https://gcc.gnu.org/bugzilla/show_bug.cgi?id=66560
Bug ID: 66560
Summary: Fails to generate ADDSUBPS
Product: gcc
Version: 6.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: ubizjak at gmail dot com
Target Milestone: ---
The second testcase (bar_v4sf) fails to generate ADDSUBPS with -O2 -mavx.
--cut here--
typedef float v4sf __attribute__((vector_size(16)));
typedef int v4si __attribute__((vector_size(16)));
v4sf foo_v4sf (v4sf x, v4sf y)
{
v4sf tem0 = x - y;
v4sf tem1 = x + y;
return __builtin_shuffle (tem0, tem1, (v4si) { 0, 5, 2, 7 });
}
v4sf bar_v4sf (v4sf x, v4sf y)
{
v4sf tem0 = x + y;
v4sf tem1 = x - y;
return __builtin_shuffle (tem0, tem1, (v4si) { 4, 1, 6, 3 });
}
--cut here--
gcc -O2 -mavx:
vaddps %xmm1, %xmm0, %xmm2
vsubps %xmm1, %xmm0, %xmm0
vblendps $5, %xmm0, %xmm2, %xmm0
ret
The combine pass says:
foo_v4sf:
Successfully matched this instruction:
(set (reg:V4SF 93 [ D.2156 ])
(vec_merge:V4SF (plus:V4SF (reg/v:V4SF 91 [ x ])
(reg:V4SF 22 xmm1 [ y ]))
(minus:V4SF (reg/v:V4SF 91 [ x ])
(reg:V4SF 22 xmm1 [ y ]))
(const_int 10 [0xa])))
bar_v4sf:
Failed to match this instruction:
(set (reg:V4SF 93 [ D.2159 ])
(vec_merge:V4SF (minus:V4SF (reg/v:V4SF 91 [ x ])
(reg/v:V4SF 92 [ y ]))
(plus:V4SF (reg/v:V4SF 91 [ x ])
(reg/v:V4SF 92 [ y ]))
(const_int 5 [0x5])))
Please note that similar testcases always generate expected ADDSUBPD:
--cut here--
typedef double v4df __attribute__((vector_size(32)));
typedef long long v4di __attribute__((vector_size(32)));
v4df foo_v4df (v4df x, v4df y)
{
v4df tem0 = x - y;
v4df tem1 = x + y;
return __builtin_shuffle (tem0, tem1, (v4di) { 0, 5, 2, 7 });
}
v4df bar_v4df (v4df x, v4df y)
{
v4df tem0 = x + y;
v4df tem1 = x - y;
return __builtin_shuffle (tem0, tem1, (v4di) { 4, 1, 6, 3 });
}
--cut here--
foo_v4df:
Successfully matched this instruction:
(set (reg:V4DF 93 [ D.2162 ])
(vec_select:V4DF (vec_concat:V8DF (minus:V4DF (reg/v:V4DF 91 [ x ])
(reg/v:V4DF 92 [ y ]))
(plus:V4DF (reg/v:V4DF 91 [ x ])
(reg/v:V4DF 92 [ y ])))
(parallel [
(const_int 0 [0])
(const_int 5 [0x5])
(const_int 2 [0x2])
(const_int 7 [0x7])
])))
bar_v4df:
Successfully matched this instruction:
(set (reg:V4DF 93 [ D.2165 ])
(vec_select:V4DF (vec_concat:V8DF (minus:V4DF (reg/v:V4DF 91 [ x ])
(reg/v:V4DF 92 [ y ]))
(plus:V4DF (reg/v:V4DF 91 [ x ])
(reg/v:V4DF 92 [ y ])))
(parallel [
(const_int 0 [0])
(const_int 5 [0x5])
(const_int 2 [0x2])
(const_int 7 [0x7])
])))
