james...
thanks for the memory information. I'll try to digest it in more detail.
You are correct that we are being sloppy in out memory usage at this
point, and we need to clean some things up.
I see one routine that really sucks up memory
libtrk.a[trk_dcgeom.o]: 24932 + 2416 + 167988 = 195336
Which has the wire positions for EACH wire. I know for a fact that
Bogdan only has 1/6 of this information...so SDA will grow once
Bogdan realizes that the drift chambers are not perfectly positioned.
The other hogs are:
libtrk.a[sda_dcgeom.o]: 29065 + 1240 + 167732 = 198037
libuser.a[user_sda.o]: 40577 + 1576 + 580 = 42733
libtrk.a[sda_geom.o]: 36350 + 4440 + 476 = 41266
libtrk.a[sda_ginit.o]: 39719 + 4116 + 228 = 44063
libdc.a[dc_getcal_geom.o]: 33456 + 7596 + 166716 = 207768
libdc.a[dc_getcal_tdly.o]: 8690 + 1360 + 166336 = 176386
libsc.a[scfit_sector.o]: 50005 + 1616 + 1932 = 53553
librecutl.a[reclog.o]: 10589 + 1372 + 33580 = 45541
librecutl.a[hsttcl_init.o]: 39845 + 7548 + 404 = 47797
libec.a[ecfit_peaks_hits.o]: 23334 + 1192 + 28740 = 53266
libec.a[ec_fill_bos.o]: 19881 + 1332 + 68716 = 89929
We should look at these routines carefully ....
Arne
First:
libec.a[ecfit_peaks_hits.o]: 23334 + 1192 + 28740 = 53266
we have subroutine ecfph.F which is doing even more then ecfit_peaks_hits.F and
has about 2 times smaller size.
Second:
libec.a[ec_fill_bos.o]: 19881 + 1332 + 68716 = 89929
here we still keeping two output BOS banks - ECRB and ECHB. After final
dicision having one of them this routine will use at least 2 times less memory.
regards Stepan.