From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from NAM12-BN8-obe.outbound.protection.outlook.com (mail-bn8nam12on2054.outbound.protection.outlook.com [40.107.237.54]) by gabe.freedesktop.org (Postfix) with ESMTPS id A58A210E2AB for ; Mon, 10 Jul 2023 17:17:29 +0000 (UTC) Message-ID: <5aca6f5f-a861-3d17-18b6-c680b03514eb@amd.com> Date: Mon, 10 Jul 2023 11:17:24 -0600 Content-Language: en-US To: Tom Chung , igt-dev@lists.freedesktop.org References: <20230710075749.2377896-1-chiahsuan.chung@amd.com> <20230710075749.2377896-5-chiahsuan.chung@amd.com> From: Alex Hung In-Reply-To: <20230710075749.2377896-5-chiahsuan.chung@amd.com> Content-Type: text/plain; charset=UTF-8; format=flowed Content-Transfer-Encoding: 7bit MIME-Version: 1.0 Subject: Re: [igt-dev] [i-g-t, v4 4/8] tests/amdgpu/amd_freesync_video_mode: Add some code from kms_vrr to resolve tearing issue List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Cc: christian.koenig@amd.com Errors-To: igt-dev-bounces@lists.freedesktop.org Sender: "igt-dev" List-ID: Reviewed-by: Alex Hung On 2023-07-10 01:57, Tom Chung wrote: > [Why] > There is a tearing video issue during the stage-2 animation test. > > [How] > 1. Refer to the kms_vrr IGT code, it needs some delay before the the next flip and > add some warmup time before do the flip and measure to make the test more accurate. > 2. Change the target refresh rate percentage to 75% to align with the kms_vrr test. > > Signed-off-by: Tom Chung > --- > tests/amdgpu/amd_freesync_video_mode.c | 22 ++++++++++++++++++---- > 1 file changed, 18 insertions(+), 4 deletions(-) > > diff --git a/tests/amdgpu/amd_freesync_video_mode.c b/tests/amdgpu/amd_freesync_video_mode.c > index 29ba1f65c..0927f4307 100644 > --- a/tests/amdgpu/amd_freesync_video_mode.c > +++ b/tests/amdgpu/amd_freesync_video_mode.c > @@ -26,7 +26,7 @@ > #include > > #define NSECS_PER_SEC (1000000000ull) > -#define TEST_DURATION_NS (10 * NSECS_PER_SEC) > +#define TEST_DURATION_NS (8 * NSECS_PER_SEC) > > #define BYTES_PER_PIXEL 4 > #define MK_COLOR(r, g, b) ((0 << 24) | (r << 16) | (g << 8) | b) > @@ -638,7 +638,7 @@ flip_and_measure( > igt_info("interval_ns=%lu\n", interval_ns); > > for (;;) { > - uint64_t event_ns; > + uint64_t event_ns, wait_ns; > int64_t diff_ns; > > data->front = !data->front; > @@ -673,6 +673,19 @@ flip_and_measure( > > if (event_ns - start_ns > duration_ns) > break; > + > + /* > + * Burn CPU until next timestamp, sleeping isn't accurate enough. > + * The target timestamp is based on the delta b/w event timestamps > + * and whatever the time left to reach the expected refresh rate. > + */ > + diff_ns = event_ns - target_ns; > + wait_ns = ((diff_ns + interval_ns - 1) / interval_ns) * interval_ns; > + wait_ns -= diff_ns; > + target_ns = event_ns + wait_ns; > + > + while (get_time_ns() < target_ns - 10) > + ; > } > > igt_info("Completed %u flips, %u were in threshold for (%llu Hz) %"PRIu64"ns.\n", > @@ -808,9 +821,10 @@ mode_transition(data_t *data, enum pipe pipe, igt_output_t *output, uint32_t sce > igt_info("stage-2: simple animation as video playback\n"); > prepare_test(data, output, pipe, mode_playback); > interval = nsec_per_frame(mode_playback->vrefresh); > + /* Do a short run with VRR before measure to make sure we measure in a stable state */ > + result = flip_and_measure(data, output, pipe, interval, 2 * NSECS_PER_SEC, ANIM_TYPE_CIRCLE_WAVE); > result = flip_and_measure(data, output, pipe, interval, TEST_DURATION_NS, ANIM_TYPE_CIRCLE_WAVE); > - igt_assert_f(result > 90, "Target refresh rate not meet(result=%d%%\n", result); > - > + igt_assert_f(result > 75, "Target refresh rate not meet 75%% (result=%d%%\n", result); > finish_test(data, pipe, output); > } >