JakeHillion/scx
1
0
Fork 0
mirror of https://github.com/JakeHillion/scx.git synced 2024-11-30 04:50:24 +00:00
Code Issues Packages Projects Releases Wiki Activity

scx_rustland: bypass user-space scheduler for short-lived kthreads

Browse Source 
Bypass the user-space scheduler for kthreads that still have more than
half of their runtime budget.

As they are likely to release the CPU soon, granting them a substantial
priority boost can enhance the overall system performance.

In the event that one of these kthreads turns into a CPU hog, it will
deplete its runtime budget and therefore it will be scheduled like
any other normal task through the user-space scheduler.

Signed-off-by: Andrea Righi <andrea.righi@canonical.com>
This commit is contained in:
Andrea Righi 2023-12-31 16:35:44 +01:00
parent 405a11308e
commit a7677fdf28
1 changed files with 14 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

15
scheds/rust/scx_rustland/src/bpf/main.bpf.c
View File

@ -314,7 +314,7 @@ static bool is_task_cpu_available(struct task_struct *p, u64 enq_flags)
struct task_ctx *tctx;
/*
* Always dispatch per-CPU kthread on the same CPU, bypassing the
* Always dispatch per-CPU kthreads on the same CPU, bypassing the
* user-space scheduler (in this way we can to prioritize critical
* kernel threads that may potentially slow down the entire system if
* they are blocked for too long).
@ -322,6 +322,19 @@ static bool is_task_cpu_available(struct task_struct *p, u64 enq_flags)
if (is_kthread(p) && p->nr_cpus_allowed == 1)
return true;
/*
* Moreover, immediately dispatch kthreads that still have more than
* half of their runtime budget. As they are likely to release the CPU
* soon, granting them a substantial priority boost can enhance the
* overall system performance.
*
* In the event that one of these kthreads turns into a CPU hog, it
* will deplete its runtime budget and therefore it will be scheduled
* like any other normal task.
*/
if (is_kthread(p) && p->scx.slice > slice_ns / 2)
return true;
/*
* No scheduling required if it's the last task running.
*/