Automatic NUMA Balancing uses a multi-stage pass to decide whether a page should migrate to a local node. This filter avoids excessive ping-ponging if a page is shared or used by threads that migrate cross-node frequently. Threads inherit both page tables and the preferred node ID from the parent. This means that threads can trigger hinting faults earlier than a new task which delays scanning for a number of seconds. As it can be load balanced very early in its lifetime there can be an unnecessary delay before it starts migrating thread-local data. This patch migrates private pages faster early in the lifetime of a thread using the sequence counter as an identifier of new tasks. With this patch applied, STREAM performance is the same as 4.17 even though processes are not spread cross-node prematurely. Other workloads showed a mix of minor gains and losses. This is somewhat expected most workloads are not very sensitive to the starting conditions of a process. 4.19.0-rc5 4.19.0-rc5 4.17.0 numab-v1r1 fastmigrate-v1r1 vanilla MB/sec copy 43298.52 ( 0.00%) 47335.46 ( 9.32%) 47219.24 ( 9.06%) MB/sec scale 30115.06 ( 0.00%) 32568.12 ( 8.15%) 32527.56 ( 8.01%) MB/sec add 32825.12 ( 0.00%) 36078.94 ( 9.91%) 35928.02 ( 9.45%) MB/sec triad 32549.52 ( 0.00%) 35935.94 ( 10.40%) 35969.88 ( 10.51%)
Unified · Split
| @@ -1392,6 +1392,17 @@ bool should_numa_migrate_memory(struct task_struct *p, struct page * page, · 1391 unmodified lines | |||
| 1392 | int last_cpupid, this_cpupid; | 1392 | int last_cpupid, this_cpupid; |
| 1393 | 1393 | ||
| 1394 | this_cpupid = cpu_pid_to_cpupid(dst_cpu, current->pid); | 1394 | this_cpupid = cpu_pid_to_cpupid(dst_cpu, current->pid); |
| 1395 | last_cpupid = page_cpupid_xchg_last(page, this_cpupid); | ||
| 1396 | |||
| 1397 | /* | ||
| 1398 | * Allow first faults or private faults to migrate immediately early in | ||
| 1399 | * the lifetime of a task. The magic number 4 is based on waiting for | ||
| 1400 | * two full passes of the "multi-stage node selection" test that is | ||
| 1401 | * executed below. | ||
| 1402 | */ | ||
| 1403 | if ((p->numa_preferred_nid == -1 || p->numa_scan_seq <= 4) && | ||
| 1404 | (cpupid_pid_unset(last_cpupid) || cpupid_match_pid(p, last_cpupid))) | ||
| 1405 | return true; | ||
| 1395 | 1406 | ||
| 1396 | /* | 1407 | /* |
| 1397 | * Multi-stage node selection is used in conjunction with a periodic | 1408 | * Multi-stage node selection is used in conjunction with a periodic |
| @@ -1410,7 +1421,6 @@ bool should_numa_migrate_memory(struct task_struct *p, struct page * page, · 12 unmodified lines | |||
| 1410 | * This quadric squishes small probabilities, making it less likely we | 1421 | * This quadric squishes small probabilities, making it less likely we |
| 1411 | * act on an unlikely task<->page relation. | 1422 | * act on an unlikely task<->page relation. |
| 1412 | */ | 1423 | */ |
| 1413 | last_cpupid = page_cpupid_xchg_last(page, this_cpupid); | ||
| 1414 | if (!cpupid_pid_unset(last_cpupid) && | 1424 | if (!cpupid_pid_unset(last_cpupid) && |
| 1415 | cpupid_to_nid(last_cpupid) != dst_nid) | 1425 | cpupid_to_nid(last_cpupid) != dst_nid) |
| 1416 | return false; | 1426 | return false; |