[PATCH i-g-t v5 8/8] benchmarks/kms_fb_stress: Create a benchmark for formats

[Louis Chauvet]

Le 22/04/24 - 19:18, Arthur Grillo a écrit :
> Create a benchmark that tests the performance of each supported format
> for the primary plane.
> 
> For VKMS, it tests the performance of the conversion of formats to/from
> the ARGB1616161 baseline format.
> 
> Signed-off-by: Arthur Grillo <arthurgrillo at riseup.net>
> ---
>  benchmarks/kms_fb_stress.c | 8 ++++++++
>  1 file changed, 8 insertions(+)
> 
> diff --git a/benchmarks/kms_fb_stress.c b/benchmarks/kms_fb_stress.c
> index d2d6b3e65eae..5535339cd287 100644
> --- a/benchmarks/kms_fb_stress.c
> +++ b/benchmarks/kms_fb_stress.c
> @@ -295,6 +295,14 @@ igt_main
>  
>  	}
>  
> +	igt_subtest("formats") {
> +		data.use_overlay = false;
> +		for (size_t i = 0; i < data.kms.primary.base->format_mod_count; i++) {
> +			data.kms.primary.format = data.kms.primary.base->formats[i];
> +			data.kms.writeback.format = data.kms.primary.base->formats[i];
> +			stress_driver(&data);
> +		}
> +	}

I just used this test to compute the performance numbers. I have a comment 
on this part:

I am not sure about the writeback format here. The writeback connector may 
only support a few formats, not all the formats supported by the primary 
plane. I think you can split this test into two: one iterating over plane 
formats, and one iterating over writeback formats, so you can compare the 
performance of writeback and plane.

>  	igt_fixture {
>  		igt_display_fini(&data.display);
> 
> -- 
> 2.44.0
> 

-- 
Louis Chauvet, Bootlin
Embedded Linux and Kernel engineering
https://bootlin.com