> And I'm surprised if you read this far. Summary: i^=1 is not likely to > win comapred to ++i, unless we increase dict size 33%. Right! I had tried i^=1 and it had near zero or slightly negative effects on performance. It resulted in more collisions, though the collisions were resolved relatively cheaply. I had also experimented with changing alignment, but nothing helped. Everything is already word aligned and that takes care of the HW issues. The only benefit to the alignment is that i^=1 guarantees a cache hit. Without alignment, the odds are 4 out of 5.3 will have a hit (since there a 5.3 entries to a line). Increasing the dict size 33% with unused space doesn't help sparseness and negatively impacts the chance cache hits you already have with smaller dictionaries. heyhey-mymy-there's-more-to-the-picture-than-meets-the-eye, Raymond Hettinger
RetroSearch is an open source project built by @garambo | Open a GitHub Issue
Search and Browse the WWW like it's 1997 | Search results from DuckDuckGo
HTML:
3.2
| Encoding:
UTF-8
| Version:
0.7.4