(In reply to Paul Bone [:pbone] from comment #11) > What do the last two columns measure, the time for nursery collections/major slices/elapsed time etc? So I don't know which rows are good/bad. The last two columns are average time per minor GC and total time spent in minor GC (note different units). > Did you test with poisoning disabled? Annoyingly I did not, so I've re-run the tests. New results: Path: Major Minor Average Average Average Total Count Count PR nursery time /uS time /mS size /MB output-19-01-10.txt 117 266 21.6% 13 5123 1362 output-19-01-15.txt 56 3530 0.5% 1 51 180 output-19-01-20.txt 79 3451 0.6% 1 57 200 output-19-01-25.txt 103 253 21.1% 13 5551 1404 output-19-01-30.txt 111 268 22.1% 12 5038 1350 output-19-02-05.txt 140 3017 0.7% 1 61 185 output-19-02-10.txt 157 2735 0.6% 1 57 156 output-19-02-15.txt 149 14934 0.4% 1 20 312 output-19-02-20.txt 169 540 8.6% 6 1397 754 output-19-02-25.txt 155 13807 0.7% 1 21 298 output-19-02-28.txt 145 13355 0.3% 1 20 271 Observations: 1. It seems random whether or not it goes into the mode where we tenure more and use a larger nursery. I don't think this is related to a change in the code -- in other words, it's still a problem. 2. Discounting runs where the above problem happened, builds from 19-02-15 with the smaller nursery size do ~4 times more minor GCs and spend ~50% more time in minor GC.
Bug 1525983 Comment 12 Edit History
Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.
(In reply to Paul Bone [:pbone] from comment #11) > What do the last two columns measure, the time for nursery collections/major slices/elapsed time etc? So I don't know which rows are good/bad. The last two columns are average time per minor GC and total time spent in minor GC (note different units). > Did you test with poisoning disabled? Annoyingly I did not, so I've re-run the tests. New results: ``` Path: Major Minor Average Average Average Total Count Count PR nursery time /uS time /mS size /MB output-19-01-10.txt 117 266 21.6% 13 5123 1362 output-19-01-15.txt 56 3530 0.5% 1 51 180 output-19-01-20.txt 79 3451 0.6% 1 57 200 output-19-01-25.txt 103 253 21.1% 13 5551 1404 output-19-01-30.txt 111 268 22.1% 12 5038 1350 output-19-02-05.txt 140 3017 0.7% 1 61 185 output-19-02-10.txt 157 2735 0.6% 1 57 156 output-19-02-15.txt 149 14934 0.4% 1 20 312 output-19-02-20.txt 169 540 8.6% 6 1397 754 output-19-02-25.txt 155 13807 0.7% 1 21 298 output-19-02-28.txt 145 13355 0.3% 1 20 271 ``` Observations: 1. It seems random whether or not it goes into the mode where we tenure more and use a larger nursery. I don't think this is related to a change in the code -- in other words, it's still a problem. 2. Discounting runs where the above problem happened, builds from 19-02-15 with the smaller nursery size do ~4 times more minor GCs and spend ~50% more time in minor GC.