Oopenvela-robotsched: unify the thread exit running task behavior.
| 文件 | 最后提交记录 | 最后更新时间 |
|---|---|---|
sched: unify the thread exit running task behavior. We already split switch_context into nxsched_suspend_scheduler and nxsched_resume_scheduler, when task_exit, set running = NULL for more compatible. The switch context is much more frequency than task_exit, we should prefer decrease the switch_context record time cost. Also the x_doirq xcp.regs save will be more unified. Signed-off-by: wangzhi16 <wangzhi16@xiaomi.com> Signed-off-by: ligd <liguiding1@xiaomi.com> | 2 个月前 | |
arch: delete -fno-optimize-sibling-calls Signed-off-by: chenxiaoyi <chenxiaoyi@xiaomi.com> | 2 个月前 | |
build: Remve the unnecessary .gitignore Signed-off-by: Xiang Xiao <xiaoxiang@xiaomi.com> | 2 个月前 | |
uarch/make: use UOBJEXT When building KERNEL mode with PERF_EVENTS_HAVE_USER_ACCESS enabled via make system, we see two issues: - Warning target 'arm_perf.o' given more than once in the same rule - Non-parallel make -j1 often fails, though parallel make succeeds. Based on brainstorming with Xin, this fixes above by using UOBJEXT suffix for userspace objects so that same source can be used for both user and kernel space objects in the same folder. This is also extended to other archs, including ceva which uses sort approach to avoid warnings. Per test with qemu-armv7a sort approach still has make -j1 issue. Signed-off-by: Yanfeng Liu <p-liuyanfeng9@xiaomi.com> | 2 个月前 |