Commit 0e9da345 authored by Alejandro Homs Puron's avatar Alejandro Homs Puron Committed by Operator for ID00
Browse files

Doc: update device configuration for passive receiver

parent 525ac667
......@@ -1719,20 +1719,18 @@ Add LimaCCDs and SlsDetector class devices.
+----------------------------------------------------------+------------------------------------------------------+
| id10/limaccds/eiger500k->LimaCameraType | SlsDetector |
+----------------------------------------------------------+------------------------------------------------------+
| id10/limaccds/eiger500k->NbProcessingThread | 23 |
| id10/limaccds/eiger500k->NbProcessingThread | 11 |
+----------------------------------------------------------+------------------------------------------------------+
| id10/limaccds/eiger500k->BufferMaxMemory | 30 |
| id10/limaccds/eiger500k->BufferMaxMemory | 25 |
+----------------------------------------------------------+------------------------------------------------------+
| LimaCCDs/eiger500k/DEVICE/SlsDetector | id10/slsdetector/eiger500k |
+----------------------------------------------------------+------------------------------------------------------+
| id10/slsdetector/eiger500k->config_fname | /users/opid00/eiger/eiger_v3.1.1/config/ |
| | beb-021-020-direct-FO-10g.config |
+----------------------------------------------------------+------------------------------------------------------+
| id10/slsdetector/eiger500k->pixel_depth_cpu_affinity_map | | { 4: (((((CPU( 6), CPU(18)), (CPU( 7), CPU(19))), |
| | (CPU(12), CPU(13), CPU(14))), |
| | | (((CPU( 9), CPU(21)), (CPU(10), CPU(22))), |
| | (CPU(15), CPU(16), CPU(17)))), |
| | | CPU(*(range( 1, 6) + range(12, 18))), |
| id10/slsdetector/eiger500k->pixel_depth_cpu_affinity_map | | { 4: ((((CPU( 6), CPU( 7)), (CPU(18), CPU(19))), |
| | | ((CPU( 9), CPU(10)), (CPU(21), CPU(22)))), |
| | | CPU(*chain(range(1, 6), range(12, 18))), |
| | | CPU(0), |
| | | (('eth0,eth1,eth2,eth4,eth6,eth7,eth8,eth9', |
| | {-1: (CPU(0), CPU(0))}), |
......@@ -1740,36 +1738,27 @@ Add LimaCCDs and SlsDetector class devices.
| | {-1: (CPU( 8), CPU(20))}), |
| | | ('eth5', |
| | {-1: (CPU(11), CPU(23))}))), |
| | | 8: (((((CPU( 6), CPU(18)), (CPU( 7), CPU(19))), |
| | (CPU(18), CPU(19), CPU(20))), |
| | | (((CPU( 9), CPU(21)), (CPU(10), CPU(22))), |
| | (CPU(21), CPU(22), CPU(23)))), |
| | | CPU(*(range( 1, 6) + range(12, 18))), |
| | | CPU(0), |
| | | (('eth0,eth1,eth2,eth4,eth6,eth7,eth8,eth9', |
| | {-1: (CPU(0), CPU(0))}), |
| | | ('eth3', |
| | {-1: (CPU( 8), CPU(20))}), |
| | | ('eth5', |
| | {-1: (CPU(11), CPU(23))}))), |
| | | 16: '@8', |
| | | 32: '@8'} |
| | | 8: '@4', |
| | | 16: '@4', |
| | | 32: '@4'} |
+----------------------------------------------------------+------------------------------------------------------+
.. note:: in order to perform high frame rate acquisitions, the CPU affinity must be fixed for
the following tasks:
* Receivers ports (2x2=4): listeners, writers
* Eiger processing threads: 3 per receiver, configurable
* Receivers' packet stream threads (passive mode equivalent to listeners): 2
* Eiger processing threads: 2 per receiver, configurable
* Lima processing threads
* Other OS processes
* Net-dev group packet dispatching for Rx queues: irq & processing
The previous example is based on a dual 6-core CPUs backend with *Hyper-Threading Technology* (12 cores,
24 threads). After the data acquisition finishes the Lima processing threads will run also on the CPUs
assigned to listeners and writers (0xfffffe), that is 23 cores in total, which is used for setting the
NbProcessingThreads. Please note that there are three network groups and four pixel_depth->cpu_affinity
settings (4-, 8-, 16- and 32-bit). The special global_affinity '@X' is a reference to pixel_depth X.
24 threads). The configuration is optimised to work in pipe-line mode: cores/threads in Socket #1
(6-11, 18-23) will perform network data acquisition and basic image reconstruction (4-to-8 bit conversion,
gap pixel insertion), and cores/threads in Socket #0 (1-5, 12-17) will perform Lima processing tasks.
That's why the NbProcessingThread is set to 11. Please note that there are three network groups and four
pixel_depth->cpu_affinity settings (4-, 8-, 16- and 32-bit). The special global_affinity '@X' is a reference
to pixel_depth X.
.. note:: The Intel 10 Gigabit Ethernet Server Adapter has multiple hardware FIFOs per port, called
queues in the OS terminology. The hardware uses a hash algorithm to dispatch packets into the active
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment