From patchwork Fri Oct 27 22:57:26 2023 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: "Chautru, Nicolas" X-Patchwork-Id: 133525 X-Patchwork-Delegate: maxime.coquelin@redhat.com Return-Path: X-Original-To: patchwork@inbox.dpdk.org Delivered-To: patchwork@inbox.dpdk.org Received: from mails.dpdk.org (mails.dpdk.org [217.70.189.124]) by inbox.dpdk.org (Postfix) with ESMTP id 9E54D4321A; Sat, 28 Oct 2023 01:05:12 +0200 (CEST) Received: from mails.dpdk.org (localhost [127.0.0.1]) by mails.dpdk.org (Postfix) with ESMTP id 4BBCB42E54; Sat, 28 Oct 2023 01:04:24 +0200 (CEST) Received: from mgamail.intel.com (mgamail.intel.com [198.175.65.9]) by mails.dpdk.org (Postfix) with ESMTP id CA3A541133; Sat, 28 Oct 2023 01:04:13 +0200 (CEST) DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple; d=intel.com; i=@intel.com; q=dns/txt; s=Intel; t=1698447855; x=1729983855; h=from:to:cc:subject:date:message-id:in-reply-to: references:mime-version:content-transfer-encoding; bh=ZZWvqrBJDFLM0ZiRR9Mp9qrTlIsP+IkNJffbtdxjDao=; b=izHCKmM1ZO4ToRoTR7FaFieYX+lGOw0H91ho0GoVugZ59QGlTb2hnIA+ Db/DGzjTQAOJnW/J6Phzhgzf3mKmUB0EbBnnwqdczRlFEdEjuKJQtwqFn WrljwD0q7gxoCS0WdP7qrhGdAOzsQmqg2gAubMMW9GiT+b6sjddhMlSa5 y92dhTkzrNNSKPtbPC6QEdND0sxDnPaUFMWDR6kMIXqlVJcBKlBp+Kijj Rk6nDqFZTbt9HFk2ftyfmEo5xuK3PAKhQuBvIVs/1uwAWKchgj/rf4oDT Tt3weryOTzANg/GZqPCgZPsuejCtiyx0LF/N2sBl61FvE08Mx3QcPH02L Q==; X-IronPort-AV: E=McAfee;i="6600,9927,10876"; a="6494092" X-IronPort-AV: E=Sophos;i="6.03,257,1694761200"; d="scan'208";a="6494092" Received: from fmsmga008.fm.intel.com ([10.253.24.58]) by orvoesa101.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 27 Oct 2023 16:04:13 -0700 X-ExtLoop1: 1 X-IronPort-AV: E=McAfee;i="6600,9927,10876"; a="825496978" X-IronPort-AV: E=Sophos;i="6.03,257,1694761200"; d="scan'208";a="825496978" Received: from spr-npg-bds1-eec2.sn.intel.com (HELO spr-npg-bds1-eec2..) ([10.233.181.123]) by fmsmga008.fm.intel.com with ESMTP; 27 Oct 2023 16:04:11 -0700 From: Nicolas Chautru To: dev@dpdk.org, maxime.coquelin@redhat.com Cc: hemant.agrawal@nxp.com, david.marchand@redhat.com, hernan.vargas@intel.com, stable@dpdk.org Subject: [PATCH v2 08/10] test/bbdev: add MLD support Date: Fri, 27 Oct 2023 22:57:26 +0000 Message-Id: <20231027225728.563914-9-nicolas.chautru@intel.com> X-Mailer: git-send-email 2.34.1 In-Reply-To: <20231027225728.563914-1-nicolas.chautru@intel.com> References: <20231027225728.563914-1-nicolas.chautru@intel.com> MIME-Version: 1.0 X-BeenThere: dev@dpdk.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: DPDK patches and discussions List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Errors-To: dev-bounces@dpdk.org From: Hernan Vargas Adding test-bbdev support for the MLD-TS processing specific to the VRB2 variant. Signed-off-by: Hernan Vargas Reviewed-by: Maxime Coquelin --- app/test-bbdev/test_bbdev_perf.c | 519 +++++++++++++++++++++++++++++ app/test-bbdev/test_bbdev_vector.c | 132 ++++++++ app/test-bbdev/test_bbdev_vector.h | 1 + 3 files changed, 652 insertions(+) diff --git a/app/test-bbdev/test_bbdev_perf.c b/app/test-bbdev/test_bbdev_perf.c index 4f8e226e58..8a349fdb03 100644 --- a/app/test-bbdev/test_bbdev_perf.c +++ b/app/test-bbdev/test_bbdev_perf.c @@ -139,6 +139,7 @@ struct test_op_params { struct rte_bbdev_dec_op *ref_dec_op; struct rte_bbdev_enc_op *ref_enc_op; struct rte_bbdev_fft_op *ref_fft_op; + struct rte_bbdev_mldts_op *ref_mldts_op; uint16_t burst_sz; uint16_t num_to_process; uint16_t num_lcores; @@ -165,6 +166,7 @@ struct thread_params { struct rte_bbdev_dec_op *dec_ops[MAX_BURST]; struct rte_bbdev_enc_op *enc_ops[MAX_BURST]; struct rte_bbdev_fft_op *fft_ops[MAX_BURST]; + struct rte_bbdev_mldts_op *mldts_ops[MAX_BURST]; }; /* Stores time statistics */ @@ -472,6 +474,18 @@ check_dev_cap(const struct rte_bbdev_info *dev_info) return TEST_FAILED; } return TEST_SUCCESS; + } else if (op_cap->type == RTE_BBDEV_OP_MLDTS) { + const struct rte_bbdev_op_cap_mld *cap = &op_cap->cap.mld; + if (!flags_match(test_vector.mldts.op_flags, cap->capability_flags)) { + printf("Flag