Off Line Processing of Large GDSII Files II

Some figures will be helpful in showing the advantages of off line processing.


Normal Operation
We've measured the scan/quad-tree build time for a wide range of GDSII files. This is the time required to "begin" the display process. (depending on how much detail you have requested the screen draw may take only a second or many seconds ...)

One can see that at around 4GB it takes 7 1/2 minutes to get to the initial display -- this is enough time to run down the hall, pick up a coffee, gossip a bit and return. However the 14 GB file takes 33 minutes to initially display -- more than enough time to run to the cafeteria and eat a hamburger, fries and coke.

These measurements were made on a Sun Ultra Sparc Blade equipped with 2X 750MHz CPU's and 8 GBytes of RAM. (the dual CPU's are of no help as the scan and quad-tree processes cannot run in parallel or otherwise benefit from multiple processors.)

 
File Size

(MB)
Scan Time

(sec)
Quad Tree
Load/Build
(sec)
Time to
Initial Display
(min)
314 23 46 1
417 24 25 < 1
496 39 53 1.5
679 53 72 2
694 41 43 1.5
766 46 48 1.6
1058 78 75 2.5
1631 117 163 4.7
1862 137 140 4.6
1985 145 143 4.8
2075 128 193 5.4
3797 219 231 7.5
4710 237 343 9.66
14,177 889 1079 32.8

Timings for Preprocessed GDSII Files

Let's say that the night before we ran all of the above files through the off-line processing so that we have memory maps of the scan/quad data prior to coming in for the day's work. Now how long does it take to open each file?





For the smaller files you will see the initial display before you've taken more than two breaths ... and even for all but the largest file the time to get to the initial display is a fraction of a minute.

 
File Size

(MB)
Scan
Load Time
(sec)
Quad Tree
Load
(sec)
Time to
Initial Display
(min)
314 0 7 0.11
417 0 5 0.08
496 0 9 0.15
679 0 13 0.21
694 1 7 0.13
766 1 5 0.08
1058 0 7 0.11
1631 1 22 0.38
1862 0 17 .28
1985 0 14 0.23
2075 0 41 0.68
3797 0 24 0.40
4710 0 34 0.57
14,177 13 83 1.60


Page:   1   |   [2]   |   3