Hi, I do not think the added time is bad, but wasn't testdisplay considered for eviction as it was not testing anything else but the fact that the kernel does not blow up on modeset? Could it be improved to check that we generate a valid image? This might require pipe writeback though, so we shouldn't block on that. Also, how slow is it when we have 6 pipes and 6 DP displays connected? Wouldn't it be taking a lot of time? Martin On 2020-08-13 10:24, Sharma, Swati2 wrote: > @martin peres any thoughts? > > On 10-Aug-20 1:59 PM, Sharma, Swati2 wrote: >> BAT results >> https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_4858/index.html?testfilter=testdisplay >> >> >> Execution time of few CI m/c. >> |fi-tgl-dsi|fi-tgl-u2|fi-tgl-y|fi-ehl-1|fi-icl-u2|fi-icl-y|fi-cml-u2| >> |  0.352s  |  0.064s | 0.548s |    4.686s |  5.311s | 0.101s | 1.464s  | >> >> Execution time depends upon which panel is connected if its edp max it >> supports 2 resolutions => less time. For HDMI; having multiple >> resolutions its taking more time. >> >> On 06-Aug-20 4:38 PM, Swati Sharma wrote: >>> testdisplay should be added to BAT. testdisplay is considered >>> to be the most basic test to test display functionality. >>> With this test we will be able to test modesetting. >>> >>> Signed-off-by: Swati Sharma >>> Cc: Petri Latvala >>> Cc: Karthik B S >>> --- >>>   tests/intel-ci/fast-feedback.testlist | 1 + >>>   1 file changed, 1 insertion(+) >>> >>> diff --git a/tests/intel-ci/fast-feedback.testlist >>> b/tests/intel-ci/fast-feedback.testlist >>> index b796b264..b245b7a7 100644 >>> --- a/tests/intel-ci/fast-feedback.testlist >>> +++ b/tests/intel-ci/fast-feedback.testlist >>> @@ -149,6 +149,7 @@ igt@vgem_basic@dmabuf-mmap >>>   igt@vgem_basic@mmap >>>   igt@vgem_basic@second-client >>>   igt@vgem_basic@sysfs >>> +igt@testdisplay >>>   # All tests that do module unloading and reloading are executed last. >>>   # They will sometimes reveal issues of earlier tests leaving the >>> >> >