From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: X-Spam-Checker-Version: SpamAssassin 3.4.0 (2014-02-07) on aws-us-west-2-korg-lkml-1.web.codeaurora.org Received: from gabe.freedesktop.org (gabe.freedesktop.org [131.252.210.177]) (using TLSv1.2 with cipher ECDHE-RSA-AES256-GCM-SHA384 (256/256 bits)) (No client certificate requested) by smtp.lore.kernel.org (Postfix) with ESMTPS id 873E2C3DA4A for ; Fri, 4 Apr 2025 04:39:39 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id 2710010E127; Fri, 4 Apr 2025 04:39:34 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.b="CrVxX1oW"; dkim-atps=neutral Received: from mgamail.intel.com (mgamail.intel.com [192.198.163.15]) by gabe.freedesktop.org (Postfix) with ESMTPS id 6DA9310E127 for ; Fri, 4 Apr 2025 04:39:31 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1743741571; x=1775277571; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=eXK6lXg5uhu0dsnHRZVPYvXCv2plKwm5mdkLBHxIhkg=; b=CrVxX1oWA5Bl8WGxndCdN7OJFC872DfrutbTUeKLIu4IwXSbhPlol22M SpgrmNjU9/pM51f/noAAR2O6EInshcLGx6ueC5xIpv74TxkNVDDCwvXQI EObe0wqxb14uCaFELp3Ow28qLzEK4kNwFeq5NWsZBo5gdSBhViInb4rS1 Cxv+glnl+NQwa86CUY8QxLVjLcG/lUNd1alLjhrSUOl9gqTBB2nqrieyc SeQHOiKRxj3lH5kbWmorRlaYx6BjE/ZfQtGn13oVTsqE0aMLLyH32yNgR oGfDIOuzMQm7wARffIKbBQF2JVFZqke9ASZCF56NorxFkB4THmn2DocfI g==; X-CSE-ConnectionGUID: 83VQAg7ORW6s3ocjwy8XMg== X-CSE-MsgGUID: IAP0tuLtTbeWyqcssNc4rw== X-IronPort-AV: E=McAfee;i="6700,10204,11393"; a="45294692" X-IronPort-AV: E=Sophos;i="6.15,187,1739865600"; d="scan'208";a="45294692" Received: from fmviesa009.fm.intel.com ([10.60.135.149]) by fmvoesa109.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 03 Apr 2025 21:39:29 -0700 X-CSE-ConnectionGUID: qujYqLxjQRyPjAcFs9n23g== X-CSE-MsgGUID: CCVqodItTlO+vEbm4/igIg== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.15,187,1739865600"; d="scan'208";a="128062179" Received: from mgolanimitul-x299-ud4-pro.iind.intel.com ([10.190.239.114]) by fmviesa009.fm.intel.com with ESMTP; 03 Apr 2025 21:39:27 -0700 From: Mitul Golani To: igt-dev@lists.freedesktop.org Cc: uma.shankar@intel.com, ramanaidu.naladala@intel.com, mitulkumar.ajitkumar.golani@intel.com, ankit.k.nautiyal@intel.com Subject: [PATCH v3 1/3] tests/kms_vrr: Bucketize refresh rate tolerance Date: Fri, 4 Apr 2025 10:07:06 +0530 Message-ID: <20250404043709.955109-2-mitulkumar.ajitkumar.golani@intel.com> X-Mailer: git-send-email 2.48.1 In-Reply-To: <20250404043709.955109-1-mitulkumar.ajitkumar.golani@intel.com> References: <20250404043709.955109-1-mitulkumar.ajitkumar.golani@intel.com> MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit X-BeenThere: igt-dev@lists.freedesktop.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Development mailing list for IGT GPU Tools List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: igt-dev-bounces@lists.freedesktop.org Sender: "igt-dev" Reduce false failures while preserving timing accuracy. Introduce a small tolerance buffer based on refresh rate which accounts for HW/SW latency without compromising validation on HRR panel. Although an imperical number but already IGT is living with that. This also ensures that asked refresh rate is not too off and always catch the real HW/software issues. --v2: - Refresh rate criteria changes. --v3: - Comment changes (Uma). - Add FIXME Signed-off-by: Mitul Golani --- tests/kms_vrr.c | 59 ++++++++++++++++++++++++++++++++++++++++++++++--- 1 file changed, 56 insertions(+), 3 deletions(-) diff --git a/tests/kms_vrr.c b/tests/kms_vrr.c index c4bb30f6a..ccf32f453 