slp: support complex FMA and complex FMA conjugate
This adds support for FMA and FMA conjugated to the slp pattern matcher.
Example of instructions matched:
#include <stdio.h>
#include <complex.h>
#define N 200
#define ROT
#define TYPE float
#define TYPE2 float
void g (TYPE2 complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
{
for (int i=0; i < N; i++)
{
c[i] += a[i] * (b[i] ROT);
}
}
void g_f1 (TYPE2 complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
{
for (int i=0; i < N; i++)
{
c[i] += conjf (a[i]) * (b[i] ROT);
}
}
void g_s1 (TYPE2 complex a[restrict N], TYPE complex b[restrict N], TYPE complex c[restrict N])
{
for (int i=0; i < N; i++)
{
c[i] += a[i] * conjf (b[i] ROT);
}
}
void caxpy_add(double complex * restrict y, double complex * restrict x, size_t N, double complex f) {
for (size_t i = 0; i < N; ++i)
y[i] += x[i]* f;
}
gcc/ChangeLog:
* internal-fn.def (COMPLEX_FMA, COMPLEX_FMA_CONJ): New.
* optabs.def (cmla_optab, cmla_conj_optab): New.
* doc/md.texi: Document them.
* tree-vect-slp-patterns.c (vect_match_call_p,
class complex_fma_pattern, vect_slp_reset_pattern,
complex_fma_pattern::matches, complex_fma_pattern::recognize,
complex_fma_pattern::build): New.