.. index::
    Nanos6 FPGA Architecture configuration

.. This file was automatically generated after a push to master branch of https://github.com/bsc-pm-ompss-at-fpga/nanos6-fpga git repository

.. _nanos6_fpga_config:

Nanos6 FPGA Architecture configuration
::::::::::::::::::::::::::::::::::::::

The Nanos6 behavior can be tuned with different configuration options. They are summarized and briefly described in the Nanos6 default configuration file as well as in the `OmpSs-2 user guide <https://pm.bsc.es/ftp/ompss-2/doc/user-guide/nanos6/index.html>`_, the FPGA architecture section is shown below:

.. code:: bash

	[devices]
		directory = true
		[devices.fpga]
			# Enable/disable the reverse offload service
			reverse_offload = false
			# Byte alignment of the fpga memory allocations
			alignment = 16
			# If xtasks supports async copies, it can be "async", if not, the runtime can use the default xtasks memcpy and
			# simulate an asynchronous copy spawning a new thread with "forced async". Copies can also be synchronous with "sync".
			mem_sync_type = "sync"
			page_size = 0x8000
			requested_fpga_memory = 0x40000000
			# Enable FPGA device service threads. It is useful to disable them when using the broadcaster, because in that case the 
			# FPGAs are handled by the broadcaster device service and the FPGA services are not used.
			enable_services = true
			# Maximum number of FPGA tasks running at the same time
			streams = 16
			[devices.fpga.polling]
				# Indicate whether the FPGA services should constantly run while there are FPGA tasks
				# running on their FPGA. Enabling this option may reduce the latency of processing FPGA
				# tasks at the expenses of occupying a CPU from the system. Default is true
				pinned = true
				# The time period in microseconds between FPGA service runs. During that time, the CPUs
				# occupied by the services are available to execute ready tasks. Setting this option to 0
				# makes the services to constantly run. Default is 1000
				period_us = 1000