Wednesday 01/23/13

10:30 A – 8:30P, 9:30P – 11:45P

(8:45-10:30A workout)

Fly Work

  • collect virgins: Df/CyO,esc[2],  2x mat-tub Gal4.
  • collect males: esc[6]/Sp.
  • cross (1 new bottle) esc[6]/Sp x Df/CyO,esc[2]
  • (should have used Gla, will need to screen Sp for all flies that go into collection cage)
  • Cleared sim[D] and Espl[D], collect one more round of virgins for one more round of backcross screening.  Should have decided this earlier, bottles are mostly past it.

STORM

  • Calibrate STORM2 setup
  • STORM imaging of BX-C cy5 single label.  100,00 + frames, no activator 405 required.
  • choosing cells to continue imaging O/N. — slide slipped, positions lost.  Restarted steve.
  • 8P, O/N imaging up and running of BX-C complexes.  still mostly switching for 100,000 frames without activator

Coding

  • debugged GPU code, now runs through STORMfinder
  • multifit with overlap up to 5 causes inevitable crash for all size movies.
  • multifit with 2 works well and fast.
  • multifitter finds too many molecules to keep in memory (>15 GB by frame 9000 of HP1).  Need to write molecules straight to disk, not save all in memory.   Should also convert molecules to singles (should double the memory available).  — This is already handled  by GPUmulti.
  • found bug in call of GPU multi, was analyzing progressively larger daxfiles from start instead of analyzing in chunks.  Now fixed.  solves memory handling things too.
  • GPU multifit currently running about 4,000 frames / sec ~11,000 molecules.  (+30 seconds of load time for dax file and send to GPU, should definetely increase batch size).

 

This entry was posted in Summaries and tagged , , . Bookmark the permalink.