Mismatch\n"); + return TEST_FAILED; + } + if (nb_inputs > cap->num_buffers_src) { + printf("Too many inputs defined: %u, max: %u\n", + nb_inputs, cap->num_buffers_src); + return TEST_FAILED; + } + return TEST_SUCCESS; } } @@ -822,6 +836,9 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info, conf.arb_fft[i].gbr_threshold1 = VRB_QOS_GBR; conf.arb_fft[i].gbr_threshold1 = VRB_QOS_GBR; conf.arb_fft[i].round_robin_weight = VRB_QMGR_RR; + conf.arb_mld[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_mld[i].gbr_threshold1 = VRB_QOS_GBR; + conf.arb_mld[i].round_robin_weight = VRB_QMGR_RR; } conf.input_pos_llr_1_bit = true; @@ -847,6 +864,10 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info, conf.q_fft.num_qgroups = VRB_QMGR_NUM_QGS; conf.q_fft.first_qgroup_index = VRB_QMGR_INVALID_IDX; conf.q_fft.num_aqs_per_groups = VRB_QMGR_NUM_AQS; + conf.q_mld.num_qgroups = VRB_QMGR_NUM_QGS; + conf.q_mld.first_qgroup_index = VRB_QMGR_INVALID_IDX; + conf.q_mld.num_aqs_per_groups = VRB_QMGR_NUM_AQS; + conf.q_mld.aq_depth_log2 = VRB_QMGR_AQ_DEPTH; /* setup PF with configuration information */ ret = rte_acc_configure(info->dev_name, &conf); @@ -1979,6 +2000,31 @@ copy_reference_fft_op(struct rte_bbdev_fft_op **ops, unsigned int n, } } +static void +copy_reference_mldts_op(struct rte_bbdev_mldts_op **ops, unsigned int n, + unsigned int start_idx, + struct rte_bbdev_op_data *q_inputs, + struct rte_bbdev_op_data *r_inputs, + struct rte_bbdev_op_data *outputs, + struct rte_bbdev_mldts_op *ref_op) +{ + unsigned int i, j; + struct rte_bbdev_op_mldts *mldts = &ref_op->mldts; + for (i = 0; i < n; i++) { + ops[i]->mldts.c_rep = mldts->c_rep; + ops[i]->mldts.num_layers = mldts->num_layers; + ops[i]->mldts.num_rbs = mldts->num_rbs; + ops[i]->mldts.op_flags = mldts->op_flags; + for (j = 0; j < RTE_BBDEV_MAX_MLD_LAYERS; j++) + ops[i]->mldts.q_m[j] = mldts->q_m[j]; + ops[i]->mldts.r_rep = mldts->r_rep; + ops[i]->mldts.c_rep = mldts->c_rep; + ops[i]->mldts.r_input = r_inputs[start_idx + i]; + ops[i]->mldts.qhy_input = q_inputs[start_idx + i]; + ops[i]->mldts.output = outputs[start_idx + i]; + } +} + static int check_dec_status_and_ordering(struct rte_bbdev_dec_op *op, unsigned int order_idx, const int expected_status) @@ -2039,6 +2085,21 @@ check_fft_status_and_ordering(struct rte_bbdev_fft_op *op, return TEST_SUCCESS; } +static int +check_mldts_status_and_ordering(struct rte_bbdev_mldts_op *op, + unsigned int order_idx, const int expected_status) +{ + TEST_ASSERT(op->status == expected_status, + "op_status (%d) != expected_status (%d)", + op->status, expected_status); + + TEST_ASSERT((void *)(uintptr_t)order_idx == op->opaque_data, + "Ordering error, expected %p, got %p", + (void *)(uintptr_t)order_idx, op->opaque_data); + + return TEST_SUCCESS; +} + static inline int validate_op_chain(struct rte_bbdev_op_data *op, struct op_data_entries *orig_op) @@ -2554,6 +2615,57 @@ validate_op_fft_chain(struct rte_bbdev_op_data *op, struct op_data_entries *orig return TEST_SUCCESS; } +static inline int +validate_op_mldts_chain(struct rte_bbdev_op_data *op, + struct op_data_entries *orig_op) +{ + uint8_t i; + struct rte_mbuf *m = op->data; + uint8_t nb_dst_segments = orig_op->nb_segments; + /*the result is not bit exact*/ + int16_t thres_hold = 3; + int16_t delt, abs_delt; + uint32_t j, data_len_iq; + uint32_t error_num; + int8_t *ref_out; + int8_t *op_out; + + TEST_ASSERT(nb_dst_segments == m->nb_segs, + "Number of segments differ in original (%u) and filled (%u) op mldts", + nb_dst_segments, m->nb_segs); + + /* Due to size limition of mbuf, MLDTS doesn't use real mbuf. */ + for (i = 0; i < nb_dst_segments; ++i) { + uint16_t offset = (i == 0) ? op->offset : 0; + uint32_t data_len = op->length; + + TEST_ASSERT(orig_op->segments[i].length == data_len, + "Length of segment differ in original (%u) and filled (%u) op mldts", + orig_op->segments[i].length, data_len); + data_len_iq = data_len; + ref_out = (int8_t *)(orig_op->segments[i].addr); + op_out = rte_pktmbuf_mtod_offset(m, int8_t *, offset), + error_num = 0; + for (j = 0; j < data_len_iq; j++) { + + delt = ref_out[j] - op_out[j]; + abs_delt = delt > 0 ? delt : -delt; + error_num += (abs_delt > thres_hold ? 