search

LSSTDESC / DC2_visitList

Code to create the list of visits to simulate for DC2.
1 stars 0 forks source link

Run 1 vs. Run 2: number of chips that need to be simulated? #2

Open humnaawan opened 6 years ago

humnaawan commented 6 years ago

The total number of WFD visits for Run 2 is larger (50,415) than for Run 1 (8,389), i.e. there are ~6x more visits when we go from 25deg2 to 300deg2. This increase is not equivalent to 6x more sensors since the visits translationally dithered and not all visits contain full sensor rafts.

The number of DD visits in the visit list has not changed (23,160), although only a subset was chosen for Run 1.

In order to estimate the storage/computational needs, we need to know how many sensors need to be simulated.

humnaawan commented 6 years ago

Okay, in an attempt to find the number of sensors, I ran an adapted version of the DC1-chip-counting-code on the DD and WFD visit lists produced for Run 1 (at nside 512) and Run 2 (nside 1024). The latter took forever to run so I decided to produce the Run 2 visit lists at Nside 256 to get some estimates. Here's a summary of the outputs:

Run/Survey Nside nobsIDs nobsIDs (chip-count code) nChips (science sensors) Time taken on login node (hrs)
Run 1 DD 512 23,160 23,160 888,758 16.20
Run 2 DD 256 23,160 23,160 855,299 4.65
Run 2 DD 1024 23,160 23,160 846,139 50.69
--- --- --- --- --- ---
Run 1 WFD 512 8,389 8,374 616,072 10.75
Run 2 WFD 256 51,122 51,063 5,612,971 31.19
Run 2 WFD 1024 50,415 ?? ?? ?

Now, the chip count for DD goes down between Nside 256 and Nside 1024, so if we assume a similar trend for WFD, we'll be in the range of 5M sensors for all WFD Run 2 visits. Comparable to that from 27K undithered visits mentioned in the DC2 planning document, which translate to ~27k x 189= 5,103,000 science sensors.