#define FLR_4G_TIMEOUT 610
#endif
+#ifdef RTE_LIBRTE_PMD_BBDEV_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")
+#define VF_UL_5G_QUEUE_VALUE 4
+#define VF_DL_5G_QUEUE_VALUE 4
+#define UL_5G_BANDWIDTH 3
+#define DL_5G_BANDWIDTH 3
+#define UL_5G_LOAD_BALANCE 128
+#define DL_5G_LOAD_BALANCE 128
+#define FLR_5G_TIMEOUT 610
+#endif
+
#define OPS_CACHE_SIZE 256U
#define OPS_POOL_SIZE_MIN 511U /* 0.5K per queue */
"Failed to configure 4G FPGA PF for bbdev %s",
info->dev_name);
}
+#endif
+#ifdef RTE_LIBRTE_PMD_BBDEV_FPGA_5GNR_FEC
+ if ((get_init_device() == true) &&
+ (!strcmp(info->drv.driver_name, FPGA_5GNR_PF_DRIVER_NAME))) {
+ struct fpga_5gnr_fec_conf conf;
+ unsigned int i;
+
+ printf("Configure FPGA 5GNR FEC Driver %s with default values\n",
+ info->drv.driver_name);
+
+ /* clear default configuration before initialization */
+ memset(&conf, 0, sizeof(struct fpga_5gnr_fec_conf));
+
+ /* Set PF mode :
+ * true if PF is used for data plane
+ * false for VFs
+ */
+ conf.pf_mode_en = true;
+
+ for (i = 0; i < FPGA_5GNR_FEC_NUM_VFS; ++i) {
+ /* Number of UL queues per VF (fpga supports 8 VFs) */
+ conf.vf_ul_queues_number[i] = VF_UL_5G_QUEUE_VALUE;
+ /* Number of DL queues per VF (fpga supports 8 VFs) */
+ conf.vf_dl_queues_number[i] = VF_DL_5G_QUEUE_VALUE;
+ }
+
+ /* UL bandwidth. Needed for schedule algorithm */
+ conf.ul_bandwidth = UL_5G_BANDWIDTH;
+ /* DL bandwidth */
+ conf.dl_bandwidth = DL_5G_BANDWIDTH;
+
+ /* UL & DL load Balance Factor to 64 */
+ conf.ul_load_balance = UL_5G_LOAD_BALANCE;
+ conf.dl_load_balance = DL_5G_LOAD_BALANCE;
+
+ /**< FLR timeout value */
+ conf.flr_time_out = FLR_5G_TIMEOUT;
+
+ /* setup FPGA PF with configuration information */
+ ret = fpga_5gnr_fec_configure(info->dev_name, &conf);
+ TEST_ASSERT_SUCCESS(ret,
+ "Failed to configure 5G FPGA PF for bbdev %s",
+ info->dev_name);
+ }
#endif
nb_queues = RTE_MIN(rte_lcore_count(), info->drv.max_num_queues);
nb_queues = RTE_MIN(nb_queues, (unsigned int) MAX_QUEUES);
/* Clear active devices structs. */
memset(active_devs, 0, sizeof(active_devs));
nb_active_devs = 0;
+
+ /* Disable interrupts */
+ intr_enabled = false;
}
static int
rte_atomic16_add(&tp->nb_dequeued, deq);
}
+static int
+throughput_intr_lcore_ldpc_dec(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_dec_op *ops[num_to_process];
+ struct test_buffers *bufs = NULL;
+ struct rte_bbdev_info info;
+ int ret, i, j;
+ struct rte_bbdev_dec_op *ref_op = tp->op_params->ref_dec_op;
+ uint16_t num_to_enq, enq;
+
+ bool loopback = check_bit(ref_op->ldpc_dec.op_flags,
+ RTE_BBDEV_LDPC_INTERNAL_HARQ_MEMORY_LOOPBACK);
+ bool hc_out = check_bit(ref_op->ldpc_dec.op_flags,
+ RTE_BBDEV_LDPC_HQ_COMBINE_OUT_ENABLE);
+
+ 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];
+
+ rte_atomic16_clear(&tp->processing_status);
+ rte_atomic16_clear(&tp->nb_dequeued);
+
+ while (rte_atomic16_read(&tp->op_params->sync) == SYNC_WAIT)
+ rte_pause();
+
+ ret = rte_bbdev_dec_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_ldpc_dec_op(ops, num_to_process, 0, bufs->inputs,
+ bufs->hard_outputs, bufs->soft_outputs,
+ bufs->harq_inputs, bufs->harq_outputs, ref_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) {
+ if (!loopback)
+ rte_pktmbuf_reset(
+ ops[i]->ldpc_dec.hard_output.data);
+ if (hc_out || loopback)
+ mbuf_reset(
+ ops[i]->ldpc_dec.harq_combined_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_ldpc_dec_ops(
+ tp->dev_id,
+ queue_id, &ops[enqueued],
+ num_to_enq);
+ } while (unlikely(num_to_enq != 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.
