[backend](cuda): faster uncontiguous concat #17496
| Job | Run time |
|---|---|
| 12m 18s | |
| 19m 2s | |
| 6m 29s | |
| 12m 7s | |
| 3m 10s | |
| 2m 54s | |
| 2m 34s | |
| 2m 52s | |
| 1m 10s | |
| 11m 29s | |
| 3m 2s | |
| 1m 2s | |
| 5m 2s | |
| 1m 6s | |
| 1m 50s | |
| 3m 57s | |
| 4m 37s | |
| 2m 47s | |
| 1m 46s | |
| 4m 34s | |
| 3m 5s | |
| 1m 39s | |
| 11m 54s | |
| 3m 56s | |
| 4m 41s | |
| 12m 11s | |
| 2m 56s | |
| 4m 54s | |
| 5m 12s | |
| 4m 54s | |
| 5m 6s | |
| 6m 56s | |
| 4m 47s | |
| 5m 10s | |
| 5m 50s | |
| 3m 46s | |
| 3m 53s | |
| 6m 21s | |
| 15m 24s | |
| 0s | |
| 0s | |
| 3h 36m 23s |