IceCube simulation with PPC Dmitry Chirkin, UW Madison photon propagation code
Direct photon tracking with PPC simulating flasher/standard candle photons same code for muon/cascade simulation using precise scattering function: linear combination of HG+SAM using tabulated (in 10 m depth slices) layered ice structure employing 6-parameter ice model to extrapolate in wavelength tilt in the ice layer structure is properly taken into account transparent folding of acceptance and efficiencies precise tracking through layers of ice, no interpolation needed precise simulation of the longitudinal development of cascades and angular distribution of particles emitting Cherenkov photons photon propagation code
Updates to ppc since last meeting PPC: LONG: simulate longitudinal cascade development ANGW: smear cherenkov cone due to shower development Corrected ice density to average at detector center Made the code scalable with the number of GPU multiprocessors The flasher simulation now uses the wavelength profile read from file wv.dat Randomized the simulation based on system time (with us resolution) Modified code to run CPU and GPU parts concurrently Added option to disable a multiprocessor Added the implementation of the simple approximate Mie scattering function Added a configuration file "cfg.txt" New oversized DOM treatment (designed for minimum bias compared to oversize=1): oversize only in direction perpendicular to the photon time needed to reach the nominal (non-oversized) DOM surface is added re-use the photon after it hits a DOM and ensure the causality in the flasher simulation nominal DOM oversized DOM oversized ~ 5 times photon
Timing of oversized DOM MC xR=1 default do not track back to detected DOM do not track after detection no ovesize delta correction! do not check causality del=(sqrtf(b*b+(1/(e.zR*e.zR-1)*c)-D)*e.zR-h del=e.R-OMR Flashing xR=1 default
Photon angular profile from thesis of Christopher Wiebusch
New ice density: mwe handbook of chemistry and physics T.Gow's data of density near the surface T= *d+5.822e-6*d (fit to AMANDA data) Fit to (1-p 1 *exp(-p 2 *d))*f(T(d))*(1+0.94e-12*9.8*917*d)
Simplified Mie Scattering Single radius particles, described better as smaller angles by SAM Also known as the Liu scattering function Introduced by Jon Miller
New approximation to Mie f SAM
ppc icetray module at uses a wrapper: private/ppc/i3ppc.cxx, which compiles by cmake system into the libppc.so it is necessary to compile an additional library libxppc.so by running make in private/ppc/gpu: “make glib” compiles gpu-accelerated version (needs cuda tools) “make clib” compiles cpu version (from the same sources!) link to libxppc.so and libcudart.so (if gpu version) from build/lib directory this library file must be loaded before the libppc.so wrapper library These steps are automated with a resouces/make.sh script
ppc example script run.py if(len(sys.argv)!=6): print "Use: run.py [corsika/nugen/flasher] [gpu] [seed] [infile/num of flasher events] [outfile]" sys.exit() … det = "ic86" detector = False … os.putenv("PPCTABLESDIR", expandvars("$I3_BUILD/ppc/resources/ice/mie")) … if(mode == "flasher"): … str=63 dom=20 nph=8.e9 tray.AddModule("I3PhotoFlash", "photoflash")(…) os.putenv("WFLA", "405") # flasher wavelength; set to 337 for standard candles os.putenv("FLDR", "-1") # direction of the first flasher LED … # Set FLDR=x+(n-1)*360, where 0 0 to simulate n LEDs in a # symmetrical n-fold pattern, with first LED centered in the direction x. # Negative or unset FLDR simulates a symmetric in azimuth pattern of light. tray.AddModule("i3ppc", "ppc")( ("gpu", gpu), ("bad", bad), ("nph", nph*0.1315/25), # corrected for efficiency and DOM oversize factor; eff(337)= ("fla", OMKey(str, dom)), # set str=-str for tilted flashers, str=0 and dom=1,2 for SC1 and 2 ) else:
ppc-pick and ppc-eff ppc-pick: restrict to primaries below MaxEpri load("libppc-pick") tray.AddModule("I3IcePickModule ","emax")( ("DiscardEvents", True), ("MaxEpri", 1.e9*I3Units.GeV) ) ppc-eff: reduce efficiency from 1.0 to eff load("libppc-eff") tray.AddModule("AdjEff", "eff")( ("eff", eff) )
Todo list from the last meeting need to: verify that it works for V of simulation add code to treat high-efficient DOMs correctly verify that it works for IC59 improve flasher simulation (interface with photoflash) figure out the best way to compile All done! Done?
