net/virtio: fix mbuf count on Rx queue setup
[dpdk.git] / app / test-bbdev / test_bbdev_perf.c
index 9245d6f..469597b 100644 (file)
 #define GET_SOCKET(socket_id) (((socket_id) == SOCKET_ID_ANY) ? 0 : (socket_id))
 
 #define MAX_QUEUES RTE_MAX_LCORE
-#define TEST_REPETITIONS 1000
+#define TEST_REPETITIONS 100
+#define WAIT_OFFLOAD_US 1000
 
-#ifdef RTE_LIBRTE_PMD_BBDEV_FPGA_LTE_FEC
+#ifdef RTE_BASEBAND_FPGA_LTE_FEC
 #include <fpga_lte_fec.h>
 #define FPGA_LTE_PF_DRIVER_NAME ("intel_fpga_lte_fec_pf")
 #define FPGA_LTE_VF_DRIVER_NAME ("intel_fpga_lte_fec_vf")
@@ -39,7 +40,7 @@
 #define FLR_4G_TIMEOUT 610
 #endif
 
-#ifdef RTE_LIBRTE_PMD_BBDEV_FPGA_5GNR_FEC
+#ifdef RTE_BASEBAND_FPGA_5GNR_FEC
 #include <rte_pmd_fpga_5gnr_fec.h>
 #define FPGA_5GNR_PF_DRIVER_NAME ("intel_fpga_5gnr_fec_pf")
 #define FPGA_5GNR_VF_DRIVER_NAME ("intel_fpga_5gnr_fec_vf")
@@ -52,7 +53,7 @@
 #define FLR_5G_TIMEOUT 610
 #endif
 
-#ifdef RTE_LIBRTE_PMD_BBDEV_ACC100
+#ifdef RTE_BASEBAND_ACC100
 #include <rte_acc100_cfg.h>
 #define ACC100PF_DRIVER_NAME   ("intel_acc100_pf")
 #define ACC100VF_DRIVER_NAME   ("intel_acc100_vf")
@@ -371,14 +372,14 @@ check_dev_cap(const struct rte_bbdev_info *dev_info)
                        if (nb_harq_inputs > cap->num_buffers_hard_out) {
                                printf(
                                        "Too many HARQ inputs defined: %u, max: %u\n",
-                                       nb_hard_outputs,
+                                       nb_harq_inputs,
                                        cap->num_buffers_hard_out);
                                return TEST_FAILED;
                        }
                        if (nb_harq_outputs > cap->num_buffers_hard_out) {
                                printf(
                                        "Too many HARQ outputs defined: %u, max: %u\n",
-                                       nb_hard_outputs,
+                                       nb_harq_outputs,
                                        cap->num_buffers_hard_out);
                                return TEST_FAILED;
                        }
@@ -577,17 +578,17 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info,
 /* Configure fpga lte fec with PF & VF values
  * if '-i' flag is set and using fpga device
  */
-#ifdef RTE_LIBRTE_PMD_BBDEV_FPGA_LTE_FEC
+#ifdef RTE_BASEBAND_FPGA_LTE_FEC
        if ((get_init_device() == true) &&
                (!strcmp(info->drv.driver_name, FPGA_LTE_PF_DRIVER_NAME))) {
-               struct fpga_lte_fec_conf conf;
+               struct rte_fpga_lte_fec_conf conf;
                unsigned int i;
 
                printf("Configure FPGA LTE FEC Driver %s with default values\n",
                                info->drv.driver_name);
 
                /* clear default configuration before initialization */
-               memset(&conf, 0, sizeof(struct fpga_lte_fec_conf));
+               memset(&conf, 0, sizeof(struct rte_fpga_lte_fec_conf));
 
                /* Set PF mode :
                 * true if PF is used for data plane
@@ -615,13 +616,13 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info,
                conf.flr_time_out = FLR_4G_TIMEOUT;
 
                /* setup FPGA PF with configuration information */
-               ret = fpga_lte_fec_configure(info->dev_name, &conf);
+               ret = rte_fpga_lte_fec_configure(info->dev_name, &conf);
                TEST_ASSERT_SUCCESS(ret,
                                "Failed to configure 4G FPGA PF for bbdev %s",
                                info->dev_name);
        }
 #endif
-#ifdef RTE_LIBRTE_PMD_BBDEV_FPGA_5GNR_FEC
+#ifdef RTE_BASEBAND_FPGA_5GNR_FEC
        if ((get_init_device() == true) &&
                (!strcmp(info->drv.driver_name, FPGA_5GNR_PF_DRIVER_NAME))) {
                struct rte_fpga_5gnr_fec_conf conf;
@@ -665,7 +666,7 @@ add_bbdev_dev(uint8_t dev_id, struct rte_bbdev_info *info,
                                info->dev_name);
        }
 #endif
-#ifdef RTE_LIBRTE_PMD_BBDEV_ACC100
+#ifdef RTE_BASEBAND_ACC100
        if ((get_init_device() == true) &&
                (!strcmp(info->drv.driver_name, ACC100PF_DRIVER_NAME))) {
                struct rte_acc100_conf conf;
@@ -956,6 +957,9 @@ init_op_data_objs(struct rte_bbdev_op_data *bufs,
                        if ((op_type == DATA_INPUT) && large_input) {
                                /* Allocate a fake overused mbuf */
                                data = rte_malloc(NULL, seg->length, 0);
+                               TEST_ASSERT_NOT_NULL(data,
+                                       "rte malloc failed with %u bytes",
+                                       seg->length);
                                memcpy(data, seg->addr, seg->length);
                                m_head->buf_addr = data;
                                m_head->buf_iova = rte_malloc_virt2iova(data);
@@ -1257,7 +1261,7 @@ copy_reference_dec_op(struct rte_bbdev_dec_op **ops, unsigned int n,
        struct rte_bbdev_op_turbo_dec *turbo_dec = &ref_op->turbo_dec;
 
        for (i = 0; i < n; ++i) {
-               if (turbo_dec->code_block_mode == 0) {
+               if (turbo_dec->code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK) {
                        ops[i]->turbo_dec.tb_params.ea =
                                        turbo_dec->tb_params.ea;
                        ops[i]->turbo_dec.tb_params.eb =
@@ -1305,7 +1309,7 @@ copy_reference_enc_op(struct rte_bbdev_enc_op **ops, unsigned int n,
        unsigned int i;
        struct rte_bbdev_op_turbo_enc *turbo_enc = &ref_op->turbo_enc;
        for (i = 0; i < n; ++i) {
-               if (turbo_enc->code_block_mode == 0) {
+               if (turbo_enc->code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK) {
                        ops[i]->turbo_enc.tb_params.ea =
                                        turbo_enc->tb_params.ea;
                        ops[i]->turbo_enc.tb_params.eb =
@@ -1660,7 +1664,7 @@ copy_reference_ldpc_dec_op(struct rte_bbdev_dec_op **ops, unsigned int n,
        struct rte_bbdev_op_ldpc_dec *ldpc_dec = &ref_op->ldpc_dec;
 
        for (i = 0; i < n; ++i) {
-               if (ldpc_dec->code_block_mode == 0) {
+               if (ldpc_dec->code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK) {
                        ops[i]->ldpc_dec.tb_params.ea =
                                        ldpc_dec->tb_params.ea;
                        ops[i]->ldpc_dec.tb_params.eb =
@@ -1714,7 +1718,7 @@ copy_reference_ldpc_enc_op(struct rte_bbdev_enc_op **ops, unsigned int n,
        unsigned int i;
        struct rte_bbdev_op_ldpc_enc *ldpc_enc = &ref_op->ldpc_enc;
        for (i = 0; i < n; ++i) {
-               if (ldpc_enc->code_block_mode == 0) {
+               if (ldpc_enc->code_block_mode == RTE_BBDEV_TRANSPORT_BLOCK) {
                        ops[i]->ldpc_enc.tb_params.ea = ldpc_enc->tb_params.ea;
                        ops[i]->ldpc_enc.tb_params.eb = ldpc_enc->tb_params.eb;
                        ops[i]->ldpc_enc.tb_params.cab =
@@ -2238,7 +2242,7 @@ calc_dec_TB_size(struct rte_bbdev_dec_op *op)
        uint8_t i;
        uint32_t c, r, tb_size = 0;
 
