New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Doublet tuning and new vertex algo #158

Merged

fwyzard merged 38 commits into cms-patatrack:CMSSW_10_2_X_Patatrack from VinInn:DoubletTuningAndVertex

Sep 12, 2018

VinInn commented Sep 4, 2018 •

edited by fwyzard

Loading

This PR contains two independent components

tuning and speed up of the doublet algorithm
a new concurrent vertex algorithm based on DBSCAN
(it reuses pieces of pixel cluster algo...)

The vertex algorithm will require tuning (to say the least) even to reach the performance of current DivisiveClustering algo used at HLT.

MTV:
http://innocent.home.cern.ch/innocent/RelVal/gpuverI2/
current DivisiveClustering algo:
https://innocent.web.cern.ch/innocent/RelVal/divverI/

Baseline is offline doublets and DA vertexing.

Rebased onto master from #149.

VinInn and others added 30 commits

August 13, 2018 09:44


          use hit to sort (not faster)

4d3a0a5


          reduce memory, optimize bin iteration

b148a8e


          reduce memory occpancy, use jmax

bc7615d


          remove redundant checks

5e272ec


          new pt cut

13b741a


          new interface


          prototype of vertex finder on gpu

6214e0f


          initial commit

a8bdab1


          more code, does not compile yet

5047e6d


          compiles

3dc134e


          works, disabled to run old divisive...

6e96759


          now gpu algo running

f0963c4


          tranfer reverse pointers

57b8db3


          added tracks to vertex, printouts removed

68fad3d


          expose all params

1ce01f2


          added chi2, error rescaled (small effect)

fc770d6


          sort in pt2

10ad7e9


          more doc, some cleanup

3d67018


          as usual the test is more complex than the algo

bcb7c44


          use new utility function

cdb91a3


          properly rescale error with chi2

794a796


          support 8 and 64 bit int

aa8913e


          support unsigned

80308a5


          SFINAE

77dcfa6


          const

f156524


          derivation of fast dphi cut

fd5886b


          python notebook to tune cuts for doublets and triplets

4c16b56


          Add MTV instance for pixel tracks from PV

88ff2cd


          transfrom errors to global

4f0b74f


          fix copy to gpu

08467e6

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

f4e4bea

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

fdd752b

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

0a62757

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Add new concurrent vertex algo (#158)

43a044c

Add a new concurrent vertex algorithm based on DBSCAN (that reuses parts of the pixel cluster algorithm).
It still needs to be tuned to reach at least the performance of current DivisiveClustering algorithm used at HLT.

fwyzard pushed a commit that referenced this pull request


          Add new concurrent vertex algo (#158)

bf36f0d

Add a new concurrent vertex algorithm based on DBSCAN (that reuses parts of the pixel cluster algorithm).
It still needs to be tuned to reach at least the performance of current DivisiveClustering algorithm used at HLT.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

8515df7

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

b4bbc39

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

d27ae35

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

d4a1999

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

a444afb

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Add new concurrent vertex algo (#158)

94ec80e

Add a new concurrent vertex algorithm based on DBSCAN (that reuses parts of the pixel cluster algorithm).
It still needs to be tuned to reach at least the performance of current DivisiveClustering algorithm used at HLT.

fwyzard pushed a commit that referenced this pull request


          Add new concurrent vertex algo (#158)

3c15f00

Add a new concurrent vertex algorithm based on DBSCAN (that reuses parts of the pixel cluster algorithm).
It still needs to be tuned to reach at least the performance of current DivisiveClustering algorithm used at HLT.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

4803ceb

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Add new concurrent vertex algo (#158)

9f9e43b

Add a new concurrent vertex algorithm based on DBSCAN (that reuses parts of the pixel cluster algorithm).
It still needs to be tuned to reach at least the performance of current DivisiveClustering algorithm used at HLT.

fwyzard added a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

a711fe9

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard added a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

7a85217

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

c33e812

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

22c9309

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard added a commit that referenced this pull request


          Add new concurrent vertex algo (#158)

32a6715

Add a new concurrent vertex algorithm based on DBSCAN (that reuses parts of the pixel cluster algorithm).
It still needs to be tuned to reach at least the performance of current DivisiveClustering algorithm used at HLT.

fwyzard added a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

72d8383

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

cf35be0

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Add new concurrent vertex algo (#158)

293da66

Add a new concurrent vertex algorithm based on DBSCAN (that reuses parts of the pixel cluster algorithm).
It still needs to be tuned to reach at least the performance of current DivisiveClustering algorithm used at HLT.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

ca90611

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Add new concurrent vertex algo (#158)

8c27094

Add a new concurrent vertex algorithm based on DBSCAN (that reuses parts of the pixel cluster algorithm).
It still needs to be tuned to reach at least the performance of current DivisiveClustering algorithm used at HLT.

fwyzard pushed a commit that referenced this pull request


          Add new concurrent vertex algo (#158)

ebe462b

Add a new concurrent vertex algorithm based on DBSCAN (that reuses parts of the pixel cluster algorithm).
It still needs to be tuned to reach at least the performance of current DivisiveClustering algorithm used at HLT.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

fdba916

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

5673b02

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

fwyzard pushed a commit that referenced this pull request


          Tune and speed up doublet algo (#158)

f76d1a9

Tune and speed up the pixel doublet alforithm, and take advantage of GPU read-only memory for a further speedup.

Includes a python notebook to tune the cuts for doublets and triplets.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels