Now, I would like to see REAL assembler code for a binary search that takes into consideration all the items being discussed and works for a varying length tables, including an odd number of rows in the table. To make it simple, assume: R2 = @ first entry R3 = @ last entry R4 = length of each entry R5 = @ of key to match offset and length of key are EQUs KOFF and KLGH
Tony Thigpen -----Original Message ----- From: Don Higgins Sent: 10/24/2013 06:50 AM
All I see discussion about optimizing the binary search, but I don't see any discussion about optimizing linear search which might make it much faster than binary depending on the search history. Putting the last key found at front of most recently used list or just holding last found as a first test may result in significant reduction in total search compares if there are frequent repeat requests. Don Higgins d...@higgins.net