-       if (op->turbo_dec.code_block_mode) {
+       if (op->turbo_dec.code_block_mode == RTE_BBDEV_CODE_BLOCK) {
                tb_size = op->turbo_dec.tb_params.k_neg;
        } else {
                c = op->turbo_dec.tb_params.c;
@@ -2258,7 +2262,7 @@ calc_ldpc_dec_TB_size(struct rte_bbdev_dec_op *op)
        uint32_t c, r, tb_size = 0;
        uint16_t sys_cols = (op->ldpc_dec.basegraph == 1) ? 22 : 10;
 
-       if (op->ldpc_dec.code_block_mode) {
+       if (op->ldpc_dec.code_block_mode == RTE_BBDEV_CODE_BLOCK) {
                tb_size = sys_cols * op->ldpc_dec.z_c - op->ldpc_dec.n_filler;
        } else {
                c = op->ldpc_dec.tb_params.c;
@@ -2276,7 +2280,7 @@ calc_enc_TB_size(struct rte_bbdev_enc_op *op)
        uint8_t i;
        uint32_t c, r, tb_size = 0;
 
-       if (op->turbo_enc.code_block_mode) {
+       if (op->turbo_enc.code_block_mode == RTE_BBDEV_CODE_BLOCK) {
                tb_size = op->turbo_enc.tb_params.k_neg;
        } else {
                c = op->turbo_enc.tb_params.c;
@@ -2296,7 +2300,7 @@ calc_ldpc_enc_TB_size(struct rte_bbdev_enc_op *op)
        uint32_t c, r, tb_size = 0;
        uint16_t sys_cols = (op->ldpc_enc.basegraph == 1) ? 22 : 10;
 
