| Scott James Remnant | 56686d6 | 2009-11-09 18:38:51 +0000 | [diff] [blame] | 1 | TODO: |
| 2 | |
| Scott James Remnant | 56686d6 | 2009-11-09 18:38:51 +0000 | [diff] [blame] | 3 | * Continue to investigate how to improve performance of the open() pass |
| 4 | on hard-drives. |
| Scott James Remnant | cc2943b | 2009-11-29 15:24:15 +0000 | [diff] [blame] | 5 | |
| 6 | - The inode preloading is a benefit, but the threshold of how many hits |
| 7 | is a delicate one and seems to vary; worthy of further investigation. |
| 8 | |
| 9 | - It seems that doing the preloading in a separate pass to open() is |
| 10 | generally better; I tried preloading inode groups before the first |
| 11 | open() call in that inode group, but that was slower. |
| 12 | |
| 13 | - Likewise it seems that the separate open() pass is much better, read |
| 14 | ahead took twice as long when the calls were mixed. |
| 15 | |
| 16 | - How to sort the paths? By inode group seems to be the best so far, |
| 17 | giving a slight edge over by path. |