SUNphi
1.0
|
CANCEL_DUPLICATED_NNARY_SMET_CALL
why can't we make only const & on rhs?
we need to enforce SmET
add costRead
add costEval
add isAliasing
the array must be replaced with a tuple, whose types must be deduced when instatiating the struct, such that int or long int or whatever is appropriately used!
the array must be replaced with a tuple, whose types must be deduced when instatiating the struct, such that int or long int or whatever is appropriately used!
the array must be replaced with a tuple, whose types must be deduced when instatiating the struct, such that int or long int or whatever is appropriately used!
the array must be replaced with a tuple, whose types must be deduced when instatiating the struct, such that int or long int or whatever is appropriately used!
the array must be replaced with a tuple, whose types must be deduced when instatiating the struct, such that int or long int or whatever is appropriately used!
the array must be replaced with a tuple, whose types must be deduced when instatiating the struct, such that int or long int or whatever is appropriately used!
the array must be replaced with a tuple, whose types must be deduced when instatiating the struct, such that int or long int or whatever is appropriately used!
the array must be replaced with a tuple, whose types must be deduced when instatiating the struct, such that int or long int or whatever is appropriately used!
the array must be replaced with a tuple, whose types must be deduced when instatiating the struct, such that int or long int or whatever is appropriately used!
Add the possibility that the accumulated size is more than needed, so we can split the outermost component to allow for other kind of optimization
Include a variation of the vectorization according to the kind of SIMD vector