-       if (op->turbo_enc.code_block_mode) {
+       if (op->ldpc_enc.code_block_mode == RTE_BBDEV_CODE_BLOCK) {
                tb_size = sys_cols * op->ldpc_enc.z_c - op->ldpc_enc.n_filler;
        } else {
                c = op->turbo_enc.tb_params.c;
@@ -2467,7 +2471,7 @@ retrieve_harq_ddr(uint16_t dev_id, uint16_t queue_id,
 {
        uint16_t j;
        int save_status, ret;
-       uint32_t harq_offset = (uint32_t) queue_id * HARQ_INCR * 1024;
+       uint32_t harq_offset = (uint32_t) queue_id * HARQ_INCR * MAX_OPS;
        struct rte_bbdev_dec_op *ops_deq[MAX_BURST];
        uint32_t flags = ops[0]->ldpc_dec.op_flags;
        bool loopback = flags & RTE_BBDEV_LDPC_INTERNAL_HARQ_MEMORY_LOOPBACK;
@@ -2513,20 +2517,20 @@ preload_harq_ddr(uint16_t dev_id, uint16_t queue_id,
                bool preload)
 {
        uint16_t j;
-       int ret;
-       uint32_t harq_offset = (uint32_t) queue_id * HARQ_INCR * 1024;
-       struct rte_bbdev_op_data save_hc_in, save_hc_out;
-       struct rte_bbdev_dec_op *ops_deq[MAX_BURST];
+       int deq;
+       uint32_t harq_offset = (uint32_t) queue_id * HARQ_INCR * MAX_OPS;
+       struct rte_bbdev_op_data save_hc_in[MAX_OPS], save_hc_out[MAX_OPS];
+       struct rte_bbdev_dec_op *ops_deq[MAX_OPS];
        uint32_t flags = ops[0]->ldpc_dec.op_flags;
        bool mem_in = flags & RTE_BBDEV_LDPC_INTERNAL_HARQ_MEMORY_IN_ENABLE;
        bool hc_in = flags & RTE_BBDEV_LDPC_HQ_COMBINE_IN_ENABLE;
        bool mem_out = flags & RTE_BBDEV_LDPC_INTERNAL_HARQ_MEMORY_OUT_ENABLE;
        bool hc_out = flags & RTE_BBDEV_LDPC_HQ_COMBINE_OUT_ENABLE;
        bool h_comp = flags & RTE_BBDEV_LDPC_HARQ_6BIT_COMPRESSION;
-       for (j = 0; j < n; ++j) {
-               if ((mem_in || hc_in) && preload) {
-                       save_hc_in = ops[j]->ldpc_dec.harq_combined_input;
-                       save_hc_out = ops[j]->ldpc_dec.harq_combined_output;
+       if ((mem_in || hc_in) && preload) {
+               for (j = 0; j < n; ++j) {
+                       save_hc_in[j] = ops[j]->ldpc_dec.harq_combined_input;
+                       save_hc_out[j] = ops[j]->ldpc_dec.harq_combined_output;
                        ops[j]->ldpc_dec.op_flags =
                                RTE_BBDEV_LDPC_INTERNAL_HARQ_MEMORY_LOOPBACK +
                                RTE_BBDEV_LDPC_INTERNAL_HARQ_MEMORY_OUT_ENABLE;
@@ -2536,16 +2540,23 @@ preload_harq_ddr(uint16_t dev_id, uint16_t queue_id,
                        ops[j]->ldpc_dec.harq_combined_output.offset =
                                        harq_offset;
                        ops[j]->ldpc_dec.harq_combined_input.offset = 0;
-                       rte_bbdev_enqueue_ldpc_dec_ops(dev_id, queue_id,
-                                       &ops[j], 1);
-                       ret = 0;
-                       while (ret == 0)
-                               ret = rte_bbdev_dequeue_ldpc_dec_ops(
-                                       dev_id, queue_id, &ops_deq[j], 1);
+                       harq_offset += HARQ_INCR;
+               }
+               rte_bbdev_enqueue_ldpc_dec_ops(dev_id, queue_id, &ops[0], n);
+               deq = 0;
+               while (deq != n)
+                       deq += rte_bbdev_dequeue_ldpc_dec_ops(
+                                       dev_id, queue_id, &ops_deq[deq],
+                                       n - deq);
+               /* Restore the operations */
+               for (j = 0; j < n; ++j) {
                        ops[j]->ldpc_dec.op_flags = flags;
-                       ops[j]->ldpc_dec.harq_combined_input = save_hc_in;
-                       ops[j]->ldpc_dec.harq_combined_output = save_hc_out;
+                       ops[j]->ldpc_dec.harq_combined_input = save_hc_in[j];
+                       ops[j]->ldpc_dec.harq_combined_output = save_hc_out[j];
                }
+       }
+       harq_offset = (uint32_t) queue_id * HARQ_INCR * MAX_OPS;
+       for (j = 0; j < n; ++j) {
                /* Adjust HARQ offset when we reach external DDR */
                if (mem_in || hc_in)
                        ops[j]->ldpc_dec.harq_combined_input.offset
@@ -3231,11 +3242,9 @@ bler_pmd_lcore_ldpc_dec(void *arg)
                                mbuf_reset(
                                ops_enq[j]->ldpc_dec.harq_combined_output.data);
                }
-               if (extDdr) {
-                       bool preload = i == (TEST_REPETITIONS - 1);
+               if (extDdr)
                        preload_harq_ddr(tp->dev_id, queue_id, ops_enq,
-                                       num_ops, preload);
-               }
+                                       num_ops, true);
                start_time = rte_rdtsc_precise();
 
                for (enq = 0, deq = 0; enq < num_ops;) {
@@ -3362,11 +3371,9 @@ throughput_pmd_lcore_ldpc_dec(void *arg)
                                mbuf_reset(
                                ops_enq[j]->ldpc_dec.harq_combined_output.data);
                }
-               if (extDdr) {
-                       bool preload = i == (TEST_REPETITIONS - 1);
+               if (extDdr)
                        preload_harq_ddr(tp->dev_id, queue_id, ops_enq,
-                                       num_ops, preload);
-               }
+                                       num_ops, true);
                start_time = rte_rdtsc_precise();
 
                for (enq = 0, deq = 0; enq < num_ops;) {
@@ -3715,21 +3722,25 @@ bler_test(struct active_device *ad,
                        RTE_ALIGN(sizeof(struct thread_params) * num_lcores,
                                RTE_CACHE_LINE_SIZE));
 
