chroma

• G4DAEChroma
  • G4DAEChroma
    • Objective
    • Recent Changes
    • TODO : Test on Workstation GPU
    • Dependencies
    • Current Approach Aiming at
    • Ultimate photon list compression approach
    • Transform cache problem
    • Geometry Consistency between GPU and MEMORY trees
    • Appropriate identifiers
    • Generalization of PMT Identifier handling
    • Incorporate IDMAP Handling into G4DAEChroma/G4DAEGeometry
    • Usage Stages
      • BeginOfRunAction
      • StackAction::ClassifyNewTrack OR customized Processes
      • StackAction::NewStage
    • How to organize code
    • MockNuWa code development
    • Real NuWa hookup for machinery test
      • csa : ChromaStackAction
      • export-
      • GiGaRunActionBase
    • GiGaRunActionExport
    • Initialize in RunAction?
      • Looking for hooks
    • Issues
      • Development Cycle too slow
      • GPU Hit handling : SensDet
      • Accessing SD
      • Detector Specific Code
    • Trace channel_id/PmtId from CUDA kernel backwards
  • G4DAEChroma Implementation Details
    • G4DAEChroma Dependencies
    • High Level API
    • Collectable/Transportable Object Definitions
    • Collectable/Transportable Lists
    • Manual Lists to be migrated to common machinery
    • Transport infrastructure
    • Array Collection and Transport
    • Config and control Infrastructure
    • Utililites
    • Geometry
    • Geant4 Hit Collection Integration
    • Daya Bay specializations from DetSimChroma
    • G4DAEChroma Config and Initialization
    • Soon to be removed
• Chroma Issues
  • EXECUTIVE SUMMARIES ONLY
  • Chroma CUDA Instability, GPU Panics
  • Chroma Camera is uncontrollable
  • Chroma Surface model mismatch G4LogicalBoundarySurface
• Chroma Intersection
  • Box intersection with axis aligned photons
  • CUDA float operations IEEE 754
  • intersect_mesh
• CHROMA
  • Top Level Link
  • Update for CUDA 7, Dec 2015
  • Installation Overview
    • Dependencies Install Overview
    • IMPORTANT
    • chroma deps server
    • D: modify chroma geant4 build to add GDML/persistency
  • Chroma Dependencies
    • OSX 10.9.1 Mavericks
    • Common
    • Careful with errant eggs
    • chroma-deps pycuda build failure
    • root 5.34.11 freetype issue
    • rerun root
    • g4py_chroma
    • pyzmq
    • ZMQ
    • markupsafe
    • pip logs
    • root-5.34.14
      • TTF : header name clash between freetype and ftgl
      • rebuild pycuda with opengl capability
  • Chroma Install
  • Chroma Rebuild
  • Chroma Install Rerun from my bitbucket fork
  • Moved retainable chroma mods into bitbucket fork
• How Chroma Works ?
  • questions
  • chroma/generator/photon.py
    • G4GeneratorProcess
    • G4ParallelGenerator
  • chroma/sim.py
    • Simulation
• Chroma Development
  • Primary Repo
  • Forks
  • Dev approach
• Chroma Refs
  • Search
  • Author presentations
  • CUDA
  • GPU from python, PyCuda PyOpenCL
    • PyCuda
    • PyOpenCL
  • Geant4Py aka g4py
• Chroma Prerequisites
  • Software
  • uncertainties
  • pycuda
  • pyublas
    • src/G4chroma.cc
  • pygame
  • spnav
  • pyzmq
• Chroma Geant4 Integration
  • Chroma Use of G4
  • Boost python C++ _g4chroma
    • ChromaPhysicsList
    • PhotonTrackingAction
    • Attempt to add a PhotonSteppingAction
• Chroma/NuWa/Geant4 Integration with ZeroMQ
  • Related
  • Thoughts
    • Photon Collection
    • Send to server
    • Handle Results from server
  • NuWa/Geant4 Integration
    • NuWa DetSim
    • Where is DsOpStackAction hooked up in NuWa ?
    • How to make ZMQ call from inside C++ code ?
