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 31F0CC369C7 for ; Thu, 17 Apr 2025 09:17:27 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id CCDB010EA70; Thu, 17 Apr 2025 09:17:26 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.b="PcgiPOps"; dkim-atps=neutral Received: from mgamail.intel.com (mgamail.intel.com [192.198.163.8]) by gabe.freedesktop.org (Postfix) with ESMTPS id 8196610E1B1 for ; Thu, 17 Apr 2025 09:17:24 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1744881445; x=1776417445; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=c7GqBcDNHz00UPA3b0MyYliiqSEdGJnw5vF4K8PLxJg=; b=PcgiPOpskLzR2+s7O3GkR8g0oNXKErqpupyL9unMaTXqIFQ2lO3WiSYg w3pqnEBE8iVACZblmgAsP1PjO0iICdD2ihZKGWHrhwBRfBs3SUQb/xM6x okzvgfXkLqTIMnA4YV68cL5pkz8TmYle7/9nGZ5cgI4PTczUbcydarEAM vtefd/uzSwSbNBlQNEU1T8gu4KkXolPa26vqVuK5G8RsAxIrenGBkH1qX r0gSE2/Ju/ZGjL/RFBZlOi0sXpgfMHco/KMFVRKEBHe6bGlTs4qzgqPCt jkcZBmI0E/7A7RbF7Yx8VqnXnkbq29kGTj65MazTjsS9mmOGqW0cpkHET Q==; X-CSE-ConnectionGUID: Qy7AA6f7S5CTwZT//f+LVw== X-CSE-MsgGUID: bU1IiCNnSU25UVs9j8Ny7g== X-IronPort-AV: E=McAfee;i="6700,10204,11405"; a="63996141" X-IronPort-AV: E=Sophos;i="6.15,218,1739865600"; d="scan'208";a="63996141" Received: from fmviesa001.fm.intel.com ([10.60.135.141]) by fmvoesa102.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 17 Apr 2025 02:17:25 -0700 X-CSE-ConnectionGUID: rP7gJ3poQymsW2piq894Aw== X-CSE-MsgGUID: xhBbd6TQSrWy4e7hONquIg== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.15,218,1739865600"; d="scan'208";a="161792644" Received: from mgolanimitul-x299-ud4-pro.iind.intel.com ([10.190.239.114]) by fmviesa001.fm.intel.com with ESMTP; 17 Apr 2025 02:17:23 -0700 From: Mitul Golani To: igt-dev@lists.freedesktop.org Cc: ankit.k.nautiyal@intel.com, ramanaidu.naladala@intel.com Subject: [PATCH v4 1/2] tests/kms_vrr: Bucketize refresh rate tolerance Date: Thu, 17 Apr 2025 14:45:12 +0530 Message-ID: <20250417091514.1794527-2-mitulkumar.ajitkumar.golani@intel.com> X-Mailer: git-send-email 2.48.1 In-Reply-To: <20250417091514.1794527-1-mitulkumar.ajitkumar.golani@intel.com> References: <20250417091514.1794527-1-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). --v4: Use pre-table for refresh rate checks (Ankit). Commit message changes (Ankit). Signed-off-by: Mitul Golani --- tests/kms_vrr.c | 82 +++++++++++++++++++++++++++++++++++++++++++++++-- 1 file changed, 79 insertions(+), 3 deletions(-) diff --git a/tests/kms_vrr.c b/tests/kms_vrr.c index c4bb30f6a..c9059c897 100644 --- a/tests/kms_vrr.c +++ b/tests/kms_vrr.c @@ -124,6 +124,24 @@ typedef struct vtest_ns { uint64_t max; } vtest_ns_t; +typedef struct { + int min_rate; /* Min refresh rate in Hz */ + int max_rate; /* Max refresh rate in Hz */ + int tolerance; /* Tolerance in Hz */ +} tolerance_config; + +/* + * Tolerance values are determined based on predefined ranges + * ≤ 120 Hz: ±1 + * 121-240 Hz: ±10 + * > 240 Hz: ±20 + */ +tolerance_config tolerance_table[] = { + {0, 120, 1}, + {121, 240, 10}, + {241, INT_MAX, 20} +}; + typedef struct data { igt_display_t display; int drm_fd; @@ -410,6 +428,57 @@ 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. + * 5. Tolerance Check: Determine if the flip completion time falls within + * the acceptable range. + * + * We set a tolerance value as the acceptable range of time within which a + * flip completion event should occur. 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. + */ + + for(int i = 0; i < ARRAY_SIZE(tolerance_table); i++) { + if (refresh_rate >= tolerance_table[i].min_rate && + refresh_rate <= tolerance_table[i].max_rate) { + *threshold_hi = + NSECS_PER_SEC / (((float)NSECS_PER_SEC / rates_ns) + tolerance_table[i].tolerance); + *threshold_lo = + NSECS_PER_SEC / (((float)NSECS_PER_SEC / rates_ns) - tolerance_table[i].tolerance); + + return; + } + } +} + /* * Flips at the given rate and measures against the expected value. * Returns the pass rate as a percentage from 0 - 100. @@ -439,9 +508,7 @@ 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 +564,15 @@ 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