(noise, color='k', ls="--")
ax[0].set_ylim(2, 5e2)
lbls = ['$b_1$', '$b_2$', '$b_s$']
for ii in range(3):
ax[1].plot(kk, paramsk[ii], lw=2, color='C%d'%ii, label=lbls[ii])
ax[1].axhline(params[ii], lw=1, color='C%d'%ii, ls="--")
ax[1].set_ylabel(r'$b(k)$', fontdict=font)
for axis in ax:
axis.set_xscale('log')
axis.grid(which='both')
axis.legend(prop=fontmanage, ncol=2)
axis.set_xlabel(r'$k\quad [h\,{\rm Mpc}^{-1}]$', fontdict=font)
axis.set_xscale('log')
# Put on some more labels.
for axis in ax:
axis.set_xscale('log')
for tick in axis.xaxis.get_major_ticks():
tick.label.set_fontproperties(fontmanage)
for tick in axis.yaxis.get_major_ticks():
tick.label.set_fontproperties(fontmanage)
##and finish
plt.tight_layout(rect=[0, 0, 1, 0.95])
plt.savefig(figpath + '/bfitk_L%04d_%04d.pdf'%(bs, aa*10000))
def make_bias2p_plot():
"""Does the work of making the real-space xi(r) and b(r) figure."""
hmesh = BigFileMesh(dpath+'ZA/opt_s999_h1massA_fourier/datap', 'mapp').paint()
grid = pm.mesh_coordinates()*bs/nc
noises = np.loadtxt('/global/u1/c/chmodi/Programs/21cm/21cm_cleaning/data/summaryHI.txt').T
for i in range(noises[0].size):
if noises[0][i] == np.round(1/aa-1, 2): noise = noises[3][i]
print(noise)
zamod = BigFileMesh(dpath+'ZA/opt_s999_h1massA_fourier/fitp/', 'mapp').paint()
try: pmmod = BigFileMesh(dpath+'T05-B1/opt_s999_h1massA_fourier/fitp/', 'mapp').paint()
except:
lin = BigFileMesh(dpath+'ZA/opt_s999_h1massA_fourier/datap/', 's').paint()
dyn = BigFileCatalog('/global/cscratch1/sd/chmodi/m3127/cm_lowres/5stepT-B1/%d-%d-9100-fixed/fastpm_%0.4f/1'%(bs, nc, aa))
fpos = dyn['Position']
_, pmod = getbias(pm, basemesh=lin, hmesh=hmesh, pos=fpos, grid=grid, fpos=fpos)
fin = BigFileMesh(dpath+'ZA/opt_s999_h1massA_fourier/datap/', 'd').paint()
fin /= fin.cmean()
fin -= 1
finsm = ft.smooth(fin, 3, 'gauss')
params, finmod = getbias(pm, basemesh=finsm, hmesh=hmesh, pos=grid, grid=grid)
models = [zamod, pmmod, finmod, finsm.copy()]
lbls = ['ZA shift', 'PM shift', 'Eulerian (R=3)', 'Eulerian (R=3), $b_1$']
lss = ['-', '-', '-', '--']
fig, ax = plt.subplots(1, 2, figsize=(9, 4))
for ii, mod in enumerate(models):
pmod = FFTPower(mod, mode='1d').power
k, pmod = pmod['k'], pmod['power']
ph = FFTPower(hmesh, mode='1d').power['power']
if ii == 3:
mod *= (ph[1]/pmod[1]).real**0.5
pmod *= (ph[1]/pmod[1]).real
pxmodh = FFTPower(hmesh, second=mod, mode='1d').power['power']
perr = FFTPower(hmesh -mod, mode='1d').power['power']
ax[0].plot(k, pxmodh/(pmod*ph)**0.5, label=lbls[ii], lw=2, ls=lss[ii])
ax[0].set_ylabel('$r_{cc}$', fontdict=font)
#if ii == 3: pmod *= ph[1]/pmod[1]
ax[1].plot(k,(pmod/ph)**0.5, lw=2, ls=lss[ii])
ax[1].set_ylabel(r'$T_f$', fontdict=font)
#ax[1].set_ylabel(r'$\sqrt{P_{\rm mod}/P_{hh}}$', fontdict=font)
for axis in ax:
axis.set_xscale('log')
axis.grid(which='both')
axis.legend(prop=fontmanage)
axis.set_xlabel(r'$k\quad [h\,{\rm Mpc}^{-1}]$', fontdict=font)
axis.set_xscale('log')