• Chroma NuWa Integration
  • Thoughts
  • Geant4/DetSim reminder
  • Questions
    • how/when to give OP tracks back to G4/reconstruction code ?
    • how does G4 OP propagation end ?
    • translation of DYB solid geomety into surface tris
    • what about some magic optransport physics process ?
    • OP collection and propagation, kicked off where ?
• Chroma CUDA Photon handling
• Chroma COLLADA
  • GDML wavelength dependant properties
• Getting to know how Chroma uses PyCUDA
  • PyCUDA compilation in Chroma
  • CUDA modules available to python
  • CUDA model of grid/blocks/threads
  • GPU Hardware Limits
    • GeForce GT 750M
  • chroma grid/blocks
  • chroma grid
  • sharing between threads, sharing geometry
• Chroma Use of PyUblas
  • using extension from ipython
  • test setters
  • extension setup
  • build extension with setters added
  • Usage of G4chroma extension
  • G4chroma extension
  • PyUblas
• Chroma Geometry Source Overview
  • GPU memory usage of chroma geometry
  • where is geometry populated
  • chroma/cuda/geometry_types.h
  • chroma/gpu/geometry.py
    • GPUGeometry
    • material and surface indices packed into material_codes
  • chroma/loader.py
  • chroma/stl.py
  • chroma/geometry.py
    • Q: where all these properties getting set ?
  • Chroma BVH class chroma/bvh/bvh.py
    • chroma/cuda/geometry_types.h
  • chroma/cuda/mesh.h
• Chroma Materials
  • Material handling in Chroma
  • chroma/geometry.py
    • Solid
    • Material
    • Surface
  • chroma/gpu/geometry.py
  • chroma/cuda/propagate.cu
    • propagate
  • chroma/cuda/photon.h
    • State
    • fill_state
    • material_codes
• Photon Propagation
  • python side GPUPhotons
  • cuda entry points
  • cuda propagate
    • chroma/cuda/geometry_types.h
    • chroma/cuda/photon.h
    • surface_index
• Chroma Physics
  • Is the modeling equivalent to Geant4 ?
• Chroma Modeling of Sensitive Detectors
  • Overview
  • GPU hit formation
    • cutdown CPL to just the hits
    • ChromaManager singleton
    • G4VSensitiveDetector
  • correspondence between the COLLADA boundgeom.matrix and G4 TopTransform
    • volume 0
    • volume 1
    • last volume
    • all volumes compared
    • geant4 transforms
    • pycollada boundgeom nodes matrix
    • ProcessHits Hit formation
  • Adding Detection Surfaces
    • Using a surface per-PMT exceeds a limit
  • Questions
  • chroma/geometry.py
    • detector_material not used ?
    • detsim does EFFICIENCY lookup on Bialkali materialof the cathode
    • cathode example
  • chroma/detector.py
    • add_solid/add_pmt
  • chroma/gpu/detector.py
  • chroma/gpu/daq.py
    • run_daq
  • SURFACE_DETECT
  • chroma/cuda/photon.h
  • chroma/cuda/daq.cu
• Chroma Server
  • Transport Mechanism : Feeding the server from Geant4 C++
  • Alternatives to ROOT TObject Serialization ?
    • boost serialization
    • numpy from C ?
• Chroma ZeroMQ
  • Where does ZeroMQ server fit in ?
