On Tue, May 20, 2014 at 2:22 PM, Seima Rao <seimarao@xxxxxxxxx> wrote:
> Hi,
>
> When compiling a simple matrix dot product example on both
> alpha & ia64 using -O3 and -fmodulo-sched options,
> I am unable to find a software pipelined loop in the
> generated assembly.
The following is the C++ source code for which sw pipelining is not achieved
-------------------------------------------------------------------------------------------------------
// file vectorproduct.h
#include <cstddef>
using namespace std;
template<
class T,
size_t N,
T A__[N],
const T B__[N],
const T C__[N],
const T D__[N]
>
void
vectorproduct(void)
{
for(size_t i = 0; i < N; ++i) {
A__[i] = B__[i] * C__[i] + D__[i];
}
}
// file vectorproduct.int.cpp
#include "vectorproduct.h"
int A[1024], B[1024], C[1024], D[1024];
extern void
foo (
void
)
{
vectorproduct<int, 1024, A, B, C, D>();
}
Sincerely,
Seima Rao.
