Hi experts,
Currently I'm working on optimization of some audio codec on Cortex-A8, I have coded some functions with Neon Intrinsics and some in asm. when I tested on Rvds4.0 simulator every thing is fine but when I tested on board I'm seeing some bit exactness issues in generated outputs. Original C code outputs are fine on board. I have taken care of un-initialised variables, memory leaks. for device testing I'm using make file with arm gcc 4.4.3
Did you find the solution to this problem ?
One more thing I could think of that needs verifying: volatile on global variables that you're using from within an interrupt.
... I can't even say rounding, because normally you would round towards zero, in order to avoid overflows.
But I can imagine that the simulator would maybe not catch a volatile problem. Try disabling optimization using -O0 and see if it makes any difference.