[dpdk-dev] [PATCH v4 4/4] app/testpmd: use per-core variable in flowgen

[Zhihong Wang <wangzhihong.wzh@bytedance.com>]

Use per-core variable for flow indexing to solve cache contention in
multi-core scenarios.

v4: use loop local variable to improve performance

Signed-off-by: Zhihong Wang <wangzhihong.wzh@bytedance.com>
---
 app/test-pmd/flowgen.c | 6 +++++-
 1 file changed, 5 insertions(+), 1 deletion(-)

diff --git a/app/test-pmd/flowgen.c b/app/test-pmd/flowgen.c
index 229794ee9c..b541485304 100644
--- a/app/test-pmd/flowgen.c
+++ b/app/test-pmd/flowgen.c
@@ -53,6 +53,8 @@ static struct rte_ether_addr cfg_ether_dst =
 
 #define IP_DEFTTL  64   /* from RFC 1340. */
 
+RTE_DEFINE_PER_LCORE(int, _next_flow);
+
 /*
  * Multi-flow generation mode.
  *
@@ -80,7 +82,7 @@ pkt_burst_flow_gen(struct fwd_stream *fs)
 	uint32_t retry;
 	uint64_t tx_offloads;
 	uint64_t start_tsc = 0;
-	static int next_flow = 0;
+	int next_flow = RTE_PER_LCORE(_next_flow);
 
 	get_start_cycles(&start_tsc);
 
@@ -193,6 +195,8 @@ pkt_burst_flow_gen(struct fwd_stream *fs)
 		} while (++nb_tx < nb_pkt);
 	}
 
+	RTE_PER_LCORE(_next_flow) = next_flow;
+
 	get_end_cycles(fs, start_tsc);
 }
 
-- 
2.11.0