Global constants

Global variables and constants are defined upon initialization of the pisa package (pisa/__init__.py) and are available to all of its modules. They can be imported via from pisa import <constant>.

Here we keep track of which global constants are available, what their purpose is, and by which stages they are used.

Description

Constant	Description	Default	Overwritten by environment variables (priority indicated where necessary)
NUMBA_CUDA_AVAIL	Availability of Numba’s CUDA interface	False (unless CUDA-capable GPU available)
TARGET	Numba compilation target	cpu	PISA_TARGET (GPU target only possible if NUMBA_CUDA_AVAIL)
OMP_NUM_THREADS	Number of threads allocated to OpenMP	1	OMP_NUM_THREADS
PISA_NUM_THREADS	Global limit for number of threads (also upper limit for OMP_NUM_THREADS)	1 (numba.config.NUMBA_NUM_THREADS) for TARGET='cpu' ('parallel')	PISA_NUM_THREADS
PISA_HIST_THREADING	Multi-threading mode for PISA (fast) histogramming	'off'	PISA_HIST_THREADING
FTYPE	Global floating-point data type	np.float64	PISA_FTYPE
CTYPE	Global complex-valued floating-point data type	np.complex128 (np.complex64) for FTYPE=np.float64(32)
ITYPE	Global integer data type	np.int64 (np.int32) for FTYPE=np.float64(32)
HASH_SIGFIGS	Number of significant digits used for hashing numbers, depends on FTYPE	12(5) for FTYPE=np.float64(32)
EPSILON	Best numerical precision, derived from HASH_SIGFIGS	10**(-HASH_SIGFIGS)
C_FTYPE	C floating-point type corresponding to FTYPE	'double' ('float') for FTYPE=np.float64(32)
C_PRECISION_DEF	C precision of floating-point calculations, derived from FTYPE	'DOUBLE_PRECISION' ('SINGLE_PRECISION') for FTYPE=np.float64(32)
CACHE_DIR	Root directory for storing PISA cache files	'~/.cache/pisa'	1.PISA_CACHE_DIR, 2.XDG_CACHE_HOME/pisa

Usage

The table below depicts which services make use of a select set of global constants. Note that the table entries are derived both from the module files themselves (where the services are defined) and from any pisa.utils objects they make use of (e.g., reliance on “PISA-tailored” jit in numba_tools). Constants which are implicitly used by all services via pisa.core objects (e.g. HASH_SIGFIGS, CACHE_DIR) are not shown. Also note that where a service implements FTYPE and relies on C extension code, the simultaneous implementation of C_FTYPE and C_PRECISION_DEF is implied.

Legend

• ✓: implements

• ✗: does not implement but does not fail (i.e., ignores)

	TARGET	PISA_NUM_THREADS	FTYPE	PISA_HIST_THREADING
absorption.earth_absorption	✗	✗	✓	✗
aeff.aeff	✗	✗	✗	✗
aeff.weight	✗	✗	✗	✗
aeff.weight_hnl	✗	✗	✗	✗
background.atm_muons	✗	✗	✗	✗
cont_sys.snowstorm_hist	✓	✓	✓	✓
data.csv_data_hist	✗	✗	✓	✗
data.csv_icc_hist	✗	✗	✓	✗
data.csv_loader	✗	✗	✓	✗
data.freedom_hdf5_loader	✗	✗	✓	✗
data.grid	✗	✗	✓	✗
data.licloader_weighter	✗	✗	✓	✗
data.meows_loader	✗	✗	✓	✗
data.simple_data_loader	✗	✗	✓	✗
data.simple_signal	✗	✗	✓	✗
data.sqlite_loader	✗	✗	✓	✗
data.toy_event_generator	✗	✗	✓	✗
discr_sys.hypersurfaces	✗	✗	✓	✗
discr_sys.ultrasurfaces	✗	✗	✓	✗
flux.airs	✗	✗	✓	✗
flux.astrophysical	✗	✗	✓	✗
flux.barr_simple	✓	✓	✓	✗
flux.daemon_flux	✗	✗	✓	✗
flux.hillasg	✗	✗	✓	✗
flux.honda_ip	✗	✗	✓	✗
flux.mceq_barr	✓	✓	✓	✗
flux.mceq_barr_red	✓	✗	✓	✗
likelihood.generalized_llh_params	✗	✗	✓	✗
osc.decoherence	✓	✓	✓	✗
osc.globes	✓	✓	✓	✗
osc.nusquids	✗	✓	✓	✗
osc.prob3	✓	✓	✓	✗
osc.two_nu_osc	✓	✓	✓	✗
reco.resolutions	✗	✗	✗	✗
reco.simple_param	✗	✗	✓	✗
utils.add_indices	✗	✗	✗	✗
utils.adhoc_sys	✗	✗	✓	✗
utils.bootstrap	✗	✗	✗	✗
utils.fix_error	✗	✗	✓	✗
utils.hist	✓	✓	✓	✓
utils.kde	✗	✗	✗	✗
utils.kfold	✗	✗	✓	✗
utils.resample	✓	✓	✓	✗
utils.set_variance	✓	✓	✓	✗
xsec.dis_sys	✓	✓	✓	✗
xsec.genie_sys	✗	✗	✗	✗
xsec.nutau_xsec	✓	✓	✓	✗