dayblob tests

Database name:	wdi93
SMILES rooted datatrees:	37037

"Chal"	SGI Challenge L, 2x100MHz MIPS R4400, IRIX 5.1.1.2
"Sun1K"	Sun SparcServer-1000, 8x50MHz, SuperSparc CPUs, SunOS 5.3
"Sun60"	Sun Ultra-60, 2x360MHz SuperSparc CPUs, SunOS 5.6

1a	Stringsearch in SMILES for "As" (93 hits)
1b	Stringsearch in SMILES for "Au" (42 hits)
1c	Regular expression search in SMILES for "A[su]" (134 hits)
1d	Superstructure search for [As] (93 hits)
1e	Superstructure search for [Au] (42 hits)
2a	Superstructure search for N#CN=CN (44 hits)
2b	Superstructure search + sort for N#CN=CN (44 hits)
3a	Superstructure search for dopamine (1086 hits)
3b	Superstructure search + sort for dopamine (1086 hits)
3c	Superstructure SMARTS-search for dopamine (1086 hits)
4a	Superstructure search for cefalosporin-g1 (243 hits)
4b	Superstructure search + sort for cefalosporin-g1 (243 hits)
4c	Superstructure SMARTS-search for cefalosporin-g1 (243 hits)
5a	Similarity search (high, T > 0.75) for cefedrolor (35 hits)
5b	Similarity sort for cefedrolor
5c	Superstructure search for cefedrolor (1 hit)
5d	Substructure search for cefedrolor (70 hits)
6a	ASCII sort of SMILES
6b	Length-sort of SMILES
6c	Numerical sort of molecular weight
6d	Numerical value search for molecular weight 300-310 (1136 hits)

dayblob supports only structural searching; tests of merlinserver functionality which are not available in dayblob (e.g., string searching, sorting) are indicated by '~'.
dayblob uses 512-bit fixed fingerprints vs. folded fingerprints used by merlinserver, resulting in difference in results of test 5a.
merlinserver461 provides special-purpose implementations of certain functions which earlier versions and dayblob do not. Sorting a hitlist by similarity is one of these, which explains the 45x performance difference in test 5b.
Challenge/4.32 is the standard against which other hardware and software versions are compared (this test suite was established 4-1/2 years ago). For typical structural searches, dayblob is 3.4x faster than merlinserver432 running on a Challenge L. Including pathological cases, it is 2.5x faster.
The Sun 1000 is another machine in the same generation as the Challenge, which is shown for comparison only.
Sun Ultra60/4.61 is a current high-end workstation running the latest Daylight software. For typical structural searches, dayblob is about the same speed as merlinserver on an Ultra60 (30% slower to 15% faster, depending on query). Including pathological cases, it is 2.5x slower.