[Intel-gfx] [i-g-t] Fixing the latency of hrtimer

[Kaparthi, SowmyaX]

Hi Lionel,

Thanks for your suggestions. I will send the patch to igt-dev at lists.freedesktop.org.

This patch is regarding the git lab issue #1542. Based on the ftrace, a 2ms hrtimer results in around 5000 calls to read (because test duration is 10s). Based on average read times (30us), the total kernel_ns is larger than the value being compared and this fix is to allow user to have a larger timer value than the default 5ms to avoid waking up too frequently.

Thanks and Regards
Sowmya

-----Original Message-----
From: Landwerlin, Lionel G <lionel.g.landwerlin at intel.com> 
Sent: 27 August 2020 13:20
To: Kaparthi, SowmyaX <sowmyax.kaparthi at intel.com>; intel-gfx at lists.freedesktop.org; Nerlige Ramappa, Umesh <umesh.nerlige.ramappa at intel.com>
Subject: Re: [i-g-t] Fixing the latency of hrtimer

Hi Sowmya,

Thanks for the patch. If you could send it to the igt-dev at lists.freedesktop.org list instead, this is where the IGT patches go.

Could you refresh my memory as to what this is fixing?
It sounds like this is just adjusting a value to match more common settings.

Cheers,

-Lionel

On 27/08/2020 10:38, Sowmya Kaparthi wrote:
> The blocking/polling parameterized tests were introduced to test 
> different hrtimer configurations.These tests check how many times the 
> process wakes up to read the reports with different hrtimer values (= 
> duration of test / hrtimer value). A user is more likely to choose a 
> larger hrtimer value than the default 5ms to avoid wake up too frequently.
>
> Cc: Landwerlin, Lionel G <lionel.g.landwerlin at intel.com>
> Signed-off-by: Sowmya Kaparthi <sowmyax.kaparthi at intel.com>
> ---
>   tests/i915/perf.c | 4 ++--
>   1 file changed, 2 insertions(+), 2 deletions(-)
>
> diff --git a/tests/i915/perf.c b/tests/i915/perf.c index 
> a894fd38..5fd1193f 100644
> --- a/tests/i915/perf.c
> +++ b/tests/i915/perf.c
> @@ -4995,7 +4995,7 @@ igt_main
>   			      40 * 1000 * 1000 /* default 40ms hrtimer */);
>   		test_blocking(500 * 1000 /* 500us oa period */,
>   			      true /* set_kernel_hrtimer */,
> -			      2 * 1000 * 1000 /* default 2ms hrtimer */);
> +			      10 * 1000 * 1000 /* default 10ms hrtimer */);
>   	}
>   
>   	igt_describe("Test polled read with default hrtimer frequency"); @@ 
> -5014,7 +5014,7 @@ igt_main
>   			     40 * 1000 * 1000 /* default 40ms hrtimer */);
>   		test_polling(500 * 1000 /* 500us oa period */,
>   			     true /* set_kernel_hrtimer */,
> -			     2 * 1000 * 1000 /* default 2ms hrtimer */);
> +			     10 * 1000 * 1000 /* default 10ms hrtimer */);
>   	}
>   
>   	igt_describe("Test polled read with buffer size smaller than 
> available data");