Skip to content
Snippets Groups Projects
Select Git revision
20 results
Search by author
  • Any Author
0 authors
  1. Jun 21, 2011
  3. Jun 20, 2011
  5. Jun 18, 2011
    • Wouter Verkerke's avatar
      · 29698b17
      Wouter Verkerke authored 
        o  RooDataSet

     - Introduce pool-based memory allocation scheme

     - In ctor for composite datasets with Link() ensure that 
       order of observables is preserved to allow use of assignFast()
       

  o  RooRealSumPdf

     - Use getNorm() in expectedEvents() rather than createIntegral()


  o  RooAbsCollection

     - Do no allocate hash tables by default

     - Introduce contents claim counter to counter 'foreign' ownership
       claims of contents (needed for RooAbsOptTestStatistic::setData()
       support)

     - If collection is owning, only delete contents if claim count is also zero

     - In snapshot() raise hash table threshold from 100 to 1000


  o RooRefCountedList

     - Disable hash tables by default


  o RooStudyPackage

     - Make study package own clone of incoming workspace

     - Implement copy ctor


  o RooAbsTestStatistic

     - Implement 'master' setData() method. Forward to setDataSlave()
       of self for monolithic likelihoods or forward data components
       to likelihood components in case node is SimMaster. If data
       is not fast-splittable, execute slow split.


  o RooAbsData

     - Implement 'live check' to be able to conclude if a given dataset
       pointer points to a live or dead object

     - Claim contents of _vars through new claim mechanism. Release
       in dtor, allow _vars to delete contents if claim count is zero

     - Add concept of owned component datasets to support fast-splitting
       of dataset

     - Add method canSplitFast() that indicate if fast split is available
       (only possible if dataset was created as composite)

     - Add method getSimData() that retrieves pointer to component
       RooDataSet representation for a given index state


   o RooObjCacheManager

     - Add static mechanism that allows to deleted cached observables lists used for
       cache optimization triggers to be deleted on a sterilize() action. This
       is useful on toy studies with setData() operation, that would otherwise
       collect very large lists of observable clones during the toy cycle.


   o RooAbsPdf

     - Set contraint term to ADirty to avoid synchronization problems when using
       setData() in the likelihood


   o RooAbsArg

     - In clone operations set hash table threshold at 1000 nodes

     - In getObservables() use pass-by-value iterators rather than heap iterators
       to making this faster. Also make getObservables() non-virtual, as it is in
       fact not overloaded in any subclass (anymore). Also remove looping over
       branch nodes to call getObservablesHook() as no class implements method
       anymore.

   
   o RooAddition

     - Implement a setData() method that forwards to its components. Allows transparent
       operation of setData() on likelihood objects with constraint terms


   o RooLinkedList

     - Add method that returns iterator by value (simply forwards to similar implementation
       in RooLinkedList)


   o RooSimGenContext   

     - Move calculation of fraction table from ctor to dedicated updateFraction() method

     - Before each generation sequence call updateFractions() so that multi-generation
       sequence are robust against parameter changes between calls

     - In createDataSet() always make a composite dataset if the index is present, as this
       allows faster use in toymc operations


   o RooAbsOptTestStatistic

     - Initialize all variables in default ctor

     - In dtor and ctor handle ownership of dataset observables, on which we put a claim
       (i.e. we need to delete contents if the test statistics survives the dataset)

     - Reimplement existing setData() as setDataSlave(), and add claim management on
       dataset observables. When attached new data to function, request that object caches
       clear full content in sterilize() operation to avoid buildup of cache objects in
       long toy runs.

     - In setData() adjust event count of test statistics, for correct functioning of extended
       terms after update


   o RooArgSet

      - In ctor for list and a single object, only add object if not already in list
        (is already checked, but new procedure preserves order of list elements in case
         object was already present)

   o LinkDef

      - Add list<RooAbsData*> and map<string,RooAbsData*> to support streaming of composite datasets


   o RooAbsStudy

      - Implement clone method to support polymorphic array cloning

 
   o RooAbsReal

      - Add interface method setData() 


   o RooProfileLL

      - When storing coordinates of absolute minimum, store only those that are non-constant.
        Otherwise option to restart each minimization from known absmin may alter constant
        parameters


git-svn-id: http://root.cern.ch/svn/root/trunk@39817 27541ba8-7e3a-0410-8455-c3a389f83636
      29698b17
  7. Jun 17, 2011
  9. Jun 16, 2011
  11. Jun 15, 2011
  13. Jun 14, 2011
    • Fons Rademakers's avatar
      upgrade libz from version 1.2.3 to 1.2.5: · fb8ecc8a
      Fons Rademakers authored 
      - Version 1.2.5 fixes bugs in gzseek() and gzeof() that were present in
  version 1.2.4 (March 2010). All users are encouraged to upgrade immediately.
- Version 1.2.4 has many changes over 1.2.3, including these improvements:
  - Fixed bugs in adler32_combine(), compressBound(), and deflateBound()
  - Wholesale replacement of gz* functions with faster versions
  - As part of that, added gzbuffer(), gzoffset(), gzclose_r(), and gzclose_w()
    functions
  - Faster Z_HUFFMAN_ONLY and Z_RLE compression for images and other
    specialized compression
  - Added flush options Z_BLOCK to deflate() and Z_TREES to inflate() for
    finer control
  - Added inflateReset2() and inflateMark() functions, the latter to aid
    in random access applications
  - Added LFS (Large File Summit) support for 64-bit file offsets and many
    other portability improvements

However, note, from the zlib release notes:
If you are using libxml version 2.7.6 or earlier, you will need to update
libxml to version 2.7.7 or later before installing zlib version 1.2.4 or
later. libxml 2.7.6 and earlier made unnecessary assumptions about the
undocumented internal structure of zlib that were changed in zlib 1.2.4
and result in libxml crashing. This was fixed in libxml 2.7.7.

OSX 10.6 has libxml 2.7.8 and is ok, but all current SLC and Ubuntu
distributions still have an older libxml (check with xml2-config --version).

Zlib 1.2.5 is between 10 and 20% faster then version 1.2.3. 
To enable it do:
   ./configure --enable-builtin-zlib


git-svn-id: http://root.cern.ch/svn/root/trunk@39730 27541ba8-7e3a-0410-8455-c3a389f83636
      fb8ecc8a