[photran-dev] Photran Indexer Measurements - Round 2

[Jeffrey Overbey <subscribe-photran-dev@xxxxxxxxxxxx>]

Based on a profile of indexing POP, I changed the serialization 
mechanism and got some improvement.  The database is half as large in 
POP's case (!?!) and the execution times are a little better across the 
board.  Not stellar, but OK.  This will make our discussion on Tuesday 
more interesting.


> Parallel Ocean Program
>     77,900 LOC in 68 files
>     38.9 seconds indexing (several module loading errors)

      Now 28.2 seconds (2,762 lines/sec)
          29.8 with forced write
          20.0 without reference collection or write

> IBEAM (Post-Preprocessing)
>     62,026 LOC in 220 files
>     51.2 seconds indexing

      Now 34.9 seconds (1,777 lines/sec)
          36.5 with forced write
          27.7 without reference collection or write

> Atmospheric Dispersion Simulation
>     52,968 LOC in 29 files
>     19.1 seconds indexing (3 fixed form parsing errors)

      Now 14.9 seconds (3,555 lines/sec)
          15.3 with forced write
          11.2 without reference collection or write

> FMLIB
>     56,418 LOC in 5 files
>     1:18.7 indexing

      Now 1:11.3 (803 lines/sec)
          1:15.2 with forced write
          0:47.5 without reference collection or write

> LAPACK
>     910,973 LOC in 2,811 files
>     6:26.5 indexing (many fixed form parsing errors)

      Now 4:57.4 (3,067 lines/sec)
          5:10.2 with forced write
          2:22.7 without reference collection or write