I'm trying to use a TypedBufferArg to get at a small array of
uint64_ts, but I think the following snippet of code from that class is
tripping me up:
T &operator[](int i) { return ((T *)bufPtr)[i]; }
T in this case is uint64_t[3], and I think it has to be because the
amount of space allocated for the buffer is equal to sizeof(T). If I
want to be able to hold all the elements in the buffer, I can't for
instance make it a uint64_t *. In the line above, what happens is that
the bufPtr is cast to a T*, or in other words a (uint64_t[3]) *. That's
probably syntactically wrong, but you get the idea. Then the subscript
of i is applied, so for example if i was 2 you'd end up with bufPtr +
sizeof(uint64_t[3]) * 2 which is off the end of the buffer. The data
type you get out the end is also not useful because it's a uint64_t[3],
not a uint64_t like I would expect. I'm not sure what the right way to
implement what this code seems to be going for is, but I'm pretty sure
this is not it. Does anyone have a suggestion for how to fix it? Maybe
we need a new type called TypedBufferArgArray? If it's too obnoxious to
fix we should probably nuke that operator and I'll rework my code to do
without it.
Gabe
_______________________________________________
m5-dev mailing list
[email protected]
http://m5sim.org/mailman/listinfo/m5-dev
