https://gcc.gnu.org/bugzilla/show_bug.cgi?id=72747
Bug ID: 72747
Summary: powerpc: wrong code generated for vec_splats in
cascading assignment
Product: gcc
Version: 5.3.1
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: wschmidt at gcc dot gnu.org
Target Milestone: ---
A statement such as "v = vec_splats (1);" correctly initializes a vector.
However, a statement such as "v[1] = v[0] = vec_splats (1);" initializes both
v[1] and v[0] to random garbage. The following code can be used to verify the
problem:
#include <altivec.h>
#include <iostream>
#include <cstdlib>
using std::cout;
using std::endl;
int main (int argc, char *argv[])
{
int i, j;
__vector short v1[2];
v1[1] = v1[0] = vec_splats ((short) 1);
for (i = 0; i < 2; ++i)
{
cout << "v1[" << i << "]: ";
for (j = 0; j < 8; ++j)
cout << v1[i][j] << " ";
cout << endl;
}
return EXIT_SUCCESS;
}
Probable cause is the special handling of vec_splats in rs6000-c.c during
parsing. Compiled at -O0 with current trunk, the "optimized" dump shows:
<bb 2>:
_1 = v1[0];
v1[1] = _1;
where v1[0] is uninitialized.
Reported on GCC 5.3.1, but confirmed to occur also in GCC 6.1 and current
trunk.