Difference: Hlt2016WinterTasks (1 vs. 18)

Revision 182016-01-25 - MikaVesterinen

Line: 1 to 1
 
META TOPICPARENT name="LHCbTrigger"

Tasks for the 2016 HLT commissioning

Line: 16 to 16
 
TCKsh, and documentation Anton, Albert, ...  
Resurrection of FEST, and pit installation procedure ..  
Make Ghost probability available in Hlt1 ..  
Added:
>
>
Make ProbNN variables available in Hlt2 Kevin, Johannes  
 

Memory profiling

Line: 92 to 93
 

Make Ghost probability available in Hlt1

Added:
>
>

Make the ProbNN variables available in Hlt2

 -- MikaVesterinen - 2015-09-24 \ No newline at end of file

Revision 172015-12-18 - SaschaStahl

Line: 1 to 1
 
META TOPICPARENT name="LHCbTrigger"

Tasks for the 2016 HLT commissioning

Line: 15 to 15
 
HLT standard particles ...  
TCKsh, and documentation Anton, Albert, ...  
Resurrection of FEST, and pit installation procedure ..  
Added:
>
>
Make Ghost probability available in Hlt1 ..  
 

Memory profiling

Line: 88 to 89
 

Resurrection of FEST

Added:
>
>

Make Ghost probability available in Hlt1

  -- MikaVesterinen - 2015-09-24

Revision 162015-12-18 - MikaVesterinen

Line: 1 to 1
 
META TOPICPARENT name="LHCbTrigger"

Tasks for the 2016 HLT commissioning

Line: 8 to 8
 
L0 BW division Conor, Eric, Miriam, Alvaro Preliminary results shown at PPG on 17/12/2015
HLT timing and farm loading studies Sascha, ...  
HLT bandwidth division ....  
Changed:
<
<
>
>
Fix all the nightly tests Rosen, Roel, ....  
Add downstream tracking in HLT1 ...  
Speed up the muon lines ...  
HLT emulation for MC ...  
HLT standard particles ...  
TCKsh, and documentation Anton, Albert, ...  
Resurrection of FEST, and pit installation procedure ..  
 

Memory profiling

Line: 47 to 53
 
  • Automating the procedure as much as possible, including tests
  • Clear definition of the emergency release protocol.
Changed:
<
<

Resurrection of FEST

! StdLoose style particle inputs

>
>

! HltStdParticles

  • We shouldn't have every HLT line doing its own filtering on pT, IP etc...
  • Should have stripping style StdLooseXXX
 
Added:
>
>
 

TCKsh improvements, and auto generation of TCK web pages

  • Like these for the stripping
Changed:
<
<

Validation of the Tesla 2015 reprocessing

>
>

New tools for the turbo stream

 
Added:
>
>
 

Trigger emulation for MC

  • For Run-II, this should be straightforward
  • For Run-I, we need something to work with pGun productions (no PV) and fast MC productions without all sub-detectors.
Added:
>
>
 

Speed up Muon lines

  • Use the momentum estimate from MatchVeloMuonTT in the Forward tracking to make it faster.
  • Contact Sascha if interested.
Changed:
<
<

Add Downstream tracks in Hlt1.

  • An upfront Seeding seems impossible but the Seeding has an option to search in a region of interest. This could be provided e.g. by a muon candidate.
  • Contact Manuel and or Sascha if interested.
>
>
 

Testing

  • Currently most of the nightly tests are failing
  • Should also add more tests to the PR framework (independence, memory profile etc...)
Added:
>
>

Add Downstream tracks in Hlt1.

  • An upfront Seeding seems impossible but the Seeding has an option to search in a region of interest. This could be provided e.g. by a muon candidate.
  • Contact Manuel and or Sascha if interested.

Resurrection of FEST

  -- MikaVesterinen - 2015-09-24

Revision 152015-12-17 - MikaVesterinen

Line: 1 to 1
 
META TOPICPARENT name="LHCbTrigger"
Changed:
<
<

Tasks to prepare for 2016 datataking

>
>

Tasks for the 2016 HLT commissioning

 
Added:
>
>
Task People Status
Memory profiling Roel, .... First results at HLT ops 8/1/2015?
Monitoring Kevin and Roel First results at HLT ops 8/1/2015?
L0 BW division Conor, Eric, Miriam, Alvaro Preliminary results shown at PPG on 17/12/2015
HLT timing and farm loading studies Sascha, ...  
HLT bandwidth division ....  

 

Memory profiling

Changed:
<
<
  • A potential improvement in the memory usage would be the LoKi functor cache.
>
>
  • See Roel's slides from the HLT parallel session in the December 2015 LHCb week.
  • We can currently only load the farm to 70-80% because we can't fit enough Moores into the available memory on the nodes
  • It is critical that we try to reduce the memory footprint of Moore
 
Added:
>
>
 

Monitoring (Kevin)

Added:
>
>
* See Roel's slides from the HLT parallel session in the December 2015 LHCb week. * One area that can be improved is the flexibility, by decoupling some of the monitoring from the TCK.
 
