suppose I have in a file type.h
#ifdef DTYPE==FP16
typedef half Dtype;
#else
typdef float Dtype;
#endif
I have a function declaration in test.h
#include "type.h"
Dtype fill_data(Dtype *buff, Dtype alpha, int N);
and definition in test.c
#include "test.h"
Dtype fill_data(Dtype *buff, Dtype alpha, int N)
{
int i;
for (i = 0; i < N; ++i) X[i*] = ALPHA;
}
Now, another file uses that fill_data function and it is like this.
Dtype *buf1, alpha;
int n;
...
file_data(buf1, 1, n);
The second argument is integer 1 and the g++ complains it can't convert int to Dtype. When the code was declared and defined for float (instead of Dtype), g++ converted int to float using implicit conversion.
Can we let g++ do this implicit conversion when the argument is not a built-in type by specifying any compile option? Of course compiler error is removed if I used (Dtype)1 instead of 1.
question from:
https://stackoverflow.com/questions/65931490/can-we-make-automatic-type-casting-possible-in-c-for-function-argument-when-t 与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…