100644 --- a/tests/kms_vrr.c +++ b/tests/kms_vrr.c @@ -410,6 +410,52 @@ do_flip(data_t *data, igt_fb_t *fb) igt_reset_timeout(); } +static void +calculate_tolerance(uint64_t *threshold_hi, uint64_t *threshold_lo, uint64_t rates_ns) +{ + uint32_t refresh_rate = NSECS_PER_SEC / rates_ns; + + if (refresh_rate < 0) + return; + + /* + * Current IGT implementation follows this sequence: + * 1. Perform a page flip (`do_flip`). + * 2. Wait for the flip completion event. + * 3. Compare the timestamp of the flip completion event with the previous frame’s + * completion timestamp. + * 4. Adjust CPU cycle burning based on the relative frame time. + * + * If a flip completes too early or too late, it is marked as out of tolerance. + * As a result, additional CPU cycles are burned to match the `target_ns`. + * Even if the next frame is on time, the total frame time now includes: + * Burned CPU cycle time (from the previous frame) + Flip completion event time. + * This leads to miscalculation, causing **false out-of-range detections**. + * The impact is more significant on High Refresh Rate (HRR) panels, where: + * The allowed tolerance window is smaller and more correction time is required. + * i.e. for 210hz (4.762ms), allowed range is 209hz(4.784ms) to 211hz(4.739ms). + * This comes just 23 microsecond tolerance, which is much lesser + * for accounting HW/SW latency, CPU burn cycle latency and correction logic + * applied in igt for validation. + * + * To address this implement a Bucketing Strategy: + * Provide a small tolerance buffer to allow IGT tests to account for correction. + * Based on range of asked refresh rate. This prevents excessive failures due to minor + * timing adjustments. + */ + + if (refresh_rate <= 120) { + *threshold_hi = NSECS_PER_SEC / (((float)NSECS_PER_SEC / rates_ns) + 1); + *threshold_lo = NSECS_PER_SEC / (((float)NSECS_PER_SEC / rates_ns) - 1); + } else if (refresh_rate >= 120 && refresh_rate <= 240) { + *threshold_hi = NSECS_PER_SEC / (((float)NSECS_PER_SEC / rates_ns) + 5); + *threshold_lo = NSECS_PER_SEC / (((float)NSECS_PER_SEC / rates_ns) - 5); + } else { + *threshold_hi = NSECS_PER_SEC / (((float)NSECS_PER_SEC / rates_ns) + 10); + *threshold_lo = NSECS_PER_SEC / (((float)NSECS_PER_SEC / rates_ns) - 10); + } +} + /* * Flips at the given rate and measures against the expected value. * Returns the pass rate as a percentage from 0 - 100. @@ -439,9 +485,7 @@ flip_and_measure(data_t *data, igt_output_t *output, enum pipe pipe, else exp_rate_ns = vtest_ns.max; - /* Allow ~1 Hz deviation for different reasons causing delay. */ - threshold_hi[i] = NSECS_PER_SEC / (((float)NSECS_PER_SEC / exp_rate_ns) + 1); - threshold_lo[i] = NSECS_PER_SEC / (((float)NSECS_PER_SEC / exp_rate_ns) - 1); + calculate_tolerance(&threshold_hi[i], &threshold_lo[i], exp_rate_ns); igt_info("Requested rate[%d]: %"PRIu64" ns, Expected rate between: %"PRIu64" ns to %"PRIu64" ns\n", i, rates_ns[i], threshold_hi[i], threshold_lo[i]); @@ -497,6 +541,15 @@ flip_and_measure(data_t *data, igt_output_t *output, enum pipe pipe, wait_ns -= diff_ns; target_ns = event_ns + wait_ns; + /* + * FIXME: This logic makes next immediate frame time calculation + * in inconsistent state, even if next flip comes on correct time, + * it will be marked as fail due to time difference from previous + * flip. Needs to reset at every cycle for correct measurement. + * Once this is corrected, igt can ask for more stricter pass + * criteria. + */ + while (get_time_ns() < target_ns - 10); } -- 2.48.1