Hello!
For our project, we are using the TDA4 dev board with on it the Ti C71x DSP.
We are currently in the process of writing kernels to run our algorithms on the C71x. The algorithms rely heavy on look up tables, that's why we investigated to usability of the C71x's LUT intrinsics/acceleration. As far as we are aware off, the C71x can look-up in parallel 16 different LUT's using the on-board LUT acceleration. This on its own is off course a very powerful feature, but not what we would need. Our algorithm uses just 4 look-up table's and we would like to query ONE LUT in a SIMD way, so for example looking up 16 values in ONE LUT with one instruction (vectorized).
So far the available documentation could not provide us with the information we need so thats why i'm posting this on the forums. Would it be possible to do a SIMD (uint16) look-up to just one LUT ?
Many thanks in advance!
Chris