Sent from my iPhone
On May 14, 2010, at 2:18 AM, "tkoenig at gcc dot gnu dot org" <gcc-bugzi...@gcc.gnu.org
> wrote:
This code leads to the adding of 0.0, which is a nop. Any
signalling should have been done previously.
It is not signalling that matters here but signed zero. 0.0 + -0.0 ==
0.0. So without the 0.0 +, you can get a negative zero.
i...@linux-fd1f:/tmp> cat mult.f90
subroutine foo(a,b,c)
real, intent(in) :: a,b
real, intent(out) :: c
c = 0.0
c = c + a*b
end subroutine foo
i...@linux-fd1f:/tmp> gfortran -O3 -fdump-tree-optimized -S mult.f90
i...@linux-fd1f:/tmp> cat mult.f90.142t.optimized
;; Function foo (foo_)
foo (real(kind=4) & restrict a, real(kind=4) & restrict b, real
(kind=4) &
restrict c)
{
real(kind=4) D.1542;
real(kind=4) D.1541;
real(kind=4) D.1540;
real(kind=4) D.1539;
<bb 2>:
D.1539_4 = *a_3(D);
D.1540_6 = *b_5(D);
D.1541_7 = D.1539_4 * D.1540_6;
D.1542_8 = D.1541_7 + 0.0;
*c_1(D) = D.1542_8;
return;
}
i...@linux-fd1f:/tmp> cat mult.s
.file "mult.f90"
.text
.p2align 4,,15
.globl foo_
.type foo_, @function
foo_:
.LFB0:
movss (%rdi), %xmm0
mulss (%rsi), %xmm0
addss .LC0(%rip), %xmm0
movss %xmm0, (%rdx)
ret
.LFE0:
.size foo_, .-foo_
.section .rodata.cst4,"aM",@progbits,4
.align 4
.LC0:
.long 0
.section .eh_frame,"a",@progbits
.Lframe1:
.long .LECIE1-.LSCIE1
.LSCIE1:
.long 0
.byte 0x1
.string "zR"
.uleb128 0x1
.sleb128 -8
.byte 0x10
.uleb128 0x1
.byte 0x3
.byte 0xc
.uleb128 0x7
.uleb128 0x8
.byte 0x90
.uleb128 0x1
.align 8
.LECIE1:
.LSFDE1:
.long .LEFDE1-.LASFDE1
.LASFDE1:
.long .LASFDE1-.Lframe1
.long .LFB0
.long .LFE0-.LFB0
.uleb128 0
.align 8
.LEFDE1:
.ident "GCC: (GNU) 4.6.0 20100513 (experimental)"
.section .note.GNU-stack,"",@progbits
--
Summary: Unneeded +0.0 for c = 0.0 ; c = c+ a*b
Product: gcc
Version: 4.6.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: enhancement
Priority: P3
Component: middle-end
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: tkoenig at gcc dot gnu dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=44134