-       if (test_vector.op_type == RTE_BBDEV_OP_LDPC_DEC)
+       if ((test_vector.op_type == RTE_BBDEV_OP_LDPC_DEC) &&
+                       !check_bit(test_vector.ldpc_dec.op_flags,
+                       RTE_BBDEV_LDPC_INTERNAL_HARQ_MEMORY_LOOPBACK)
+                       && !check_bit(test_vector.ldpc_dec.op_flags,
+                       RTE_BBDEV_LDPC_LLR_COMPRESSION))
                bler_function = bler_pmd_lcore_ldpc_dec;
        else
                return TEST_SKIPPED;
 
        rte_atomic16_set(&op_params->sync, SYNC_WAIT);
 
-       /* Master core is set at first entry */
+       /* Main core is set at first entry */
        t_params[0].dev_id = ad->dev_id;
        t_params[0].lcore_id = rte_lcore_id();
        t_params[0].op_params = op_params;
        t_params[0].queue_id = ad->queue_ids[used_cores++];
        t_params[0].iter_count = 0;
 
-       RTE_LCORE_FOREACH_SLAVE(lcore_id) {
+       RTE_LCORE_FOREACH_WORKER(lcore_id) {
                if (used_cores >= num_lcores)
                        break;
 
@@ -3746,7 +3757,7 @@ bler_test(struct active_device *ad,
        rte_atomic16_set(&op_params->sync, SYNC_START);
        ret = bler_function(&t_params[0]);
 
-       /* Master core is always used */
+       /* Main core is always used */
        for (used_cores = 1; used_cores < num_lcores; used_cores++)
                ret |= rte_eal_wait_lcore(t_params[used_cores].lcore_id);
 
@@ -3840,14 +3851,14 @@ throughput_test(struct active_device *ad,
 
        rte_atomic16_set(&op_params->sync, SYNC_WAIT);
 
-       /* Master core is set at first entry */
+       /* Main core is set at first entry */
        t_params[0].dev_id = ad->dev_id;
        t_params[0].lcore_id = rte_lcore_id();
        t_params[0].op_params = op_params;
        t_params[0].queue_id = ad->queue_ids[used_cores++];
        t_params[0].iter_count = 0;
 
-       RTE_LCORE_FOREACH_SLAVE(lcore_id) {
+       RTE_LCORE_FOREACH_WORKER(lcore_id) {
                if (used_cores >= num_lcores)
                        break;
 
@@ -3864,7 +3875,7 @@ throughput_test(struct active_device *ad,
        rte_atomic16_set(&op_params->sync, SYNC_START);
        ret = throughput_function(&t_params[0]);
 
-       /* Master core is always used */
+       /* Main core is always used */
        for (used_cores = 1; used_cores < num_lcores; used_cores++)
                ret |= rte_eal_wait_lcore(t_params[used_cores].lcore_id);
 
@@ -3888,7 +3899,7 @@ throughput_test(struct active_device *ad,
        /* In interrupt TC we need to wait for the interrupt callback to deqeue
         * all pending operations. Skip waiting for queues which reported an
         * error using processing_status variable.
-        * Wait for master lcore operations.
+        * Wait for main lcore operations.
         */
        tp = &t_params[0];
        while ((rte_atomic16_read(&tp->nb_dequeued) <
@@ -3901,7 +3912,7 @@ throughput_test(struct active_device *ad,
        tp->mbps /= TEST_REPETITIONS;
        ret |= (int)rte_atomic16_read(&tp->processing_status);
 
-       /* Wait for slave lcores operations */
+       /* Wait for worker lcores operations */
        for (used_cores = 1; used_cores < num_lcores; used_cores++) {
                tp = &t_params[used_cores];
 
@@ -3999,12 +4010,14 @@ latency_test_dec(struct rte_mempool *mempool,
        return i;
 }
 