  • chroma/generator/photon.py
  • chroma/sim.py
  • chroma/io/root.py
  • bin/chroma-server
  • client usage of this server
    • Used to communicate between chroma and g4py subprocesses
  • bin/chroma-sim
• chroma_camera
  • Chroma Camera Usage
    • docs
    • keys using pygame
    • chroma-cam
  • Chroma Camera DAE Debug
    • Initial Shakedown
      • First Run
      • After avoid None material
      • Setting defaults
      • Black Window Issue
    • Simple Camera Profile
  • Chroma Camera Enhancements
    • Remote Steering/Lookat/Viewpointmarks
    • GPU resident pixels
    • Calculate ray directions in CUDA code
    • Quaternion Trackball Control
      • transformations
  • render_pbo
    • Render PBO
      • next
      • called twice
      • glumpy frame size
      • daeviewgl integration
      • CUDA Device Reset
      • render_pbo kernel launch observations
      • kernel pixel times figure
      • tri count metric
      • profile shows first launch much more expensive
      • standard chroma cam
  • Chroma Raycaster
    • Raytrace performance
• chroma_sim
  • Chroma Sim Usage
    • Simplify
    • chroma-sim
    • Simulation
    • GPUDaq
• Chroma CUDA Freeze
  • Symptoms March 10, 2014
  • Rare Catch
  • Examine Logs
    • search
  • CUDA Drivers
    • Update History
    • New Release 5.5.47
    • Search
• Chroma CUDA
  • CUDA 101
  • Chroma Shared Geometry
• Chroma Visualisation
  • CUDA OpenGL Interop
    • Numpy OpenGL
    • OpenCL OpenGL Interop
• Photon Mapping
  • Accelerated Stochastic Progressive Photon Mapping On GPU
  • Opposite Renderer
  • MNRT
• Geant4 Background
  • G4UserTrackingAction
  • G4TrackStatus
• Cuda Threads and Blocks
  • Host/device coordination
  • Blocks and/or Threads
    • Why Bother with Threads?
• Bounding Volume Hierarchy (BVH)
  • Background
• Hardware
  • Requirements
    • Operating System
    • CUDA Drivers
    • Which Nvidia GPU ?
    • What GPUs are commonly used with Chroma ?
    • GPU Reviews
    • Costs
  • Reference
  • NTU Planned GPU Cluster Upgrade
  • Apple rMBP and iMac with Mobile variants
• install issues
  • ROOT issues
    • freetype ftgl issue with v5.34.14 (2014/01/17)
  • Python issues
    • Locale
    • help commands trys to install stuff
    • python version mixup
    • test run
  • pygame
    • resources
    • chroma pygame segv
    • version
    • macports pygame
    • macports trawl
    • pygame sysfont hardcoded paths
    • pygame.tests font fails
      • failing font tests
      • freesansbold.ttf
    • pygame built upon SDL
    • pygame sysfont access
      • listing fonts fc-list
    • font book installation
• tests
  • Overview of Chroma Tests
    • benchmark
    • chroma_cam
    • chroma_sim
  • Unittests
    • Initially, Feb/Mar 2014
      • atexit mess avoidance
      • parent module ‘pycuda’ not found while handling absolute import
      • testTime Failure repeats but with different values
    • Again April 2014
    • nosetests : a nicer way to run them
  • chroma benchmark
    • gzip issue
    • io module shadowing
    • hg rename did not delete the io folder
    • cuMemAlloc failed with demo.detector()
    • with demo.tiny() detector
    • Update for changed API
    • After uncertainties refactor
  • bin/chroma-cam
    • solution to pygame segfault : dont fork
    • pygame parachute segmentation fault
    • issue avoided by not forking
    • lldb not informative
    • supplying an input file, somehow yields a trace
    • Looking for minimal tickle
    • checking the cache
    • red herrings
  • Forking Multiprocessing pycuda
    • crux of chroma-sim
    • Observations
  • cube example
    • Controls on macbook pro
    • Issues
      • Needs nvcc in PATH
      • CUDA Compilation Warnings
      • Initial Crash, no device
    • Working
  • simple detector example
    • Runtime Mavericks ROOT/python issue with root-5.34.11 AVOIDED
    • Runtime hang, probably zmq ? INVALID
    • XQuartz DISPLAY issue : RESOLVED
    • zombie test.root file : INVALID
    • working
    • how to see what is happening on GPU
    • with more logging
  • uncertainties
    • uncertainties refactor
• ChromaZMQRootTest
  • ZMQRoot/MyTMessage dtor issue
  • CZRT : ChromaZMQRootTest
    • Memory Issue
    • Running Echoserver
    • FUNCTIONS
    • SCRIPTS
    • Local DEBUG
    • NuWa DEBUG
      • works locally
• Chroma Use Of CURand
  • init_rng
• Chroma Reemission