+ */
+ rte_atomic16_set(&tp->burst_sz, num_to_enq);
+
+ /* Wait until processing of previous batch is
+ * completed
+ */
+ while (rte_atomic16_read(&tp->nb_dequeued) !=
+ (int16_t) enqueued)
+ rte_pause();
+ }
+ if (j != TEST_REPETITIONS - 1)
+ rte_atomic16_clear(&tp->nb_dequeued);
+ }
+
+ return TEST_SUCCESS;
+}
+
static int
throughput_intr_lcore_dec(void *arg)
{
return TEST_SUCCESS;
}
+
+static int
+throughput_intr_lcore_ldpc_enc(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_enc_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];
+
+ rte_atomic16_clear(&tp->processing_status);
+ rte_atomic16_clear(&tp->nb_dequeued);
+
+ while (rte_atomic16_read(&tp->op_params->sync) == SYNC_WAIT)
+ rte_pause();
+
+ ret = rte_bbdev_enc_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_ldpc_enc_op(ops, num_to_process, 0,
+ bufs->inputs, bufs->hard_outputs,
+ tp->op_params->ref_enc_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)
+ rte_pktmbuf_reset(ops[i]->turbo_enc.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_ldpc_enc_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.
+ */
+ rte_atomic16_set(&tp->burst_sz, num_to_enq);
+
+ /* Wait until processing of previous batch is
+ * completed
+ */
+ while (rte_atomic16_read(&tp->nb_dequeued) !=
+ (int16_t) enqueued)
+ rte_pause();
+ }
+ if (j != TEST_REPETITIONS - 1)
+ rte_atomic16_clear(&tp->nb_dequeued);
+ }
+
+ return TEST_SUCCESS;
+}
+
static int
throughput_pmd_lcore_dec(void *arg)
{
if (test_vector.op_type == RTE_BBDEV_OP_TURBO_DEC)
throughput_function = throughput_intr_lcore_dec;
else if (test_vector.op_type == RTE_BBDEV_OP_LDPC_DEC)
- throughput_function = throughput_intr_lcore_dec;
+ throughput_function = throughput_intr_lcore_ldpc_dec;
else if (test_vector.op_type == RTE_BBDEV_OP_TURBO_ENC)
throughput_function = throughput_intr_lcore_enc;
else if (test_vector.op_type == RTE_BBDEV_OP_LDPC_ENC)
- throughput_function = throughput_intr_lcore_enc;
+ throughput_function = throughput_intr_lcore_ldpc_enc;
else
throughput_function = throughput_intr_lcore_enc;
offload_latency_empty_q_test_dec(uint16_t dev_id, uint16_t queue_id,
const uint16_t num_to_process, uint16_t burst_sz,
uint64_t *deq_total_time, uint64_t *deq_min_time,
- uint64_t *deq_max_time)
+ uint64_t *deq_max_time, const enum rte_bbdev_op_type op_type)
{
int i, deq_total;
struct rte_bbdev_dec_op *ops[MAX_BURST];
if (unlikely(num_to_process - deq_total < burst_sz))
burst_sz = num_to_process - deq_total;
- rte_bbdev_dequeue_dec_ops(dev_id, queue_id, ops, burst_sz);
+ if (op_type == RTE_BBDEV_OP_LDPC_DEC)
+ rte_bbdev_dequeue_ldpc_dec_ops(dev_id, queue_id, ops,
+ burst_sz);
+ else
+ rte_bbdev_dequeue_dec_ops(dev_id, queue_id, ops,
+ burst_sz);
deq_last_time = rte_rdtsc_precise() - deq_start_time;
*deq_max_time = RTE_MAX(*deq_max_time, deq_last_time);
offload_latency_empty_q_test_enc(uint16_t dev_id, uint16_t queue_id,
const uint16_t num_to_process, uint16_t burst_sz,
uint64_t *deq_total_time, uint64_t *deq_min_time,
- uint64_t *deq_max_time)
+ uint64_t *deq_max_time, const enum rte_bbdev_op_type op_type)
{
int i, deq_total;
struct rte_bbdev_enc_op *ops[MAX_BURST];
if (unlikely(num_to_process - deq_total < burst_sz))
burst_sz = num_to_process - deq_total;
- rte_bbdev_dequeue_enc_ops(dev_id, queue_id, ops, burst_sz);
+ if (op_type == RTE_BBDEV_OP_LDPC_ENC)
+ rte_bbdev_dequeue_ldpc_enc_ops(dev_id, queue_id, ops,
+ burst_sz);
+ else
+ rte_bbdev_dequeue_enc_ops(dev_id, queue_id, ops,
+ burst_sz);
deq_last_time = rte_rdtsc_precise() - deq_start_time;
*deq_max_time = RTE_MAX(*deq_max_time, deq_last_time);
return i;
}
+
#endif
static int
printf("== test: offload latency empty dequeue\ndev: %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)
+ if (op_type == RTE_BBDEV_OP_TURBO_DEC ||
+ op_type == RTE_BBDEV_OP_LDPC_DEC)
iter = offload_latency_empty_q_test_dec(ad->dev_id, queue_id,
num_to_process, burst_sz, &deq_total_time,
- &deq_min_time, &deq_max_time);
+ &deq_min_time, &deq_max_time, op_type);
else
iter = offload_latency_empty_q_test_enc(ad->dev_id, queue_id,
num_to_process, burst_sz, &deq_total_time,
- &deq_min_time, &deq_max_time);
+ &deq_min_time, &deq_max_time, op_type);
if (iter <= 0)
return TEST_FAILED;