  • Roel suggests that in the routing bits, we should separate what is needed for "streams" and what is needed for monitoring.
Added:
>
>
 

L0 Bandwidth division

Changed:
<
<

Detailed study on the farm loading

Define emergency release procedure

>
>
  • See the discussion on the 27th November 2015.
  • We now migrate to an tuple based BW division
    • The latest results were shown by Conor at the PPG on the 17th December 2016

HLT bandwidth division

  • See the HLT output rates here
  • We need to reduce the HLT1 output rate to 150 kHz
  • And we need to reduce the HLT2 output rate to the equivalent of (10+2.5) kHz

Overall timing of the HLT sequence

  • We should have two people responsible for the timing tests
  • And should follow
    • Tracking activities listed here.
    • Activities in the RICH reconstruction. One possible time saving could come from using a faster photon reconstruction option.
    • Activities in the CALO reconstruction.

Improving the HLT/Moore/TCK release procedure

  • Automating the procedure as much as possible, including tests
  • Clear definition of the emergency release protocol.
 

Resurrection of FEST

Deleted:
<
<

Automating the release and testing procedure

 

! StdLoose style particle inputs

Deleted:
<
<

Hlt2 bandwidth division

 

TCKsh improvements, and auto generation of TCK web pages

  • Like these for the stripping

Revision 142015-12-01 - MikaVesterinen

Line: 1 to 1
 
META TOPICPARENT name="LHCbTrigger"

Tasks to prepare for 2016 datataking

Line: 7 to 7
 
  • A potential improvement in the memory usage would be the LoKi functor cache.

Monitoring (Kevin)

Added:
>
>
  • Roel suggests that in the routing bits, we should separate what is needed for "streams" and what is needed for monitoring.
 

L0 Bandwidth division

Revision 112015-11-23 - MikaVesterinen

Line: 1 to 1
 
META TOPICPARENT name="LHCbTrigger"

Tasks to prepare for 2016 datataking

Deleted:
<
<

Resurrection of FEST

Automating the release and testing procedure

 

Memory profiling

Changed:
<
<

! StdLoose style particle inputs

>
>
  • A potential improvement in the memory usage would be the LoKi functor cache.
 

Monitoring (Kevin)

Changed:
<
<

Hlt2 bandwidth division

>
>

L0 Bandwidth division

 

Detailed study on the farm loading

Changed:
<
<

Persist HLT reconstruction and isolation info in the turbo stream

  • To eventually negate the need for any offline reconstruction (and potentially stripping) altogether.
>
>

Resurrection of FEST

Automating the release and testing procedure

 
Changed:
<
<

Bandwidth division

>
>

! StdLoose style particle inputs

Hlt2 bandwidth division

 
Changed:
<
<

Trigger emulation

  • To work with particle gun productions
>
>

Validation of the Tesla 2015 reprocessing

 
Changed:
<
<

LoKi functor cache

  • Makes a significant memory saving in the stripping. To be seen how much impact it could have on Moore with checkpointing....
>
>

Trigger emulation for MC

  • For Run-II, this should be straightforward
  • For Run-I, we need something to work with pGun productions (no PV) and fast MC productions without all sub-detectors.
 

Speed up Muon lines

  • Use the momentum estimate from MatchVeloMuonTT in the Forward tracking to make it faster.

Revision 62015-10-19 - MikaVesterinen

Line: 1 to 1
 
META TOPICPARENT name="LHCbTrigger"
Added:
>
>

Tasks to prepare for 2016 datataking

 
Added:
>
>

Monitoring

Hlt2 bandwidth division

Detailed study on the farm loading

Persist HLT reconstruction and isolation info in the turbo stream

  • To eventually negate the need for any offline reconstruction (and potentially stripping) altogether.

Bandwidth division

 

Trigger emulation

  • To work with particle gun productions

Revision 52015-10-05 - MikaVesterinen

Line: 1 to 1
 
META TOPICPARENT name="LHCbTrigger"

Trigger emulation

Line: 15 to 15
 
  • An upfront Seeding seems impossible but the Seeding has an option to search in a region of interest. This could be provided e.g. by a muon candidate.
  • Contact Manuel and or Sascha if interested.
Added:
>
>

Testing

  • Currently most of the nightly tests are failing
  • Should also add more tests to the PR framework (independence, memory profile etc...)
 -- MikaVesterinen - 2015-09-24 \ No newline at end of file

Revision 42015-10-01 - SaschaStahl

Line: 1 to 1
 
META TOPICPARENT name="LHCbTrigger"

Trigger emulation

Line: 7 to 7
 

LoKi functor cache

  • Makes a significant memory saving in the stripping. To be seen how much impact it could have on Moore with checkpointing....
Added:
>
>

Speed up Muon lines

  • Use the momentum estimate from MatchVeloMuonTT in the Forward tracking to make it faster.
  • Contact Sascha if interested.

Add Downstream tracks in Hlt1.

  • An upfront Seeding seems impossible but the Seeding has an option to search in a region of interest. This could be provided e.g. by a muon candidate.
  • Contact Manuel and or Sascha if interested.
 -- MikaVesterinen - 2015-09-24 \ No newline at end of file
 
This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2019 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding TWiki? Send feedback