[subexp-daq] SIS3316 implementation in NURDLIB fixed

[Håkan T Johansson]

Dear Günter,

just had a quick look:

This old code with +1 is surely not good:

                bytes_to_read =
                    ((a_words_to_read * sizeof(uint32_t)) + 1) & ~0x7;

Since sizeof(uint32_t) is 4, the addition of 1 would not have any effect.

However, I'm a bit wondering about addint 8 (and 16 in the case of ~0xf).

How about the following:

                bytes_to_read =
                    ((a_words_to_read * sizeof(uint32_t)) + 0x7) & ~0x7;

(and '+ 0xf' with '& ~0xf') ?

That ought to bring it up to the next boundary if the read count was 
unaligned.  And the if-statements would not be needed.

Cheers,
Håkan




On Fri, 22 Mar 2024, Weber, Guenter Dr. wrote:

> 
> Dear friends,
> 
> 
> we just pushed a fixed version of the SIS3316 implementation. Compared to
> the original version of REBASING_SIS3316 a lot of glitches in the code were
> fixed in a first run a few days ago and now we finally also fixed a
> long-standing problem with the readout of the averaged traces.
> 
> 
> 
> 
> 
> Best greetings and have a nice weekend everybody
> 
> Günter
> 
> 
> 
> 
> 
>