[Bug target/48607] New: Inefficient complex float parameter passing

[Bug target/48607] New: Inefficient complex float parameter passing

[hjl.tools at gmail dot com]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48607

           Summary: Inefficient complex float parameter passing
           Product: gcc
           Version: 4.7.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned@gcc.gnu.org
        ReportedBy: hjl.tools@gmail.com
                CC: ubizjak@gmail.com


[hjl@gnu-6 pr1000]$ cat z.i 
typedef _Complex float SCtype;
SCtype
foo (SCtype x1, SCtype x2, SCtype x3, SCtype x4, SCtype x5, SCtype x6,
     SCtype x7, SCtype x8, SCtype a, SCtype b)
{
  return x2;
}
[hjl@gnu-6 pr1000]$ /usr/gcc-4.7/bin/gcc -O2 -S z.i
[hjl@gnu-6 pr1000]$ cat z.s
    .file    "z.i"
    .text
    .p2align 4,,15
    .globl    foo
    .type    foo, @function
foo:
.LFB0:
    .cfi_startproc
    movq    %xmm1, -16(%rsp)
    movl    -16(%rsp), %eax
    movl    %eax, -72(%rsp)
    movl    -12(%rsp), %eax
    movl    %eax, -68(%rsp)
    movq    -72(%rsp), %xmm0
    ret
    .cfi_endproc
.LFE0:
    .size    foo, .-foo
    .ident    "GCC: (GNU) 4.7.0 20110406 (experimental) [trunk revision
172062]"
    .section    .note.GNU-stack,"",@progbits
[hjl@gnu-6 pr1000]$ 

We should simply do

movaps    %xmm1, %xmm0  

or

movq   %xmm1,%xmm0

[Bug target/48607] Inefficient complex float parameter passing

[rguenth at gcc dot gnu.org]

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=48607

Richard Guenther <rguenth at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Keywords|                            |missed-optimization
             Target|                            |x86_64-*-*
             Status|UNCONFIRMED                 |NEW
   Last reconfirmed|                            |2011.04.14 12:58:38
     Ever Confirmed|0                           |1

--- Comment #1 from Richard Guenther <rguenth at gcc dot gnu.org> 2011-04-14 12:58:38 UTC ---
STH is seriously confused - just look at the initial code after expansion
(or -O0 asm).  Somehow I see DImode <retval> being mentioned - that hints
at some backend bogosity.  I would have expected to see some CONCATs
at least.

[Bug target/48607] Inefficient complex float parameter passing

[pinskia at gcc dot gnu.org]

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=48607

Andrew Pinski <pinskia at gcc dot gnu.org> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |RESOLVED
         Resolution|---                         |DUPLICATE

--- Comment #2 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
PR 48609 is a reduced testcase of this and has the same alysis so closing as a
dup.

*** This bug has been marked as a duplicate of bug 48609 ***