[Bug c++/61292] New: auto keyword to vector reference generates wrong alignment move
vincenzo.innocente at cern dot ch
gcc-bugzilla@gcc.gnu.org
Fri May 23 10:05:00 GMT 2014
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=61292
Bug ID: 61292
Summary: auto keyword to vector reference generates wrong
alignment move
Product: gcc
Version: 4.9.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: c++
Assignee: unassigned at gcc dot gnu.org
Reporter: vincenzo.innocente at cern dot ch
given
typedef float __attribute__( ( vector_size( 16 ) ) ) float32x4_t;
typedef float __attribute__( ( vector_size( 16 ) , aligned(4) ) )
float32x4a4_t;
void add11(float * x, float y, float32x4_t v) {
auto & k1 = *(float32x4a4_t*)(x);
auto & k2 = *(float32x4a4_t*)(x);
k1 +=v;
k2 += k1+v;
}
void add98(float * x, float y, float32x4_t v) {
float32x4a4_t & k1 = *(float32x4a4_t*)(x);
float32x4a4_t & k2 = *(float32x4a4_t*)(x);
k1 +=v;
k2 += k1+v;
}
I get
c++ -std=c++1y -Ofast -march=haswell -S extNotAlign.cc; cat extNotAlign.s
__Z5add11PffU8__vectorf:
LFB4:
vaddps (%rdi), %xmm1, %xmm0
vaddps %xmm0, %xmm0, %xmm0
vaddps %xmm1, %xmm0, %xmm1
vmovaps %xmm1, (%rdi)
ret
LFE4:
.section __TEXT,__text_cold,regular,pure_instructions
LCOLDE4:
.text
LHOTE4:
.section __TEXT,__text_cold,regular,pure_instructions
LCOLDB5:
.text
LHOTB5:
.align 4,0x90
.globl __Z5add98PffU8__vectorf
__Z5add98PffU8__vectorf:
LFB5:
vaddps (%rdi), %xmm1, %xmm0
vaddps %xmm0, %xmm0, %xmm0
vaddps %xmm1, %xmm0, %xmm1
vmovups %xmm1, (%rdi)
ret
second of course is correct (vmovups not vmovaps)
More information about the Gcc-bugs
mailing list