collperf

collperf is a test program for comparing collation performance and key lengths of ICU, Windows native collation and Unix/POSIX collation. It operates on a file of lines (names, for example), and performs one of three tests:

1. Sort Key generation. Report on key lengths and key generation times.
2. Binary search. Report the average time required to look up each of the names (file lines) from the file in a sorted list of all of the names.
3. Quick Sort. Report the time required to sort the file in memory, using the C library qsort function. The file order is randomized prior to the sort.

Usage Summary

collperf -help Usage: strperf options... -help Display this message. -file file_name utf-16 format file of names -locale name ICU locale to use. Default is en_US -langid 0x1234 Windows Language ID number. Default 0x409 (en_US) see http://msdn.microsoft.com/library/psdk/winbase/nls_8xo3.htm -win Run test using Windows native services. (ICU is default) -unix Run test using Unix strxfrm, strcoll services. -uselen Use API with string lengths. Default is null-terminated strings -usekeys Run tests using sortkeys rather than strcoll -loop nnnn Loopcount for test. Adjust for reasonable total running time. -terse Terse numbers-only output. Intended for use by scripts. -french French accent ordering -norm Normalizing mode on -shifted Shifted mode -lower Lower case first -upper Upper case first -case Enable separate case level -level n Sort level, 1 to 5, for Primary, Secndary, Tertiary, Quaternary, Identical -binsearch Binary Search timing test -keygen Sort Key Generation timing test -qsort Quicksort timing test

Example

C:\>collperf -loop 200 -file latin.txt -keygen -shifted -level 4 file "latin.txt", 7604 lines. Sort Key Generation: total # of keys = 197704 Sort Key Generation: time per key = 4253 ns Key Length / character = 1.730054