StrawData: inefficient read implementations

[JonasKunze]

Currently the StrawData constructor is reading the raw data fields into separate class variables (m_packetLength = readPacketLength()). The read methods are implemented in a very inefficient way (many branches instead of simple copy instructions). One should use a packed struct to directly access the fields. Something like following should do (I don't know the actual format, just guessing based on your code):

struct STRAW_DATA_HDR {
    uint16_t packetLength;
    uint8_t triggerTypeWord;
    uint8_t chamber:2;
    uint8_t view:2;
    uint8_t halfView:2;
    uint8_t reserved:2;
}__attribute__ ((__packed__));

To access the fields one can do following:

(STRAW_DATA_HDR*) hdr = static_cast<STRAW_DATA_HDR*>(dataPointer);
m_packetLength = hdr->packetLength;

This way you have one single instruction (but two lines of code) to read the packetLength field.

Instead of reading all fields in the constructor one could check if directly accessing the packed struct is really slower then caching them. At least one should use larger data types like uint_fast8_t if caching is useful for class variables that are used very often.

Another way would be to implement the read methods with something like following (untested):
m_L0TriggerDecision = static_cast<uint8_t>(dataPointer+2);