ppc homepage
GPU scaling Original:1/2.081/2.70 CPU c++: Assembly: GTX 295: GTX/Ori: C1060: C2050: GTX 480: On GTX 295: GHz Running on 30 MPs x 448 threads Kernel uses: l=0 r=35 s=8176 c=62400 On GTX 480: GHz Running on 15 MPs x 768 threads Kernel uses: l=0 r=40 s=3960 c=62400 On C1060: GHz Running on 30 MPs x 448 threads Kernel uses: l=0 r=35 s=3992 c=62400 On C2050: GHz Running on 14 MPs x 768 threads Kernel uses: l=0 r=41 s=3960 c=62400 Uses cudaGetDeviceProperties() to get the number of multiprocessors, Uses cudaFuncGetAttributes() to get the maximum number of threads
Kernel time calculation Run 3232 (corsika) IC86 processing on cuda002 (per file): GTX 295: Device time: (in-kernel: ) [ms] GTX 480: Device time: (in-kernel: ) [ms] If more than 1 thread is running using same GPU: Device time: (in-kernel: ) [ms] 3 counters:1. time difference before/after kernel launch in host code 2. in-kernel, using cycle counter:min thread time 3.max thread time Also, real/user/sys times of top: gpus6 cpus1 cores8 files693 Real749m4.693s User3456m10.888s sys39m50.369s Device time: [ms] files: 693 real: user: gpu: kernel: [seconds] 81%-91% GPU utilization
Concurrent execution time CPUGPUCPUGPU Thread 1: CPUGPUCPUGPU Thread 2: CPU GPU CPU GPU CPU GPU CPU GPU One thread: Create track segments Copy track segments to GPU Process photon hits Copy photon hits from GPU Need 2 buffers for track segments and photon hits However: have 2 buffers: 1 on host and 1 on GPU! Just need to synchronize before the buffers are re-used
BAD multiprocessors (MPs) clist cudatest cuda cuda cuda #badmps cuda cuda cuda Disable 3 bad GPUs out of 24: 12.5% Disable 3 bad MPs out of 720: 0.4%! Configured: xR=5 eff=0.95 sf=0.2 g=0.943 Loaded 12 angsens coefficients Loaded 6x170 dust layer points Loaded random multipliers Loaded 42 wavelenth points Loaded 171 ice layers Loaded 3540 DOMs (19x19) Processing f2k muons from stdin on device 2 Total GPU memory usage: photons: hits: 991 Error: TOT was a nan or an inf 1 times! Bad MP #20 photons: hits: 393 photons: hits: 570 photons: hits: 501 photons: hits: 832 photons: hits: 717 CUDA Error: unspecified launch failure Total GPU memory usage: photons: hits: 938 Error: TOT was a nan or an inf 9 times! Bad MP #20 #20 #20 #20 photons: hits: 442 photons: hits: 627 CUDA Error: unspecified launch failure gpu]$ cat mmc.1.f2k | BADMP=20./ppc 2 > /dev/null Configured: xR=5 eff=0.95 sf=0.2 g=0.943 Loaded 12 angsens coefficients Loaded 6x170 dust layer points Loaded random multipliers Loaded 42 wavelenth points Loaded 171 ice layers Loaded 3540 DOMs (19x19) Processing f2k muons from stdin on device 2 Not using MP #20 Total GPU memory usage: photons: hits: 871 … photons: hits: 114 Device time: (in-kernel: ) [ms] Failure rates:
Typical run times corsika: run 3232: sec files ic86/spx/3232 on cuda00[123] (53.4 seconds per job) 1.2 days of real detector time in 6.5 days nugen: run 2972: event files; E^-2 weighted ic86/spx/2972 on cudatest (25.1 seconds per job) entire 10k set of files in 2.9 days this is enough for an atmnu/diffuse analysis! Considerations: Maximize GPU utilization by running only mmc+ppc parts on the GPU nodes still, IC40 mmc+ppc+detector was run with ~80% GPU utilization run with 100% DOM efficiency, save all ppc events with at least 1 MC hit apply a range of allowed efficiencies (70-100%) later with ppc-eff module