Part Number: TDA4VL-Q1
hello,TI engineer.
Recently, I was optimizing an algorithm on the c7x processor. The optimization methods I use include the following:
1. Run -O2 or -O3 optimization and start the software pipeline. The effect is remarkable.
2. Use the c6x port and c7x inline instructions to replace the arithmetic operation, the effect is not obvious.
3. Use loose floating point calculation mode, no effect. Because I didn't do a double in my algorithm.
4. When SE, SA, and vector data types are used for parallel computation, the algorithm performance deteriorates. Because the data structure of the algorithm does not have continuous memory, this calculation will require a lot of IO operations.
5. Set some variables that repeatedly participate in the calculation to the register type, basically no promotion.
In addition, can you provide me with some good optimization directions?
