While doing high throughput test on a BQL enabled NIC,
I found a very high cost in ndo_start_xmit() when accessing BQL data.
It turned out the problem was caused by compiler trying to be
smart, but involving a bad MESI transaction :
0.05 │ mov 0xc0(%rax),%edi // LOAD dql->num_queued
0.48 │ mov %edx,0xc8(%rax) // STORE dql->last_obj_cnt = count
58.23 │ add %edx,%edi
0.58 │ cmp %edi,0xc4(%rax)
0.76 │ mov %edi,0xc0(%rax) // STORE dql->num_queued += count
0.72 │ js bd8
I got an incredible 10 % gain [1] by making sure cpu do not attempt
to get the cache line in Shared mode, but directly requests for
ownership.
New code :
mov %edx,0xc8(%rax) // STORE dql->last_obj_cnt = count
add %edx,0xc0(%rax) // RMW dql->num_queued += count
mov 0xc4(%rax),%ecx // LOAD dql->adj_limit
mov 0xc0(%rax),%edx // LOAD dql->num_queued
cmp %edx,%ecx
The TX completion was running from another cpu, with high interrupts
rate.
Note that I am using barrier() as a soft hint, as mb() here could be
too heavy cost.
[1] This was a netperf TCP_STREAM with TSO disabled, but GSO enabled.
Signed-off-by: Eric Dumazet <edumazet@google.com>
Acked-by: Jesper Dangaard Brouer <brouer@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Implementation of dynamic queue limits (dql). This is a libary which
allows a queue limit to be dynamically managed. The goal of dql is
to set the queue limit, number of objects to the queue, to be minimized
without allowing the queue to be starved.
dql would be used with a queue which has these properties:
1) Objects are queued up to some limit which can be expressed as a
count of objects.
2) Periodically a completion process executes which retires consumed
objects.
3) Starvation occurs when limit has been reached, all queued data has
actually been consumed but completion processing has not yet run,
so queuing new data is blocked.
4) Minimizing the amount of queued data is desirable.
A canonical example of such a queue would be a NIC HW transmit queue.
The queue limit is dynamic, it will increase or decrease over time
depending on the workload. The queue limit is recalculated each time
completion processing is done. Increases occur when the queue is
starved and can exponentially increase over successive intervals.
Decreases occur when more data is being maintained in the queue than
needed to prevent starvation. The number of extra objects, or "slack",
is measured over successive intervals, and to avoid hysteresis the
limit is only reduced by the miminum slack seen over a configurable
time period.
dql API provides routines to manage the queue:
- dql_init is called to intialize the dql structure
- dql_reset is called to reset dynamic values
- dql_queued called when objects are being enqueued
- dql_avail returns availability in the queue
- dql_completed is called when objects have be consumed in the queue
Configuration consists of:
- max_limit, maximum limit
- min_limit, minimum limit
- slack_hold_time, time to measure instances of slack before reducing
queue limit
Signed-off-by: Tom Herbert <therbert@google.com>
Acked-by: Eric Dumazet <eric.dumazet@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Eric Dumazet