accel-config config-wq <device name>/<wq name> [<options>]
accel-config config-wq dsa0/wq0.0 --wq-size=16
- -g, --group-id=
specify the group id used by work queue, group id should be between 0 and the maximum number of groups per device. A group id of -1 disassociates the work queue from a group.
- -s, --wq-size=
specify work queue size used by a work queue. The wq-size should be between 0 and the maximum workqueue size exposed by the device in the max_work_queue_size attribute under device.
- -p, --priority
specify the priority of this work queue relative to other work queues in the same group. This field ranges from 1 to 15 where the larger value means higher priority. For example, if wq1 has a priority of 6 and wq2 has priority of 2, wq1 will issue 3 times as many descriptors to the engine compare to wq2.
- -b, --block-on-fault
toggle block on fault enable for the work queue. block-on-fault should be either 0 or 1. If block on fault is disabled, if a page fault occurs on a source or destination memory access, the operation stops and the page fault is reported to the software.
- -t, --threshold
configure the threshold for the wq. The set up threshold value should be larger than 0 and only for "shared" wq, and note that the configured threshold value should be at least 1 less than the configured wq-size value. The reason is threshold is set for unprivileged users, and need to reserve at least 1 space for the privileged user for commands.
- -y, --type
configure the type for the wq. The wq type can be "kernel" for kernel user case to request wq, "mdev" for mediated device driver use case to request wq or "user" for regular character device driver use case to wq.
- -n, --name
configure the name for the wq. This can be any string defined by the user. Typically it is used for identifying the wq in some unique way. A possible suggestion would be using a uuid string.
- -m, --mode
configure the mode for the wq. The wq mode can be set as "dedicated" or "shared".
- -c, --max-batch-size
specify the max batch size used by a work queue. The value should be between 1 and the device maximum batch size found in the max_batch_size attribute under device. If the value is not a power of 2, it will be rounded up to the next power of 2.
- -x, --max-transfer-size
specify the max transfer size used by a work queue. The value should be between 1 and the device maximum transfer size found in the max_transfer_size attribute under device. If the value is not a power of 2, it will be rounded up to the next power of 2.
Copyright (c) 2016 - 2019, Intel Corporation. License GPLv2: GNU GPL version 2 http://gnu.org/licenses/gpl.html. This is free software: you are free to change and redistribute it. There is NO WARRANTY, to the extent permitted by law.