Re: target/9068: [3.4 regression] [x86] comisd & comiss constraints are incorrect

Re: target/9068: [3.4 regression] [x86] comisd & comiss constraints are incorrect

[hubicka]

Synopsis: [3.4 regression] [x86] comisd & comiss constraints are incorrect

State-Changed-From-To: analyzed->closed
State-Changed-By: hubicka
State-Changed-When: Fri Mar 21 21:09:48 2003
State-Changed-Why:
    Fixed by my earlier patch

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=9068

Re: target/9068: [3.4 regression] [x86] comisd & comiss constraints are incorrect

[Jan Hubicka]

The following reply was made to PR target/9068; it has been noted by GNATS.

>From jh@suse.cz  Mon Jan 13 00:57:26 2003
From: Jan Hubicka <jh@suse.cz>
To: bangerth@dealii.org, gcc-bugs@gcc.gnu.org, gcc-prs@gcc.gnu.org,
	jh@suse.cz, kelleycook@comcast.net, nobody@gcc.gnu.org,
	gcc-gnats@gcc.gnu.org
Cc:  
Subject: Re: target/9068: [3.4 regression] [x86] comisd & comiss constraints are incorrect
Date: Sat, 11 Jan 2003 11:05:02 +0100

 > Old Synopsis: [x86] comisd & comiss intel-syntax constraints are incorrect
 > New Synopsis: [3.4 regression] [x86] comisd & comiss constraints are incorrect
 > 
 > State-Changed-From-To: open->analyzed
 > State-Changed-By: bangerth
 > State-Changed-When: Fri Jan 10 16:32:10 2003
 > State-Changed-Why:
 >     Confirmed. This also happens on x86-linux, and has nothing
 >     to do with intel asm syntax, since I can reproduce it
 >     with this smaller testcase
 >     ----------------------------------
 >     int foo(int count, double sum) {
 >         return (sum/count > 0.0000001);
 >     }
 >     --------------------------------
 >     and the following (shorter) command line:
 >       tmp/g> /home/bangerth/bin/gcc-3.4-pre/bin/c++ -msse2 x.c
 >       x.c: In function `int foo(int, double)':
 >       x.c:3: internal compiler error: in get_attr_length_immediate, at insn-attrtab.c
 >        :22751
 >       Please submit a full bug report,
 
 This should be independent problem I fixed yesterday.
 >     
 >     This is a regression w.r.t. 3.3, where things worked, 
 >     although probably more by chance.
 >     Jan, you are the author of the hunk of code that the patch
 >     in this report touches. Can you comment on its validity?
 >     
 >     Thinking more about it, the original report probably
 >     was about this code (which indeed fails only with intel
 >     asm syntax):
 >     -----------------------------
 >     int foo(double sum) {
 >         return (sum > 0.0000001);
 >     }
 >     -------------------------------
 >     tmp/g> /home/bangerth/bin/gcc-3.4-pre/bin/c++ -msse2 -Wall -march=pentium4 -ffast-math -mfpmath=sse -masm=intel x.c
 >     x.c: In function `int foo(double)':
 >     x.c:3: internal compiler error: output_operand: operand number missing after
 >        %-letter
 >     Please submit a full bug report,
 I will fix this too.
 
 Honza

Re: target/9068: [3.4 regression] [x86] comisd & comiss constraints are incorrect

[bangerth]

Old Synopsis: [x86] comisd & comiss intel-syntax constraints are incorrect
New Synopsis: [3.4 regression] [x86] comisd & comiss constraints are incorrect

State-Changed-From-To: open->analyzed
State-Changed-By: bangerth
State-Changed-When: Fri Jan 10 16:32:10 2003
State-Changed-Why:
    Confirmed. This also happens on x86-linux, and has nothing
    to do with intel asm syntax, since I can reproduce it
    with this smaller testcase
    ----------------------------------
    int foo(int count, double sum) {
        return (sum/count > 0.0000001);
    }
    --------------------------------
    and the following (shorter) command line:
      tmp/g> /home/bangerth/bin/gcc-3.4-pre/bin/c++ -msse2 x.c
      x.c: In function `int foo(int, double)':
      x.c:3: internal compiler error: in get_attr_length_immediate, at insn-attrtab.c
       :22751
      Please submit a full bug report,
    
    This is a regression w.r.t. 3.3, where things worked, 
    although probably more by chance.
    Jan, you are the author of the hunk of code that the patch
    in this report touches. Can you comment on its validity?
    
    Thinking more about it, the original report probably
    was about this code (which indeed fails only with intel
    asm syntax):
    -----------------------------
    int foo(double sum) {
        return (sum > 0.0000001);
    }
    -------------------------------
    tmp/g> /home/bangerth/bin/gcc-3.4-pre/bin/c++ -msse2 -Wall -march=pentium4 -ffast-math -mfpmath=sse -masm=intel x.c
    x.c: In function `int foo(double)':
    x.c:3: internal compiler error: output_operand: operand number missing after
       %-letter
    Please submit a full bug report,
    
    
    So it seems as if these were two reports in one. The latter
    example, however, never worked as long as SSE support is
    present, which is not surprising since the code the patch
    in this report touches is unchanged since the first SSE
    patch to i386.c
    
    W.

http://gcc.gnu.org/cgi-bin/gnatsweb.pl?cmd=view%20audit-trail&database=gcc&pr=9068