[ovs-dev] [PATCH v5 5/6] dpif-netdev: Change pmd selection order.

[Darrell Ball]

    On 8/23/17, 6:34 AM, "Kevin Traynor" <ktraynor at redhat.com> wrote:
    
        Up to his point rxqs are sorted by processing cycles they
        consumed and assigned to pmds in a round robin manner.
        
        Ian pointed out that on wrap around the most loaded pmd will be
        the next one to be assigned an additional rxq and that it would be
        better to reverse the pmd order when wraparound occurs.
        
        In other words, change from assigning by rr to assigning in a forward
        and reverse cycle through pmds.
        
        Also, now that the algothim has finalised, document an example.

Typo: algothim
        
        Suggested-by: Ian Stokes <ian.stokes at intel.com>
        Signed-off-by: Kevin Traynor <ktraynor at redhat.com>
        ---
         Documentation/howto/dpdk.rst | 16 ++++++++++++++++
         lib/dpif-netdev.c            | 21 ++++++++++++++++++++-
         tests/pmd.at                 |  2 +-
         3 files changed, 37 insertions(+), 2 deletions(-)
        
        diff --git a/Documentation/howto/dpdk.rst b/Documentation/howto/dpdk.rst
        index 44737e4..493e215 100644
        --- a/Documentation/howto/dpdk.rst
        +++ b/Documentation/howto/dpdk.rst
        @@ -124,4 +124,20 @@ will be used where known to assign rxqs with the highest consumption of
         processing cycles to different pmds.
         
        +For example, in the case where here there are 5 rxqs and 3 cores (e.g. 3,7,8)
        +available, and the measured usage of core cycles per rxq over the last
        +interval is seen to be:
        +
        +- Queue #0: 30%
        +- Queue #1: 80%
        +- Queue #3: 60%
        +- Queue #4: 70%
        +- Queue #5: 10%
        +
        +The rxqs will be assigned to cores 3,7,8 in the following order:
        +
        +Core 3: Q1 (80%) |
        +Core 7: Q4 (70%) | Q5 (10%)
        +core 8: Q0 (60%) | Q0 (30%)

[Darrell]
There is a typo in the example:
        +core 8: Q0 (60%) …..
should be
        +core 8: Q3 (60%) …..



        +
         Rxq to pmds assignment takes place whenever there are configuration changes.
         
        diff --git a/lib/dpif-netdev.c b/lib/dpif-netdev.c
        index afbf591..6cc0a1e 100644
        --- a/lib/dpif-netdev.c
        +++ b/lib/dpif-netdev.c
        @@ -3285,4 +3285,5 @@ struct rr_numa {
         
             int cur_index;
        +    bool idx_inc;
         };
         
        @@ -3341,4 +3342,7 @@ rr_numa_list_populate(struct dp_netdev *dp, struct rr_numa_list *rr)
                 numa->pmds = xrealloc(numa->pmds, numa->n_pmds * sizeof *numa->pmds);
                 numa->pmds[numa->n_pmds - 1] = pmd;
        +        /* At least one pmd so initialise curr_idx and idx_inc. */
        +        numa->cur_index = 0;
        +        numa->idx_inc = true;
             }
         }
        @@ -3347,5 +3351,20 @@ static struct dp_netdev_pmd_thread *
         rr_numa_get_pmd(struct rr_numa *numa)
         {
        -    return numa->pmds[numa->cur_index++ % numa->n_pmds];
        +    int numa_idx = numa->cur_index;
        +
        +    if (numa->idx_inc == true) {
        +        if (numa->cur_index == numa->n_pmds-1) {
        +            numa->idx_inc = false;
        +        } else {
        +            numa->cur_index++;
        +        }
        +    } else {

Maybe one sentence comment for others benefit?

        +        if (numa->cur_index == 0) {
        +            numa->idx_inc = true;
        +        } else {
        +            numa->cur_index--;
        +        }
        +    }
        +    return numa->pmds[numa_idx];
         }
         
        diff --git a/tests/pmd.at b/tests/pmd.at
        index b6732ea..e39a23a 100644
        --- a/tests/pmd.at
        +++ b/tests/pmd.at
        @@ -54,5 +54,5 @@ m4_define([CHECK_PMD_THREADS_CREATED], [
         
         m4_define([SED_NUMA_CORE_PATTERN], ["s/\(numa_id \)[[0-9]]*\( core_id \)[[0-9]]*:/\1<cleared>\2<cleared>:/"])
        -m4_define([SED_NUMA_CORE_QUEUE_PATTERN], ["s/\(numa_id \)[[0-9]]*\( core_id \)[[0-9]]*:/\1<cleared>\2<cleared>:/;s/\(queue-id: \)0 2 4 6/\1<cleared>/;s/\(queue-id: \)1 3 5 7/\1<cleared>/"])
        +m4_define([SED_NUMA_CORE_QUEUE_PATTERN], ["s/\(numa_id \)[[0-9]]*\( core_id \)[[0-9]]*:/\1<cleared>\2<cleared>:/;s/\(queue-id: \)1 2 5 6/\1<cleared>/;s/\(queue-id: \)0 3 4 7/\1<cleared>/"])
         m4_define([DUMMY_NUMA], [--dummy-numa="0,0,0,0"])
         
        -- 
        1.8.3.1