+/* Test case for latency/validation for LDPC Decoder */
 static int
 latency_test_ldpc_dec(struct rte_mempool *mempool,
                struct test_buffers *bufs, struct rte_bbdev_dec_op *ref_op,
                int vector_mask, 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)
+               uint64_t *total_time, uint64_t *min_time, uint64_t *max_time,
+               bool disable_et)
 {
        int ret = TEST_SUCCESS;
        uint16_t i, j, dequeued;
@@ -4026,7 +4039,7 @@ latency_test_ldpc_dec(struct rte_mempool *mempool,
                                "rte_bbdev_dec_op_alloc_bulk() failed");
 
                /* For latency tests we need to disable early termination */
-               if (check_bit(ref_op->ldpc_dec.op_flags,
+               if (disable_et && check_bit(ref_op->ldpc_dec.op_flags,
                                RTE_BBDEV_LDPC_ITERATION_STOP_ENABLE))
                        ref_op->ldpc_dec.op_flags -=
                                        RTE_BBDEV_LDPC_ITERATION_STOP_ENABLE;
@@ -4221,9 +4234,10 @@ latency_test_ldpc_enc(struct rte_mempool *mempool,
        return i;
 }
 
+/* Common function for running validation and latency test cases */
 static int
-latency_test(struct active_device *ad,
-               struct test_op_params *op_params)
+validation_latency_test(struct active_device *ad,
+               struct test_op_params *op_params, bool latency_flag)
 {
        int iter;
        uint16_t burst_sz = op_params->burst_sz;
@@ -4248,7 +4262,11 @@ latency_test(struct active_device *ad,
        TEST_ASSERT_NOT_NULL(op_type_str, "Invalid op type: %u", op_type);
 
        printf("+ ------------------------------------------------------- +\n");
-       printf("== test: validation/latency\ndev: %s, burst size: %u, num ops: %u, op type: %s\n",
+       if (latency_flag)
+               printf("== test: latency\ndev:");
+       else
+               printf("== test: validation\ndev:");
+       printf("%s, burst size: %u, num ops: %u, op type: %s\n",
                        info.dev_name, burst_sz, num_to_process, op_type_str);
 
        if (op_type == RTE_BBDEV_OP_TURBO_DEC)
@@ -4256,11 +4274,6 @@ latency_test(struct active_device *ad,
                                op_params->ref_dec_op, op_params->vector_mask,
                                ad->dev_id, queue_id, num_to_process,
                                burst_sz, &total_time, &min_time, &max_time);
-       else if (op_type == RTE_BBDEV_OP_TURBO_ENC)
-               iter = latency_test_enc(op_params->mp, bufs,
-                               op_params->ref_enc_op, ad->dev_id, queue_id,
-                               num_to_process, burst_sz, &total_time,
-                               &min_time, &max_time);
        else if (op_type == RTE_BBDEV_OP_LDPC_ENC)
                iter = latency_test_ldpc_enc(op_params->mp, bufs,
                                op_params->ref_enc_op, ad->dev_id, queue_id,
@@ -4270,13 +4283,14 @@ latency_test(struct active_device *ad,
                iter = latency_test_ldpc_dec(op_params->mp, bufs,
                                op_params->ref_dec_op, op_params->vector_mask,
                                ad->dev_id, queue_id, num_to_process,
-                               burst_sz, &total_time, &min_time, &max_time);
-       else
+                               burst_sz, &total_time, &min_time, &max_time,
+                               latency_flag);
+       else /* RTE_BBDEV_OP_TURBO_ENC */
                iter = latency_test_enc(op_params->mp, bufs,
-                                       op_params->ref_enc_op,
-                                       ad->dev_id, queue_id,
-                                       num_to_process, burst_sz, &total_time,
-                                       &min_time, &max_time);
+                               op_params->ref_enc_op,
+                               ad->dev_id, queue_id,
+                               num_to_process, burst_sz, &total_time,
+                               &min_time, &max_time);
 
        if (iter <= 0)
                return TEST_FAILED;
@@ -4295,6 +4309,18 @@ latency_test(struct active_device *ad,
        return TEST_SUCCESS;
 }
 
