[igt-dev] [PATCH i-g-t] tests/kms_rotation_crc: enable and optimize test for sim

[Juha-Pekka Heikkila]

On 25.5.2023 14.47, Swati Sharma wrote:
> Added 2 checks. First, if rotation is supported by plane or not
> before creating fb. This is generic change to optimize test.
> Second check, is to reduce execution time in simulation
> by dropping RGB565 and PO10 formats (coverage is not
> compromised, we do have NV12 and RGB888 formats being validated).
> 
> Signed-off-by: Swati Sharma <swati2.sharma at intel.com>
> ---
>   tests/kms_rotation_crc.c | 19 ++++++++++++++++++-
>   1 file changed, 18 insertions(+), 1 deletion(-)
> 
> diff --git a/tests/kms_rotation_crc.c b/tests/kms_rotation_crc.c
> index 28b4fd27..07040e31 100644
> --- a/tests/kms_rotation_crc.c
> +++ b/tests/kms_rotation_crc.c
> @@ -847,9 +847,17 @@ static void test_multi_plane_rotation(data_t *data, enum pipe pipe)
>   			p[0].fbinfo = &planeconfigs[i];
>   			pointlocation(data, p, mode, 0);
>   
> +			if (!igt_plane_has_rotation(p[0].plane, planeconfigs[i].rotation))
> +				continue;
> +
>   			for (k = 0; k < ARRAY_SIZE(multiplaneformatlist); k++) {
>   				p[0].formatindex = k;
>   
> +				if (igt_run_in_simulation() &&
> +				   (multiplaneformatlist[k] == DRM_FORMAT_P010 ||
> +				    multiplaneformatlist[k] == DRM_FORMAT_RGB565))
> +					continue;
> +
>   				for (j = 0; j < ARRAY_SIZE(planeconfigs); j++) {
>   					p[1].fbinfo = &planeconfigs[j];
>   					pointlocation(data, p, mode, 1);
> @@ -857,6 +865,14 @@ static void test_multi_plane_rotation(data_t *data, enum pipe pipe)
>   					for (l = 0; l < ARRAY_SIZE(multiplaneformatlist); l++) {
>   						p[1].formatindex = l;
>   
> +						if (!igt_plane_has_rotation(p[1].plane, planeconfigs[k].rotation))
> +							continue;
> +
> +						if (igt_run_in_simulation() &&
> +						   (multiplaneformatlist[k] == DRM_FORMAT_P010 ||
> +						    multiplaneformatlist[k] == DRM_FORMAT_RGB565))
> +							continue;
> +
>   						if (!multiplaneskiproundcheck(data, p))

Looking at those new rules for skipping round I think they could all be 
put inside this multiplaneskiproundcheck(..) so the main loop would be 
less cluttered.

>   							continue;
>   
> @@ -872,7 +888,8 @@ static void test_multi_plane_rotation(data_t *data, enum pipe pipe)
>   
>   							retcrc_sw = crclog[logindex];
>   							have_crc = true;
> -						} else if(reusecrcfromlastround(p, lastroundjformat,
> +						} else if(!igt_run_in_simulation() &&
> +							  reusecrcfromlastround(p, lastroundjformat,
>   										lastroundirotation,
>   										lastroundjrotation)) {

Here also simulation check could go into reusecrcfromlastround(..). 
Otherwise thing look ok to me.

/Juha-Pekka

>   							/*