Re: [PATCH v5 3/5] idle: enable IRQ in default idle


Chen, Jason CJ
 

On Mon, Aug 13, 2018 at 09:26:05AM +0800, Yin, Fengwei wrote:

On 08/12/2018 09:19 PM, Chen, Jason CJ wrote:
we should let cpu handle IRQ during idle
This may impact the performance of UOS. Can we get the performance data
with gvt test for this change?

Regards
Yin, Fengwei
hi, Fengwei,

I had a test based on Android GFX bench, there is no performance drop for
offscreen & T-Rex test.



Signed-off-by: Jason Chen CJ <jason.cj.chen@...>
---
hypervisor/common/schedule.c | 5 ++++-
1 file changed, 4 insertions(+), 1 deletion(-)

diff --git a/hypervisor/common/schedule.c b/hypervisor/common/schedule.c
index 1670d5f..a56d59c 100644
--- a/hypervisor/common/schedule.c
+++ b/hypervisor/common/schedule.c
@@ -172,8 +172,11 @@ void default_idle(void)
schedule();
} else if (need_offline(pcpu_id) != 0) {
cpu_dead(pcpu_id);
- } else
+ } else {
+ CPU_IRQ_ENABLE();
cpu_do_idle(pcpu_id);
+ CPU_IRQ_DISABLE();
+ }
}
}
--

Thanks

Jason

Join acrn-dev@lists.projectacrn.org to automatically receive all group messages.