# Put on some more labels.
for axis in ax:
axis.set_xscale('log')
for tick in axis.xaxis.get_major_ticks():
tick.label.set_fontproperties(fontmanage)
for tick in axis.yaxis.get_major_ticks():
tick.label.set_fontproperties(fontmanage)
##and finish
plt.tight_layout(rect=[0, 0, 1, 0.95])
plt.savefig(figpath + '/bfit2p_L%04d_%04d.pdf'%(bs, aa*10000))
R1, R2 = pdict['R1'], pdict['R2']
#model
meshdict, dummy = ntools.readfiles(
pm,
scratch + '/data/L%04d_N%04d_S%04d_05step/' % (bs, nc, seed),
R1=R1,
R2=R2,
abund=abund)
ftt = ntools.createdata(pm, meshdict, pdict['pftname'], plocal)
mftt = ntools.createdata(pm, meshdict, mftname, mlocal)
nnpred = ntools.applynet(ftt, ptup).reshape(nc, nc, nc)
nnmass = ntools.applynet(mftt, mtup).reshape(nc, nc, nc)
predict = pm.create(mode='real')
predict[...] = nnpred * nnmass
predictR = ft.smooth(predict, 3, 'fingauss')
#data
#hdictf = ntools.gridhalos(pm, scratch +'/data/L%04d_N%04d_S%04d_40step/'%(bs, 4*nc, seed), rank=num, R1=R1, R2=R2, pmesh=True)
#datapt = pm.create(mode='real', zeros=True)
#datapt[...] = hdictf[0]['halomesh']
hdictf = ntools.gridhalos(pm,
scratch + '/data/L%04d_N%04d_S%04d_40step/' %
(bs, fine * nc, seed),
R1=R1,
R2=R2,
pmesh=True,
abund=abund)[1]
datap = pm.paint(hdictf['position'][:num], hdictf['mass'][:num])
datapR = ft.smooth(datap, 3, 'fingauss')
#model
print('Reading Files')
meshdict, dummy = ntools.readfiles(pm,
scratch +
'/data/z%02d/L%04d_N%04d_S%04d_05step/' %
(zz * 10, bs, nc, seed),
R1=R1,
R2=R2,
abund=abund)
ftt = ntools.createdata(pm, meshdict, pdict['pftname'], plocal)
mftt = ntools.createdata(pm, meshdict, mftname, mlocal)
nnpred = ntools.applynet(ftt, ptup).reshape(nc, nc, nc)
predict = pm.create(mode='real', value=nnpred)
predictR = ft.smooth(predict, Rsm, 'fingauss')
if ovd: predictR[...] = (predictR[...] - predictR.cmean()) / (predictR.cmean())
#data
print('Generating data')
hdictf = ntools.gridhalos(pm,
dpath=scratch +
'/data/z%02d/L%04d_N%04d_S%04d_40step/' %
(zz * 10, bs, sfine * nc, seed),
R1=R1,
R2=R2,
pmesh=False,
abund=abund)[1]
datap = pm.paint(hdictf['position'][:num])
predict = pm.create(mode='real')
predict[...] = nnpred * nnmass
if rsd:
if not zz:
cat = BigFileCatalog(scratch +
'/data/L%04d_N%04d_S%04d_05step/dynamic/1' %
(bs, nc, seed),
header='Header')
else:
cat = BigFileCatalog(scratch +
'/data/z%02d/L%04d_N%04d_S%04d_05step/dynamic/1' %
(zz * 10, bs, nc, seed),
header='Header')
mass = pm.paint(cat['Position'])
mom = pm.paint(cat['Position'], mass=cat['Velocity'][:, 2])
masssm = ft.smooth(mass, Rrsd, 'fingauss') + 1e-5
