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 1F0E0C021B8 for ; Tue, 25 Feb 2025 18:11:04 +0000 (UTC) Received: from gabe.freedesktop.org (localhost [127.0.0.1]) by gabe.freedesktop.org (Postfix) with ESMTP id D320B10E725; Tue, 25 Feb 2025 18:11:03 +0000 (UTC) Authentication-Results: gabe.freedesktop.org; dkim=pass (2048-bit key; unprotected) header.d=intel.com header.i=@intel.com header.b="F2sH74Hn"; dkim-atps=neutral Received: from mgamail.intel.com (mgamail.intel.com [192.198.163.13]) by gabe.freedesktop.org (Postfix) with ESMTPS id D93F410E725 for ; Tue, 25 Feb 2025 18:11:01 +0000 (UTC) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1740507062; x=1772043062; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=KcAapFkjJy+fGcdoUL1EI3UfCQDZGcxvtSpdvKrx2IY=; b=F2sH74HnzS7C1mjDsWTx49r+zodgptai+FEgfAESwJFFJst2ZDsuzILG 4HSlMOQQx336ZLPgeISsCXVsLB0JFqFs4PF3QhJDr1jpmjKEEGf2UHZiM gcYhHBeyo4pSS8mXzcGleDQa6AxCNzEYylSA0+ifqus6jeCaw7tQKgRVP mYND41g731cbM9EHP85uiq0P5G+uBpkdIwP0XUOIc0o+PPgbO234Kw5Pe jYbLu/izy/kDaNzgnWaBauXnew/9TNzPOddcrA3V7VcRWNTwAK5e2CyRw TQmsDOKc0tPlnXeg4D2/afCvS4Xs8t+W7hZpru3U7ghy/KBS0Z2Z9M4/W g==; X-CSE-ConnectionGUID: rh51AC5GQASrp8JtCH6nCg== X-CSE-MsgGUID: X8AFew3zR1KV575uk0ua8Q== X-IronPort-AV: E=McAfee;i="6700,10204,11356"; a="44151834" X-IronPort-AV: E=Sophos;i="6.13,314,1732608000"; d="scan'208";a="44151834" Received: from fmviesa007.fm.intel.com ([10.60.135.147]) by fmvoesa107.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 25 Feb 2025 10:11:02 -0800 X-CSE-ConnectionGUID: /r+NpBeGQcOH/d7z8jzuGg== X-CSE-MsgGUID: LqNi3NdBQam+Md45z3gpGQ== X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="6.13,314,1732608000"; d="scan'208";a="116477483" Received: from unknown (HELO anirban-Z690I-A-ULTRA-PLUS.iind.intel.com) ([10.190.216.83]) by fmviesa007-auth.fm.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 25 Feb 2025 10:10:59 -0800 From: Sk Anirban To: igt-dev@lists.freedesktop.org Cc: anshuman.gupta@intel.com, badal.nilawar@intel.com, riana.tauro@intel.com, karthik.poosa@intel.com, chris.p.wilson@intel.com, Sk Anirban Subject: [PATCH 1/2] tests/i915/pm_rc6_residency: Replace waitboost with igt spinner and use sysfs to achieve peak frequency Date: Tue, 25 Feb 2025 23:39:29 +0530 Message-Id: <20250225180930.1966111-2-sk.anirban@intel.com> X-Mailer: git-send-email 2.34.1 In-Reply-To: <20250225180930.1966111-1-sk.anirban@intel.com> References: <20250225180930.1966111-1-sk.anirban@intel.com> MIME-Version: 1.0 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" Leverage IGT spinner to trigger frequency scaling and set RPS minimum frequency to RP0 for achieving peak frequency. This method replaces the existing waitboost mechanism used to reach maximum frequency. Signed-off-by: Sk Anirban --- tests/intel/i915_pm_rc6_residency.c | 111 ++++++++++++++++++---------- 1 file changed, 72 insertions(+), 39 deletions(-) diff --git a/tests/intel/i915_pm_rc6_residency.c b/tests/intel/i915_pm_rc6_residency.c index c9128481d..eaeccc480 100644 --- a/tests/intel/i915_pm_rc6_residency.c +++ b/tests/intel/i915_pm_rc6_residency.c @@ -264,17 +264,6 @@ static bool __pmu_wait_for_rc6(int fd) return false; } -static uint32_t batch_create(int fd) -{ - const uint32_t bbe = MI_BATCH_BUFFER_END; - uint32_t handle; - - handle = gem_create(fd, 4096); - gem_write(fd, handle, 0, &bbe, sizeof(bbe)); - - return handle; -} - static int open_pmu(int i915, uint64_t config) { int fd; @@ -286,45 +275,72 @@ static int open_pmu(int i915, uint64_t config) return fd; } -#define WAITBOOST 0x1 +#define FREQUENT_BOOST 0x1 #define ONCE 0x2 static void sighandler(int sig) { } -static void bg_load(int i915, uint32_t ctx_id, uint64_t engine_flags, unsigned int flags, unsigned long *ctl) +static uint32_t get_freq(int dirfd, uint8_t id) +{ + uint32_t val; + + igt_assert(igt_sysfs_rps_scanf(dirfd, id, "%u", &val) == 1); + + return val; +} + +static int set_freq(int dirfd, uint8_t id, uint32_t val) +{ + return igt_sysfs_rps_printf(dirfd, id, "%u", val); +} + +uint32_t *stash_min; + +static void bg_load(int i915, const intel_ctx_t *ctx, int dirfd, uint64_t engine_flags, + unsigned int flags, unsigned long *ctl, unsigned int gt) { const bool has_execlists = intel_gen(intel_get_drm_devid(i915)) >= 8; - struct drm_i915_gem_exec_object2 obj = { - .handle = batch_create(i915), - }; - struct drm_i915_gem_execbuffer2 execbuf = { - .buffers_ptr = to_user_pointer(&obj), - .buffer_count = 1, - .flags = engine_flags, - .rsvd1 = ctx_id, - }; struct sigaction act = { .sa_handler = sighandler }; + int64_t timeout = 1; + uint64_t ahnd; + int rp0; + ahnd = get_reloc_ahnd(i915, ctx->id); + rp0 = get_freq(dirfd, RPS_RP0_FREQ_MHZ); sigaction(SIGINT, &act, NULL); do { uint64_t submit, wait, elapsed; struct timespec tv = {}; + igt_spin_t *spin; igt_nsec_elapsed(&tv); - - gem_execbuf(i915, &execbuf); + spin = igt_spin_new(i915, + .ahnd = ahnd, + .ctx = ctx, + .engine = engine_flags); submit = igt_nsec_elapsed(&tv); - if (flags & WAITBOOST) { - gem_sync(i915, obj.handle); + if (flags & FREQUENT_BOOST) { + /* Set MIN freq to RP0 to achieve the peak freq */ + igt_assert_lt(0, set_freq(dirfd, RPS_MIN_FREQ_MHZ, rp0)); + igt_assert(gem_bo_busy(i915, spin->handle)); + gem_wait(i915, spin->handle, &timeout); + + /* Restore the MIN freq back to default */ + igt_assert_lt(0, set_freq(dirfd, RPS_MIN_FREQ_MHZ, stash_min[gt])); + igt_spin_end(spin); + igt_spin_free(i915, spin); + gem_quiescent_gpu(i915); if (flags & ONCE) - flags &= ~WAITBOOST; + flags &= ~FREQUENT_BOOST; } else { - while (gem_bo_busy(i915, obj.handle)) - usleep(0); + igt_assert(gem_bo_busy(i915, spin->handle)); + igt_spin_end(spin); + igt_spin_free(i915, spin); + gem_quiescent_gpu(i915); } wait = igt_nsec_elapsed(&tv); @@ -350,6 +366,7 @@ static void bg_load(int i915, uint32_t ctx_id, uint64_t engine_flags, unsigned i /* aim for ~1% busy */ usleep(min_t(elapsed, elapsed / 10, 50 * 1000)); } while (!READ_ONCE(*ctl)); + put_ahnd(ahnd); } static void kill_children(int sig) @@ -361,25 +378,29 @@ static void kill_children(int sig) signal(sig, old); } -static void rc6_idle(int i915, uint32_t ctx_id, uint64_t flags, unsigned int gt) +static void rc6_idle(int i915, const intel_ctx_t *ctx, uint64_t flags, unsigned int gt) { - const int64_t duration_ns = SLEEP_DURATION * (int64_t)NSEC_PER_SEC; + const int64_t duration_ns = 2 * SLEEP_DURATION * (int64_t)NSEC_PER_SEC; const int tolerance = 20; /* Some RC6 is better than none! */ const unsigned int gen = intel_gen(intel_get_drm_devid(i915)); + int num_gts = igt_sysfs_get_num_gt(i915); struct { const char *name; unsigned int flags; double power; } phases[] = { + { "once", FREQUENT_BOOST | ONCE }, { "normal", 0 }, - { "boost", WAITBOOST }, - { "once", WAITBOOST | ONCE }, + { "boost", FREQUENT_BOOST } }; + int dirfd = igt_sysfs_gt_open(i915, gt); struct power_sample sample[2]; unsigned long slept, cycles; unsigned long *done; uint64_t rc6, ts[2]; struct igt_power gpu; + int dirfd1; + int gt_num; int fd; fd = open_pmu(i915, __I915_PMU_RC6_RESIDENCY(gt)); @@ -409,10 +430,16 @@ static void rc6_idle(int i915, uint32_t ctx_id, uint64_t flags, unsigned int gt) done = mmap(0, 4096, PROT_WRITE, MAP_SHARED | MAP_ANON, -1, 0); + stash_min = (uint32_t *)malloc(sizeof(uint32_t) * num_gts); + for_each_sysfs_gt_dirfd(i915, dirfd1, gt_num) { + stash_min[gt_num] = get_freq(dirfd1, RPS_MIN_FREQ_MHZ); + igt_pm_ignore_slpc_efficient_freq(i915, dirfd1, true); + } + for (int p = 0; p < ARRAY_SIZE(phases); p++) { memset(done, 0, 2 * sizeof(*done)); igt_fork(child, 1) /* Setup up a very light load */ - bg_load(i915, ctx_id, flags, phases[p].flags, done); + bg_load(i915, ctx, dirfd, flags, phases[p].flags, done, gt); igt_power_get_energy(&gpu, &sample[0]); cycles = -READ_ONCE(done[1]); @@ -454,12 +481,12 @@ static void rc6_idle(int i915, uint32_t ctx_id, uint64_t flags, unsigned int gt) igt_power_close(&gpu); - if (phases[1].power - phases[0].power > 10) { - igt_assert_f(2 * phases[2].power - phases[0].power <= phases[1].power, + if (phases[2].power - phases[1].power > 20 && !gt) { + igt_assert_f(2 * phases[0].power - phases[1].power <= phases[2].power, "Exceeded energy expectations for single busy wait load\n" "Used %.1fmW, min %.1fmW, max %.1fmW, expected less than %.1fmW\n", - phases[2].power, phases[0].power, phases[1].power, - phases[0].power + (phases[1].power - phases[0].power) / 2); + phases[0].power, phases[1].power, phases[2].power, + phases[1].power + (phases[2].power - phases[1].power) / 2); } } @@ -577,6 +604,7 @@ igt_main /* Use drm_open_driver to verify device existence */ igt_fixture { i915 = drm_open_driver(DRIVER_INTEL); + intel_allocator_multiprocess_start(); } igt_subtest_with_dynamic("rc6-idle") { @@ -590,13 +618,18 @@ igt_main for_each_ctx_engine(i915, ctx, e) { if (e->instance == 0) { igt_dynamic_f("gt%u-%s", gt, e->name) - rc6_idle(i915, ctx->id, e->flags, gt); + rc6_idle(i915, ctx, e->flags, gt); } } intel_ctx_destroy(i915, ctx); } } + igt_fixture + { + intel_allocator_multiprocess_stop(); + } + igt_subtest_with_dynamic("rc6-fence") { igt_require_gem(i915); gem_quiescent_gpu(i915); -- 2.34.1