Question about changing ACT from DBF_LONG to DBF_DOUBLE

[nariox]

Hi Mark,
I'm wondering if we could switch ACT (and data/VAL) to DBF_DOUBLE/DBF_UINT64 the definition . Double would allow us to get counts all the way to 2^53 before starting to lose precision. I see the Amptek driver already computes the counts in the readFloat64 function, so it seems like think would make sense for it to be a DBF_DOUBLE instead.

The main reason for me to change this is that we have a device that will see a high rate of counts (maybe up to 1M events/s), so that gives us about 36 mins before the value overflows. It is understandably longer than what we'd use, but I just want to avoid having values overflow. If precision is a concern, we could set it so that it alarms if it goes beyond 2^53 (although that'd take 200+ years at 1Mevents/s).

Do you foresee any issues with this change?

[MarkRivers]

I think it would be no problem to change ACT to be DBF_DOUBLE. It gets its value from pstatus->totalCounts, which is a double.

mca/mcaApp/mcaSrc/mcaRecord.c

Line 703 in 46848fd

pmca->act = pstatus->totalCounts;

However, changing the datatype of the VAL field is much more difficult. There is actually an FTVL field that lets you select the data type of the VAL field, just like for the waveform record. However, the asyn interface to device support does not support this. It hardcodes the datatype of the array data to be int or epicsInt32.

mca/mcaApp/mcaSrc/devMcaAsyn.c

Line 63 in 46848fd

int *data;

mca/mcaApp/mcaSrc/devMcaAsyn.c

Line 112 in 46848fd

pPvt->data = callocMustSucceed(pmca->nmax, sizeof(epicsInt32),

mca/mcaApp/mcaSrc/devMcaAsyn.c

Line 357 in 46848fd

pPvt->pasynInt32Array->read(pPvt->asynInt32ArrayPvt, pasynUser,

All drivers would need to be changed to allow the data type to be either variable, or DBF_DOUBLE, or DBF_UINT64.