momsm = ft.smooth(mom, Rrsd, 'fingauss')
vel = momsm / masssm
gridpt = pm.generate_uniform_particle_grid(shift=0)
vgrid = vel.readout(gridpt)
vgrid *= rsdfactor
predictgrid = predict.readout(gridpt)
predict = pm.paint(gridpt + vgrid.reshape(-1, 1) * direct.reshape(1, -1),
mass=predictgrid)
predictR = ft.smooth(predict, Rsm, 'fingauss')
#data
print('Generating data')
#hdictf = ntools.gridhalos(pm, scratch +'/data/L%04d_N%04d_S%04d_40step/'%(bs, 4*nc, seed), rank=num, R1=R1, R2=R2, pmesh=True)
def standard(pm,
fofcat,
datap,
mf,
kb=6,
Rsm=7,
rsd=False,
zz=0,
M=0.3175,
mass=False,
poskey='PeakPosition'):
if rsd:
if pm.comm.rank == 0:
print(
'\n RSD! Key used to get position is --- RSPeakPosition \n\n')
position = fofcat['RS%s' % poskey].compute()
else:
position = fofcat['%s' % poskey].compute()
try:
hmass = fofcat['AMass'].compute() * 1e10
except:
hmass = fofcat['Mass'].compute() * 1e10
pks = FFTPower(datap.s, mode='1d').power['power']
pkf = FFTPower(datap.d, mode='1d').power['power']
random = pm.generate_uniform_particle_grid()
# random = np.random.uniform(0, 400, 3*128**3).reshape(-1, 3)
Rbao = Rsm / 2**0.5
aa = mf.cosmo.ztoa(zz)
ff = mf.cosmo.Fomega1(mf.cosmo.ztoa(zz))
layout = pm.decompose(position)
if mass: hmesh = pm.paint(position, mass=hmass, layout=layout)
else: hmesh = pm.paint(position, layout=layout)
hmesh /= hmesh.cmean()
hmesh -= 1
hmeshsm = ft.smooth(hmesh, Rbao, 'gauss')
#bias
layout = pm.decompose(fofcat['%s' % poskey])
hrealp = pm.paint(fofcat['%s' % poskey], layout=layout)
hrealp /= hrealp.cmean()
hrealp -= 1
pkhp = FFTPower(hrealp, mode='1d').power['power']
bias = ((pkhp[1:kb] / pkf[1:kb]).mean()**0.5).real
beta = bias / ff
print('bias = ', bias)
displist = calc_displist(pm=pm, base=hmeshsm, b=bias)
if mass:
hpshift = displace(pm,
displist,
position,
rsd=rsd,
f=ff,
beta=beta,
mass=hmass)
else:
hpshift = displace(pm,
displist,
position,
rsd=rsd,
f=ff,
beta=beta,
mass=None)
rshift = displace(pm, displist, random, rsd=rsd, f=ff, beta=beta)
recon = hpshift - rshift
return recon, hpshift, rshift
def dostd(hdict,
numd,
pkf,
datas,
kb=6,
Rsm=7,
rsd=False,
zz=0,
M=0.3175,
mass=False,
retfield=False,
retpower=False,
mode='1d',
Nmu=5,
los=[0, 0, 1],
retall=False):
#propogator is divided by bias
hpos = hdict['position']
hmass = hdict['mass']
pks = FFTPower(datas, mode=mode, Nmu=Nmu, los=[0, 0, 1]).power['power']
aa = mf.cosmo.ztoa(zz)
ff = mf.cosmo.Fomega1(mf.cosmo.ztoa(zz))
rsdfac = 100 / (aa**2 * mf.cosmo.Ha(z=zz)**1)
print('rsdfac = ', rsdfac)
hposrsd = hdict['position'] + np.array(los) * hdict['velocity'] * rsdfac
layout = pm.decompose(hpos[:int(numd * bs**3)])
if mass:
hpmesh = pm.paint(hpos[:int(numd * bs**3)],
mass=hmass[:int(numd * bs**3)],
layout=layout)
else:
hpmesh = pm.paint(hpos[:int(numd * bs**3)], layout=layout)
hpmesh /= hpmesh.cmean()
hpmesh -= 1
layout = pm.decompose(hposrsd[:int(numd * bs**3)])
if mass:
hpmeshrsd = pm.paint(hposrsd[:int(numd * bs**3)],
mass=hmass[:int(numd * bs**3)],
layout=layout)
else:
hpmeshrsd = pm.paint(hposrsd[:int(numd * bs**3)], layout=layout)
hpmeshrsd /= hpmeshrsd.cmean()
hpmeshrsd -= 1
pkhp = FFTPower(hpmesh, mode='1d').power['power']
bias = ((pkhp[1:kb] / pkf[1:kb]).mean()**0.5).real
beta = bias / ff
print('bias = ', bias)
random = pm.generate_uniform_particle_grid()
random = np.random.uniform(0, 400, 3 * 128**3).reshape(-1, 3)
Rbao = Rsm / 2**0.5
if not rsd:
hpmeshsm = ft.smooth(hpmesh, Rbao, 'gauss')
displist = calc_displist(hpmeshsm, b=bias)
if mass:
hpshift = displace(pm,
displist,
hpos[:int(numd * bs**3)],
mass=hmass[:int(numd * bs**3)])
else:
hpshift = displace(pm, displist, hpos[:int(numd * bs**3)])
rshift = displace(pm, displist, random)
recon = hpshift - rshift
pksstd = FFTPower(recon, mode=mode, Nmu=Nmu, los=[0, 0,
1]).power['power']
pkxsstd = FFTPower(recon,
second=datas,
mode=mode,
Nmu=Nmu,
los=[0, 0, 1]).power['power']
rccstd = pkxsstd / (pks * pksstd)**0.5
cksstd = pkxsstd / pks / bias
if rsd:
RSD
hpmeshrsdsm = ft.smooth(hpmeshrsd, Rbao, 'gauss')
displist = calc_displist(hpmeshrsdsm, b=bias)
hpshift = displace(pm,
displist,
hposrsd[:int(numd * bs**3)],
rsd=True,
f=ff,
beta=bias / ff)
if mass:
hpshift = displace(pm,
displist,
hposrsd[:int(numd * bs**3)],
rsd=True,
f=ff,
beta=bias / ff,
mass=hmass[:int(numd * bs**3)])
else:
hpshift = displace(pm,
displist,
hposrsd[:int(numd * bs**3)],
rsd=True,
f=ff,
beta=bias / ff)
rshift = displace(pm, displist, random, rsd=True, f=ff, beta=beta)
recon = hpshift - rshift
pksstd = FFTPower(recon, mode='1d').power['power']
pkxsstd = FFTPower(recon, second=datap[tkey].s,
mode='1d').power['power']
pksstd = FFTPower(recon, mode=mode, Nmu=Nmu, los=[0, 0,
1]).power['power']
pkxsstd = FFTPower(recon,
second=datas,
mode=mode,
Nmu=Nmu,
los=[0, 0, 1]).power['power']
rccstd = pkxsstd / (pks * pksstd)**0.5
cksstd = pkxsstd / pks / bias
if retall:
return [rccstd, cksstd], [recon, hpshift, rshift,
displist], [pkxsstd, pksstd, pks], bias
if retfield: return [rccstd, cksstd], [recon, hpshift, rshift, displist]
elif retpower: return [rccstd, cksstd], [pkxsstd, pksstd, pks], bias
else: return rccstd, cksstd