+static int
+latency_test(struct active_device *ad, struct test_op_params *op_params)
+{
+       return validation_latency_test(ad, op_params, true);
+}
+
+static int
+validation_test(struct active_device *ad, struct test_op_params *op_params)
+{
+       return validation_latency_test(ad, op_params, false);
+}
+
 #ifdef RTE_BBDEV_OFFLOAD_COST
 static int
 get_bbdev_queue_stats(uint16_t dev_id, uint16_t queue_id,
@@ -4370,15 +4396,15 @@ offload_latency_test_dec(struct rte_mempool *mempool, struct test_buffers *bufs,
                time_st->enq_acc_total_time += stats.acc_offload_cycles;
 
                /* give time for device to process ops */
-               rte_delay_us(200);
+               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_dec_ops(dev_id, queue_id,
-                                       &ops_deq[deq], 1);
-               } while (unlikely(deq != 1));
+                                       &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,
@@ -4461,15 +4487,15 @@ offload_latency_test_ldpc_dec(struct rte_mempool *mempool,
                time_st->enq_acc_total_time += stats.acc_offload_cycles;
 
                /* give time for device to process ops */
-               rte_delay_us(200);
+               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_ldpc_dec_ops(dev_id, queue_id,
-                                       &ops_deq[deq], 1);
-               } while (unlikely(deq != 1));
+                                       &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,
@@ -4549,15 +4575,15 @@ offload_latency_test_enc(struct rte_mempool *mempool, struct test_buffers *bufs,
                time_st->enq_acc_total_time += stats.acc_offload_cycles;
 
                /* give time for device to process ops */
-               rte_delay_us(200);
+               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_enc_ops(dev_id, queue_id,
-                                       &ops_deq[deq], 1);
-               } while (unlikely(deq != 1));
+                                       &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,
@@ -4632,15 +4658,15 @@ offload_latency_test_ldpc_enc(struct rte_mempool *mempool,
                time_st->enq_acc_total_time += stats.acc_offload_cycles;
 
                /* give time for device to process ops */
-               rte_delay_us(200);
+               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_ldpc_enc_ops(dev_id, queue_id,
-                                       &ops_deq[deq], 1);
-               } while (unlikely(deq != 1));
+                                       &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,
@@ -4762,6 +4788,23 @@ offload_cost_test(struct active_device *ad,
                        (double)(time_st.deq_max_time * 1000000) /
                        rte_get_tsc_hz());
 
+       struct rte_bbdev_stats stats = {0};
+       get_bbdev_queue_stats(ad->dev_id, queue_id, &stats);
+       if (op_type != RTE_BBDEV_OP_LDPC_DEC) {
+               TEST_ASSERT_SUCCESS(stats.enqueued_count != num_to_process,
+                               "Mismatch in enqueue count %10"PRIu64" %d",
+                               stats.enqueued_count, num_to_process);
+               TEST_ASSERT_SUCCESS(stats.dequeued_count != num_to_process,
+                               "Mismatch in dequeue count %10"PRIu64" %d",
+                               stats.dequeued_count, num_to_process);
+       }
+       TEST_ASSERT_SUCCESS(stats.enqueue_err_count != 0,
+                       "Enqueue count Error %10"PRIu64"",
+                       stats.enqueue_err_count);
+       TEST_ASSERT_SUCCESS(stats.dequeue_err_count != 0,
+                       "Dequeue count Error (%10"PRIu64"",
+                       stats.dequeue_err_count);
+
        return TEST_SUCCESS;
 #endif
 }
@@ -4929,6 +4972,12 @@ latency_tc(void)
        return run_test_case(latency_test);
 }
 
+static int
+validation_tc(void)
+{
+       return run_test_case(validation_test);
+}
+
 static int
 interrupt_tc(void)
 {
@@ -4960,7 +5009,7 @@ static struct unit_test_suite bbdev_validation_testsuite = {
        .setup = testsuite_setup,
        .teardown = testsuite_teardown,
        .unit_test_cases = {
-               TEST_CASE_ST(ut_setup, ut_teardown, latency_tc),
+               TEST_CASE_ST(ut_setup, ut_teardown, validation_tc),
                TEST_CASES_END() /**< NULL terminate unit test array */
        }
 };