Closed majorlin closed 5 days ago
@majorlin Thanks for finding this.
I don't see any good reason for this. It is a very old change in the library.
@majorlin I correct what I have just said. __SMLALD
requires two Q15 packed in an int32.
But this line is using *pSrcA++
to read a q15 from the tail of the vector.
So the instruction __SMLALD
cannot be used here.
https://github.com/ARM-software/CMSIS-DSP/blob/7236701c153c9bbdec00568a2821a39039988d8f/Source/BasicMathFunctions/arm_dot_prod_q15.c#L154