Teledyne SP Devices ADQ7, User Manual

Page: 8 / 13

Classification Revision
Public PA1
Document ID Print date
19-2241 2019-05-02
Condition 6 is recommended for minimum latency at buffer completion. See
gpu_streaming_defines.h
for macros to calculate correct buffer sizes from parameters. Please note that you GPU may impose
other restrictions on buffer size. The average transfer speed should also be considered:
Transfer speed [MB/s] = A-scans/s × record length × Ch × 10
− 6
If the maximum throughput of your system is exceeded more than momentarily the ADQs buffer may
overflow and data will be lost. Throughput measured in our test setup shown in Fig. 4 can be used as
a rough guideline for maximum throughput. GPU load may affect throughput so its recommended to
determine maximum throughput of your system with data processing active.
32 128 512 2048 8192 32768 131072 524288 2097152 8388608
5000
5200
5400
5600
5800
6000
6200
6400
6600
6800
7000
7200
1ch
2ch
Record length [samples]
M
iB
/s
Figure 4: Maximum throughput in test setup for different record lenghts.
4.2 B-scan and Data Valid
If no B-scan signal is connected the system will behave as if a correct B-scan is present, no writes to
data valid buffer should occur. If no data valid buffer is specified, please set data valid size to 0 in the call
to
SetupDMAP2p2D()
. This blocks any unintentional B-scan trigger from initializing a write to an invalid
address.
4.3 Set up P2P GPU with
SetupDMAP2p2D()
The input of marker addresses is vendor specific and described bellow, For general information see the
ADQAPI reference guide [ 1 ].
4.3.1 Nvidia
Marker memory is allocated by the ADQAPI. The marker address fields can be left empty.
ADQ7 GPU Peer-To-Peer – User Guide www.teledyne-spdevices.com Page 7 of 11