1 : 0); + if (error_num > 0) + printf("MLD Error %d: Exp %x %d Actual %x %d Diff %d\n", + j, ref_out[j], ref_out[j], op_out[j], op_out[j], + delt); + } + TEST_ASSERT(error_num == 0, + "MLDTS Output are not matched total (%u) errors (%u)", + data_len_iq, error_num); + + m = m->next; + } + + return TEST_SUCCESS; +} + static int validate_fft_op(struct rte_bbdev_fft_op **ops, const uint16_t n, struct rte_bbdev_fft_op *ref_op) @@ -2578,6 +2690,28 @@ validate_fft_op(struct rte_bbdev_fft_op **ops, const uint16_t n, return TEST_SUCCESS; } +static int +validate_mldts_op(struct rte_bbdev_mldts_op **ops, const uint16_t n, + struct rte_bbdev_mldts_op *ref_op) +{ + unsigned int i; + int ret; + struct op_data_entries *mldts_data_orig = + &test_vector.entries[DATA_HARD_OUTPUT]; + for (i = 0; i < n; ++i) { + ret = check_mldts_status_and_ordering(ops[i], i, ref_op->status); + TEST_ASSERT_SUCCESS(ret, + "Checking status and ordering for MLDTS failed"); + TEST_ASSERT_SUCCESS(validate_op_mldts_chain( + &ops[i]->mldts.output, + mldts_data_orig), + "MLDTS Output buffers (op=%u) are not matched", + i); + } + + return TEST_SUCCESS; +} + static void create_reference_dec_op(struct rte_bbdev_dec_op *op) { @@ -2622,6 +2756,20 @@ create_reference_fft_op(struct rte_bbdev_fft_op *op) op->fft.base_input.length += entry->segments[i].length; } +static void +create_reference_mldts_op(struct rte_bbdev_mldts_op *op) +{ + unsigned int i; + struct op_data_entries *entry; + op->mldts = test_vector.mldts; + entry = &test_vector.entries[DATA_INPUT]; + for (i = 0; i < entry->nb_segments; ++i) + op->mldts.qhy_input.length += entry->segments[i].length; + entry = &test_vector.entries[DATA_HARQ_INPUT]; + for (i = 0; i < entry->nb_segments; ++i) + op->mldts.r_input.length += entry->segments[i].length; +} + static void create_reference_enc_op(struct rte_bbdev_enc_op *op) { @@ -2730,6 +2878,14 @@ calc_fft_size(struct rte_bbdev_fft_op *op) return output_size; } +static uint32_t +calc_mldts_size(struct rte_bbdev_mldts_op *op) +{ + uint32_t output_size; + output_size = op->mldts.num_layers * op->mldts.num_rbs * op->mldts.c_rep; + return output_size; +} + static int init_test_op_params(struct test_op_params *op_params, enum rte_bbdev_op_type op_type, const int expected_status, @@ -2744,6 +2900,9 @@ init_test_op_params(struct test_op_params *op_params, else if (op_type == RTE_BBDEV_OP_FFT) ret = rte_bbdev_fft_op_alloc_bulk(ops_mp, &op_params->ref_fft_op, 1); + else if (op_type == RTE_BBDEV_OP_MLDTS) + ret = rte_bbdev_mldts_op_alloc_bulk(ops_mp, + &op_params->ref_mldts_op, 1); else ret = rte_bbdev_enc_op_alloc_bulk(ops_mp, &op_params->ref_enc_op, 1); @@ -2763,6 +2922,8 @@ init_test_op_params(struct test_op_params *op_params, op_params->ref_enc_op->status = expected_status; else if (op_type == RTE_BBDEV_OP_FFT) op_params->ref_fft_op->status = expected_status; + else if (op_type == RTE_BBDEV_OP_MLDTS) + op_params->ref_mldts_op->status = expected_status; return 0; } @@ -2831,6 +2992,8 @@ run_test_case_on_device(test_case_function *test_case_func, uint8_t dev_id, create_reference_ldpc_dec_op(op_params->ref_dec_op); else if (test_vector.op_type == RTE_BBDEV_OP_FFT) create_reference_fft_op(op_params->ref_fft_op); + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) + create_reference_mldts_op(op_params->ref_mldts_op); for (i = 0; i < ad->nb_queues; ++i) { f_ret = fill_queue_buffers(op_params, @@ -3047,6 +3210,11 @@ dequeue_event_callback(uint16_t dev_id, &tp->fft_ops[ __atomic_load_n(&tp->nb_dequeued, __ATOMIC_RELAXED)], burst_sz); + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) + deq = rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, + &tp->mldts_ops[ + __atomic_load_n(&tp->nb_dequeued, __ATOMIC_RELAXED)], + burst_sz); else /*RTE_BBDEV_OP_TURBO_ENC*/ deq = rte_bbdev_dequeue_enc_ops(dev_id, queue_id, &tp->enc_ops[ @@ -3093,6 +3261,10 @@ dequeue_event_callback(uint16_t dev_id, struct rte_bbdev_fft_op *ref_op = tp->op_params->ref_fft_op; ret = validate_fft_op(tp->fft_ops, num_ops, ref_op); rte_bbdev_fft_op_free_bulk(tp->fft_ops, deq); + } else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) { + struct rte_bbdev_mldts_op *ref_op = tp->op_params->ref_mldts_op; + ret = validate_mldts_op(tp->mldts_ops, num_ops, ref_op); + rte_bbdev_mldts_op_free_bulk(tp->mldts_ops, deq); } else if (test_vector.op_type == RTE_BBDEV_OP_LDPC_DEC) { struct rte_bbdev_dec_op *ref_op = tp->op_params->ref_dec_op; ret = validate_ldpc_dec_op(tp->dec_ops, num_ops, ref_op, @@ -3118,6 +3290,9 @@ dequeue_event_callback(uint16_t dev_id, case RTE_BBDEV_OP_FFT: tb_len_bits = calc_fft_size(tp->op_params->ref_fft_op); break; + case RTE_BBDEV_OP_MLDTS: + tb_len_bits = calc_mldts_size(tp->op_params->ref_mldts_op); + break; case RTE_BBDEV_OP_LDPC_ENC: tb_len_bits = calc_ldpc_enc_TB_size(tp->op_params->ref_enc_op); break; @@ -3593,6 +3768,88 @@ throughput_intr_lcore_fft(void *arg) return TEST_SUCCESS; } +static int +throughput_intr_lcore_mldts(void *arg) +{ + struct thread_params *tp = arg; + unsigned int enqueued; + const uint16_t queue_id = tp->queue_id; + const uint16_t burst_sz = tp->op_params->burst_sz; + const uint16_t num_to_process = tp->op_params->num_to_process; + struct rte_bbdev_mldts_op *ops[num_to_process]; + struct test_buffers *bufs = NULL; + struct rte_bbdev_info info; + int ret, i, j; + uint16_t num_to_enq, enq; + + TEST_ASSERT_SUCCESS((burst_sz > MAX_BURST), "BURST_SIZE should be <= %u", MAX_BURST); + + TEST_ASSERT_SUCCESS(rte_bbdev_queue_intr_enable(tp->dev_id, queue_id), + "Failed to enable interrupts for dev: %u, queue_id: %u", + tp->dev_id, queue_id); + + rte_bbdev_info_get(tp->dev_id, &info); + + TEST_ASSERT_SUCCESS((num_to_process > info.drv.queue_size_lim), + "NUM_OPS cannot exceed %u for this device", + info.drv.queue_size_lim); + + bufs = &tp->op_params->q_bufs[GET_SOCKET(info.socket_id)][queue_id]; + + __atomic_store_n(&tp->processing_status, 0, __ATOMIC_RELAXED); + __atomic_store_n(&tp->nb_dequeued, 0, __ATOMIC_RELAXED); + + rte_wait_until_equal_16(&tp->op_params->sync, SYNC_START, __ATOMIC_RELAXED); + + ret = rte_bbdev_mldts_op_alloc_bulk(tp->op_params->mp, ops, num_to_process); + TEST_ASSERT_SUCCESS(ret, "Allocation failed for %d ops", num_to_process); + if (test_vector.op_type != RTE_BBDEV_OP_NONE) + copy_reference_mldts_op(ops, num_to_process, 0, bufs->inputs, bufs->harq_inputs, + bufs->hard_outputs, tp->op_params->ref_mldts_op); + + /* Set counter to validate the ordering */ + for (j = 0; j < num_to_process; ++j) + ops[j]->opaque_data = (void *)(uintptr_t)j; + + for (j = 0; j < TEST_REPETITIONS; ++j) { + for (i = 0; i < num_to_process; ++i) + mbuf_reset(ops[i]->mldts.output.data); + + tp->start_time = rte_rdtsc_precise(); + for (enqueued = 0; enqueued < num_to_process;) { + num_to_enq = burst_sz; + + if (unlikely(num_to_process - enqueued < num_to_enq)) + num_to_enq = num_to_process - enqueued; + + enq = 0; + do { + enq += rte_bbdev_enqueue_mldts_ops(tp->dev_id, + queue_id, &ops[enqueued], num_to_enq); + } while (unlikely(enq != num_to_enq)); + enqueued += enq; + + /* Write to thread burst_sz current number of enqueued + * descriptors. It ensures that proper number of + * descriptors will be dequeued in callback + * function - needed for last batch in case where + * the number of operations is not a multiple of + * burst size. + */ + __atomic_store_n(&tp->burst_sz, num_to_enq, __ATOMIC_RELAXED); + + /* Wait until processing of previous batch is + * completed + */ + rte_wait_until_equal_16(&tp->nb_dequeued, enqueued, __ATOMIC_RELAXED); + } + if (j != TEST_REPETITIONS - 1) + __atomic_store_n(&tp->nb_dequeued, 0, __ATOMIC_RELAXED); + } + + return TEST_SUCCESS; +} + static int throughput_pmd_lcore_dec(void *arg) { @@ -4403,6 +4660,104 @@ throughput_pmd_lcore_fft(void *arg) return TEST_SUCCESS; } +static int +throughput_pmd_lcore_mldts(void *arg) +{ + struct thread_params *tp = arg; + uint16_t enq, deq; + uint64_t total_time = 0, start_time; + const uint16_t queue_id = tp->queue_id; + const uint16_t burst_sz = tp->op_params->burst_sz; + const uint16_t num_ops = tp->op_params->num_to_process; + struct rte_bbdev_mldts_op *ops_enq[num_ops]; + struct rte_bbdev_mldts_op *ops_deq[num_ops]; + struct rte_bbdev_mldts_op *ref_op = tp->op_params->ref_mldts_op; + struct test_buffers *bufs = NULL; + int i, j, ret; + struct rte_bbdev_info info; + uint16_t num_to_enq; + + TEST_ASSERT_SUCCESS((burst_sz > MAX_BURST), "BURST_SIZE should be <= %u", MAX_BURST); + + rte_bbdev_info_get(tp->dev_id, &info); + + TEST_ASSERT_SUCCESS((num_ops > info.drv.queue_size_lim), + "NUM_OPS cannot exceed %u for this device", + info.drv.queue_size_lim); + + bufs = &tp->op_params->q_bufs[GET_SOCKET(info.socket_id)][queue_id]; + + rte_wait_until_equal_16(&tp->op_params->sync, SYNC_START, __ATOMIC_RELAXED); + + ret = rte_bbdev_mldts_op_alloc_bulk(tp->op_params->mp, ops_enq, num_ops); + TEST_ASSERT_SUCCESS(ret, "Allocation failed for %d ops", num_ops); + + if (test_vector.op_type != RTE_BBDEV_OP_NONE) + copy_reference_mldts_op(ops_enq, num_ops, 0, bufs->inputs, bufs->harq_inputs, + bufs->hard_outputs, ref_op); + + /* Set counter to validate the ordering */ + for (j = 0; j < num_ops; ++j) + ops_enq[j]->opaque_data = (void *)(uintptr_t)j; + + for (i = 0; i < TEST_REPETITIONS; ++i) { + uint32_t time_out = 0; + for (j = 0; j < num_ops; ++j) + mbuf_reset(ops_enq[j]->mldts.output.data); + + start_time = rte_rdtsc_precise(); + + for (enq = 0, deq = 0; enq < num_ops;) { + num_to_enq = burst_sz; + + if (unlikely(num_ops - enq < num_to_enq)) + num_to_enq = num_ops - enq; + + enq += rte_bbdev_enqueue_mldts_ops(tp->dev_id, + queue_id, &ops_enq[enq], num_to_enq); + + deq += rte_bbdev_dequeue_mldts_ops(tp->dev_id, + queue_id, &ops_deq[deq], enq - deq); + time_out++; + if (time_out >= TIME_OUT_POLL) { + timeout_exit(tp->dev_id); + TEST_ASSERT_SUCCESS(TEST_FAILED, "Enqueue timeout!"); + } + } + + /* dequeue the remaining */ + time_out = 0; + while (deq < enq) { + deq += rte_bbdev_dequeue_mldts_ops(tp->dev_id, + queue_id, &ops_deq[deq], enq - deq); + time_out++; + if (time_out >= TIME_OUT_POLL) { + timeout_exit(tp->dev_id); + TEST_ASSERT_SUCCESS(TEST_FAILED, "Dequeue timeout!"); + } + } + + total_time += rte_rdtsc_precise() - start_time; + } + + if (test_vector.op_type != RTE_BBDEV_OP_NONE) { + ret = validate_mldts_op(ops_deq, num_ops, ref_op); + TEST_ASSERT_SUCCESS(ret, "Validation failed!"); + } + + rte_bbdev_mldts_op_free_bulk(ops_enq, num_ops); + + double tb_len_bits = calc_mldts_size(ref_op); + + tp->ops_per_sec = ((double)num_ops * TEST_REPETITIONS) / + ((double)total_time / (double)rte_get_tsc_hz()); + tp->mbps = (((double)(num_ops * TEST_REPETITIONS * tb_len_bits)) / + 1000000.0) / ((double)total_time / + (double)rte_get_tsc_hz()); + + return TEST_SUCCESS; +} + static void print_enc_throughput(struct thread_params *t_params, unsigned int used_cores) { @@ -4624,6 +4979,8 @@ throughput_test(struct active_device *ad, throughput_function = throughput_intr_lcore_ldpc_enc; else if (test_vector.op_type == RTE_BBDEV_OP_FFT) throughput_function = throughput_intr_lcore_fft; + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) + throughput_function = throughput_intr_lcore_mldts; else throughput_function = throughput_intr_lcore_enc; @@ -4646,6 +5003,8 @@ throughput_test(struct active_device *ad, throughput_function = throughput_pmd_lcore_ldpc_enc; else if (test_vector.op_type == RTE_BBDEV_OP_FFT) throughput_function = throughput_pmd_lcore_fft; + else if (test_vector.op_type == RTE_BBDEV_OP_MLDTS) + throughput_function = throughput_pmd_lcore_mldts; else throughput_function = throughput_pmd_lcore_enc; } @@ -5139,6 +5498,77 @@ latency_test_fft(struct rte_mempool *mempool, return i; } +static int +latency_test_mldts(struct rte_mempool *mempool, + struct test_buffers *bufs, struct rte_bbdev_mldts_op *ref_op, + uint16_t dev_id, uint16_t queue_id, + const uint16_t num_to_process, uint16_t burst_sz, + uint64_t *total_time, uint64_t *min_time, uint64_t *max_time) +{ + int ret = TEST_SUCCESS; + uint16_t i, j, dequeued; + struct rte_bbdev_mldts_op *ops_enq[MAX_BURST], *ops_deq[MAX_BURST]; + uint64_t start_time = 0, last_time = 0; + + for (i = 0, dequeued = 0; dequeued < num_to_process; ++i) { + uint16_t enq = 0, deq = 0; + uint32_t time_out = 0; + bool first_time = true; + last_time = 0; + + if (unlikely(num_to_process - dequeued < burst_sz)) + burst_sz = num_to_process - dequeued; + + ret = rte_bbdev_mldts_op_alloc_bulk(mempool, ops_enq, burst_sz); + TEST_ASSERT_SUCCESS(ret, "rte_bbdev_mldts_op_alloc_bulk() failed"); + if (test_vector.op_type != RTE_BBDEV_OP_NONE) + copy_reference_mldts_op(ops_enq, burst_sz, dequeued, + bufs->inputs, bufs->harq_inputs, + bufs->hard_outputs, + ref_op); + + /* Set counter to validate the ordering */ + for (j = 0; j < burst_sz; ++j) + ops_enq[j]->opaque_data = (void *)(uintptr_t)j; + + start_time = rte_rdtsc_precise(); + + enq = rte_bbdev_enqueue_mldts_ops(dev_id, queue_id, &ops_enq[enq], burst_sz); + TEST_ASSERT(enq == burst_sz, + "Error enqueueing burst, expected %u, got %u", + burst_sz, enq); + + /* Dequeue */ + do { + deq += rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, + &ops_deq[deq], burst_sz - deq); + if (likely(first_time && (deq > 0))) { + last_time += rte_rdtsc_precise() - start_time; + first_time = false; + } + time_out++; + if (time_out >= TIME_OUT_POLL) { + timeout_exit(dev_id); + TEST_ASSERT_SUCCESS(TEST_FAILED, "Dequeue timeout!"); + } + } while (unlikely(burst_sz != deq)); + + *max_time = RTE_MAX(*max_time, last_time); + *min_time = RTE_MIN(*min_time, last_time); + *total_time += last_time; + + if (test_vector.op_type != RTE_BBDEV_OP_NONE) { + ret = validate_mldts_op(ops_deq, burst_sz, ref_op); + TEST_ASSERT_SUCCESS(ret, "Validation failed!"); + } + + rte_bbdev_mldts_op_free_bulk(ops_enq, deq); + dequeued += deq; + } + + return i; +} + /* Common function for running validation and latency test cases */ static int validation_latency_test(struct active_device *ad, @@ -5196,6 +5626,12 @@ validation_latency_test(struct active_device *ad, ad->dev_id, queue_id, num_to_process, burst_sz, &total_time, &min_time, &max_time); + else if (op_type == RTE_BBDEV_OP_MLDTS) + iter = latency_test_mldts(op_params->mp, bufs, + op_params->ref_mldts_op, + ad->dev_id, queue_id, + num_to_process, burst_sz, &total_time, + &min_time, &max_time); else /* RTE_BBDEV_OP_TURBO_ENC */ iter = latency_test_enc(op_params->mp, bufs, op_params->ref_enc_op, @@ -5337,6 +5773,85 @@ offload_latency_test_fft(struct rte_mempool *mempool, struct test_buffers *bufs, return i; } +static int +offload_latency_test_mldts(struct rte_mempool *mempool, struct test_buffers *bufs, + struct rte_bbdev_mldts_op *ref_op, uint16_t dev_id, + uint16_t queue_id, const uint16_t num_to_process, + uint16_t burst_sz, struct test_time_stats *time_st) +{ + int i, dequeued, ret; + struct rte_bbdev_mldts_op *ops_enq[MAX_BURST], *ops_deq[MAX_BURST]; + uint64_t enq_start_time, deq_start_time; + uint64_t enq_sw_last_time, deq_last_time; + struct rte_bbdev_stats stats; + + for (i = 0, dequeued = 0; dequeued < num_to_process; ++i) { + uint16_t enq = 0, deq = 0; + + if (unlikely(num_to_process - dequeued < burst_sz)) + burst_sz = num_to_process - dequeued; + + ret = rte_bbdev_mldts_op_alloc_bulk(mempool, ops_enq, burst_sz); + TEST_ASSERT_SUCCESS(ret, "rte_bbdev_mldts_op_alloc_bulk() failed"); + if (test_vector.op_type != RTE_BBDEV_OP_NONE) + copy_reference_mldts_op(ops_enq, burst_sz, dequeued, + bufs->inputs, bufs->harq_inputs, + bufs->hard_outputs, + ref_op); + + /* Start time meas for enqueue function offload latency */ + enq_start_time = rte_rdtsc_precise(); + do { + enq += rte_bbdev_enqueue_mldts_ops(dev_id, queue_id, + &ops_enq[enq], burst_sz - enq); + } while (unlikely(burst_sz != enq)); + + ret = get_bbdev_queue_stats(dev_id, queue_id, &stats); + TEST_ASSERT_SUCCESS(ret, + "Failed to get stats for queue (%u) of device (%u)", + queue_id, dev_id); + + enq_sw_last_time = rte_rdtsc_precise() - enq_start_time - + stats.acc_offload_cycles; + time_st->enq_sw_max_time = RTE_MAX(time_st->enq_sw_max_time, + enq_sw_last_time); + time_st->enq_sw_min_time = RTE_MIN(time_st->enq_sw_min_time, + enq_sw_last_time); + time_st->enq_sw_total_time += enq_sw_last_time; + + time_st->enq_acc_max_time = RTE_MAX(time_st->enq_acc_max_time, + stats.acc_offload_cycles); + time_st->enq_acc_min_time = RTE_MIN(time_st->enq_acc_min_time, + stats.acc_offload_cycles); + time_st->enq_acc_total_time += stats.acc_offload_cycles; + + /* give time for device to process ops */ + rte_delay_us(WAIT_OFFLOAD_US); + + /* Start time meas for dequeue function offload latency */ + deq_start_time = rte_rdtsc_precise(); + /* Dequeue one operation */ + do { + deq += rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, &ops_deq[deq], enq); + } while (unlikely(deq == 0)); + + deq_last_time = rte_rdtsc_precise() - deq_start_time; + time_st->deq_max_time = RTE_MAX(time_st->deq_max_time, deq_last_time); + time_st->deq_min_time = RTE_MIN(time_st->deq_min_time, deq_last_time); + time_st->deq_total_time += deq_last_time; + + /* Dequeue remaining operations if needed*/ + while (burst_sz != deq) + deq += rte_bbdev_dequeue_mldts_ops(dev_id, queue_id, + &ops_deq[deq], burst_sz - deq); + + rte_bbdev_mldts_op_free_bulk(ops_enq, deq); + dequeued += deq; + } + + return i; +} + static int offload_latency_test_dec(struct rte_mempool *mempool, struct test_buffers *bufs, struct rte_bbdev_dec_op *ref_op, uint16_t dev_id, @@ -5734,6 +6249,10 @@ offload_cost_test(struct active_device *ad, iter = offload_latency_test_fft(op_params->mp, bufs, op_params->ref_fft_op, ad->dev_id, queue_id, num_to_process, burst_sz, &time_st); + else if (op_type == RTE_BBDEV_OP_MLDTS) + iter = offload_latency_test_mldts(op_params->mp, bufs, + op_params->ref_mldts_op, ad->dev_id, queue_id, + num_to_process, burst_sz, &time_st); else iter = offload_latency_test_enc(op_params->mp, bufs, op_params->ref_enc_op, ad->dev_id, queue_id, diff --git a/app/test-bbdev/test_bbdev_vector.c b/app/test-bbdev/test_bbdev_vector.c index 0ef1481f2a..8f464db838 100644 --- a/app/test-bbdev/test_bbdev_vector.c +++ b/app/test-bbdev/test_bbdev_vector.c @@ -244,6 +244,20 @@ op_fft_flag_strtoul(char *token, uint32_t *op_flag_value) return 0; } +/* convert MLD flag from string to unsigned long int*/ +static int +op_mld_flag_strtoul(char *token, uint32_t *op_flag_value) +{ + if (!strcmp(token, "RTE_BBDEV_MLDTS_REP")) + *op_flag_value = RTE_BBDEV_MLDTS_REP; + else { + printf("The given value is not a MLD flag\n"); + return -1; + } + + return 0; +} + /* convert turbo encoder flag from string to unsigned long int*/ static int op_encoder_flag_strtoul(char *token, uint32_t *op_flag_value) @@ -326,6 +340,10 @@ parse_turbo_flags(char *tokens, uint32_t *op_flags, if (op_fft_flag_strtoul(tok, &op_flag_value) == -1) return -1; + } else if (op_type == RTE_BBDEV_OP_MLDTS) { + if (op_mld_flag_strtoul(tok, &op_flag_value) + == -1) + return -1; } else { return -1; } @@ -355,6 +373,8 @@ op_turbo_type_strtol(char *token, enum rte_bbdev_op_type *op_type) *op_type = RTE_BBDEV_OP_LDPC_DEC; else if (!strcmp(token, "RTE_BBDEV_OP_FFT")) *op_type = RTE_BBDEV_OP_FFT; + else if (!strcmp(token, "RTE_BBDEV_OP_MLDTS")) + *op_type = RTE_BBDEV_OP_MLDTS; else if (!strcmp(token, "RTE_BBDEV_OP_NONE")) *op_type = RTE_BBDEV_OP_NONE; else { @@ -992,6 +1012,73 @@ parse_fft_params(const char *key_token, char *token, return 0; } +/* parses MLD parameters and assigns to global variable */ +static int +parse_mld_params(const char *key_token, char *token, + struct test_bbdev_vector *vector) +{ + int ret = 0, status = 0; + uint32_t op_flags = 0; + char *err = NULL; + + struct rte_bbdev_op_mldts *mld = &vector->mldts; + + if (starts_with(key_token, "qhy_input")) { + ret = parse_data_entry(key_token, token, vector, + DATA_INPUT, "qhy_input"); + } else if (starts_with(key_token, "r_input")) { + ret = parse_data_entry(key_token, token, vector, + DATA_HARQ_INPUT, "r_input"); + } else if (starts_with(key_token, "output")) { + ret = parse_data_entry(key_token, token, vector, + DATA_HARD_OUTPUT, "output"); + } else if (!strcmp(key_token, "layers")) { + mld->num_layers = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "layer1")) { + mld->q_m[0] = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "layer2")) { + mld->q_m[1] = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "layer3")) { + mld->q_m[2] = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "layer4")) { + mld->q_m[3] = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "crep")) { + mld->c_rep = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "rrep")) { + mld->r_rep = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "rbs")) { + mld->num_rbs = (uint32_t) strtoul(token, &err, 0); + ret = ((err == NULL) || (*err != '\0')) ? -1 : 0; + } else if (!strcmp(key_token, "op_flags")) { + vector->mask |= TEST_BBDEV_VF_OP_FLAGS; + ret = parse_turbo_flags(token, &op_flags, vector->op_type); + if (!ret) + mld->op_flags = op_flags; + } else if (!strcmp(key_token, "expected_status")) { + vector->mask |= TEST_BBDEV_VF_EXPECTED_STATUS; + ret = parse_expected_status(token, &status, vector->op_type); + if (!ret) + vector->expected_status = status; + } else { + printf("Not valid mld key: '%s'\n", key_token); + return -1; + } + + if (ret != 0) { + printf("Failed with convert '%s\t%s'\n", key_token, token); + return -1; + } + + return 0; +} + /* checks the type of key and assigns data */ static int parse_entry(char *entry, struct test_bbdev_vector *vector) @@ -1046,6 +1133,9 @@ parse_entry(char *entry, struct test_bbdev_vector *vector) } else if (vector->op_type == RTE_BBDEV_OP_FFT) { if (parse_fft_params(key_token, token, vector) == -1) return -1; + } else if (vector->op_type == RTE_BBDEV_OP_MLDTS) { + if (parse_mld_params(key_token, token, vector) == -1) + return -1; } return 0; @@ -1132,6 +1222,25 @@ check_fft_segments(struct test_bbdev_vector *vector) return 0; } +static int +check_mld_segments(struct test_bbdev_vector *vector) +{ + unsigned char i; + + for (i = 0; i < vector->entries[DATA_INPUT].nb_segments; i++) + if (vector->entries[DATA_INPUT].segments[i].addr == NULL) + return -1; + + for (i = 0; i < vector->entries[DATA_HARQ_INPUT].nb_segments; i++) + if (vector->entries[DATA_HARQ_INPUT].segments[i].addr == NULL) + return -1; + + for (i = 0; i < vector->entries[DATA_HARD_OUTPUT].nb_segments; i++) + if (vector->entries[DATA_HARD_OUTPUT].segments[i].addr == NULL) + return -1; + return 0; +} + static int check_decoder_llr_spec(struct test_bbdev_vector *vector) { @@ -1359,6 +1468,26 @@ check_fft(struct test_bbdev_vector *vector) return 0; } +/* checks mld parameters */ +static int +check_mld(struct test_bbdev_vector *vector) +{ + const int mask = vector->mask; + + if (check_mld_segments(vector) < 0) + return -1; + + /* Check which params were set */ + if (!(mask & TEST_BBDEV_VF_OP_FLAGS)) { + printf( + "WARNING: op_flags was not specified in vector file and capabilities will not be validated\n"); + } + if (!(mask & TEST_BBDEV_VF_EXPECTED_STATUS)) + printf( + "WARNING: expected_status was not specified in vector file and will be set to 0\n"); + return 0; +} + /* checks encoder parameters */ static int check_encoder(struct test_bbdev_vector *vector) @@ -1520,6 +1649,9 @@ bbdev_check_vector(struct test_bbdev_vector *vector) } else if (vector->op_type == RTE_BBDEV_OP_FFT) { if (check_fft(vector) == -1) return -1; + } else if (vector->op_type == RTE_BBDEV_OP_MLDTS) { + if (check_mld(vector) == -1) + return -1; } else if (vector->op_type != RTE_BBDEV_OP_NONE) { printf("Vector was not filled\n"); return -1; diff --git a/app/test-bbdev/test_bbdev_vector.h b/app/test-bbdev/test_bbdev_vector.h index 2ea271ffb7..14b8ef2764 100644 --- a/app/test-bbdev/test_bbdev_vector.h +++ b/app/test-bbdev/test_bbdev_vector.h @@ -65,6 +65,7 @@ struct test_bbdev_vector { struct rte_bbdev_op_ldpc_dec ldpc_dec; struct rte_bbdev_op_ldpc_enc ldpc_enc; struct rte_bbdev_op_fft fft; + struct rte_bbdev_op_mldts mldts; }; /* Additional storage for op data entries */ struct op_data_entries entries[DATA_NUM_TYPES];