[Bug rtl-optimization/19236] New: ICE compiling logp(x) with -ffast-math

[Bug rtl-optimization/19236] New: ICE compiling logp(x) with -ffast-math

[bredelin at ucla dot edu]

The C compiler crashes upon seeing the 'log1p()' function, as follows:

$ gcc-snap -ffast-math -c test-logsum.c
test-logsum.c: In function 'test':
test-logsum.c:6: error: unrecognizable insn:
(insn 9 8 10 1 (set (reg:XF 63)
        (float_extend:XF (reg:XF 61))) -1 (nil)
    (expr_list:REG_DEAD (reg:XF 61)
        (nil)))
test-logsum.c:6: internal compiler error: in extract_insn, at recog.c:2020
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.

Here is the (short) test-case file:

----------------------
#include <math.h>

double test () 
{
  return log1p(1.0);
}
----------------------

$ gcc-snap -v
Using built-in specs.
Configured with: ../src/configure -v --enable-languages=c,c++,java,objc,ada
--prefix=/usr/lib/gcc-snapshot --enable-shared --with-system-zlib --enable-nls
--enable-threads=posix --without-included-gettext --disable-werror
--enable-__cxa_atexit --enable-libstdcxx-allocator=mt --enable-clocale=gnu
--enable-libstdcxx-debug --enable-java-gc=boehm --enable-java-awt=gtk i486-linux-gnu
Thread model: posix
gcc version 4.0.0 20041231 (experimental)

NOTES:
1. BTW, this is a Debian snapshot.  (2004-12-31)
2. The bug also occurs with a previous package:

$ gcc-4.0 -v
Reading specs from /usr/lib/gcc/i486-linux/4.0.0/specs
Configured with: ../src/configure -v --enable-languages=c,c++,java,f95,objc,ada
--prefix=/usr --libexecdir=/usr/lib --enable-shared --with-system-zlib
--enable-nls --enable-threads=posix --without-included-gettext
--program-suffix=-4.0 --enable-__cxa_atexit --enable-libstdcxx-allocator=mt
--enable-clocale=gnu --enable-libstdcxx-debug --enable-java-gc=boehm
--enable-java-awt=gtk --enable-mpfr i486-linux
Thread model: posix
gcc version 4.0.0 20041205 (experimental) (Debian 4.0-0pre2)

3. I am *fairly* sure that the Debian version 4.0-0pre1 was in November and did
not have this problem.

4. The problem did not exist in HEAD in the beginning of October.

-- 
           Summary: ICE compiling logp(x) with -ffast-math
           Product: gcc
           Version: 4.0.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: rtl-optimization
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: bredelin at ucla dot edu
                CC: gcc-bugs at gcc dot gnu dot org
GCC target triplet: i486-linux-gnu


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19236

[Bug rtl-optimization/19236] [4.0 regression] ICE compiling logp(x) with -ffast-math

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2005-01-03 05:24 -------
Confirmed. This is a regression over 3.4.x. 
 
My mainline snapshot if from 20041214, so the problem predates that date. 
 
W. 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|normal                      |critical
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
           Keywords|                            |ice-on-valid-code
      Known to fail|                            |4.0.0
      Known to work|                            |3.4.3
   Last reconfirmed|0000-00-00 00:00:00         |2005-01-03 05:24:18
               date|                            |
            Summary|ICE compiling logp(x) with -|[4.0 regression] ICE
                   |ffast-math                  |compiling logp(x) with -
                   |                            |ffast-math
   Target Milestone|---                         |4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19236

[Bug middle-end/19236] [4.0 regression] ICE compiling logp(x) with -ffast-math

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2005-01-03 05:30 -------
The minimal testcase is this: 
-------------------- 
extern double log1p (double __x);  
double test () { 
  return log1p(1.0); 
} 
-------------------- 
 
g/x> /home/bangerth/bin/gcc-4.0-pre/bin/c++ -ffast-math -c x.i 
x.i: In function ?double test()?: 
x.i:4: error: unrecognizable insn: 
(insn 9 8 10 0 (set (reg:XF 64) 
        (float_extend:XF (reg:XF 62))) -1 (nil) 
    (expr_list:REG_DEAD (reg:XF 62) 
        (nil))) 
x.i:4: internal compiler error: in extract_insn, at recog.c:2020 
Please submit a full bug report, 
with preprocessed source if appropriate. 
 
The same testcase ICEs with the C++ compiler, btw. 
 
W. 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|rtl-optimization            |middle-end


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19236

[Bug middle-end/19236] [4.0 regression] ICE compiling logp(x) with -ffast-math

[aj at gcc dot gnu dot org]

------- Additional Comments From aj at gcc dot gnu dot org  2005-01-03 07:30 -------
Let's add Uros since he worked on the i386 backend lately and Honza as another 
expert. 

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |uros at kss-loka dot si,
                   |                            |hubicka at gcc dot gnu dot
                   |                            |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19236

[Bug target/19236] [4.0 regression] ICE compiling logp(x) with -ffast-math

[uros at kss-loka dot si]

------- Additional Comments From uros at kss-loka dot si  2005-01-03 08:53 -------
Patch here: http://gcc.gnu.org/ml/gcc-patches/2005-01/msg00062.html

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |uros at kss-loka dot si
                   |dot org                     |
             Status|NEW                         |ASSIGNED
          Component|middle-end                  |target
   Last reconfirmed|2005-01-03 05:24:18         |2005-01-03 08:53:06
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19236

[Bug target/19236] [4.0 regression] ICE compiling logp(x) with -ffast-math

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-01-03 09:55 -------
Subject: Bug 19236

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	uros@gcc.gnu.org	2005-01-03 09:54:57

Modified files:
	gcc            : ChangeLog 
	gcc/config/i386: i386.md 

Log message:
	PR target/19236
	* config/i386/i386.c (log1psf2): Change mode of operands[0,1]
	to SFmode.
	(log1pdf2): Change mode of operands[0,1] to DFmode.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.7005&r2=2.7006
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/i386/i386.md.diff?cvsroot=gcc&r1=1.596&r2=1.597



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19236

[Bug target/19236] [4.0 regression] ICE compiling logp(x) with -ffast-math

[uros at kss-loka dot si]

------- Additional Comments From uros at kss-loka dot si  2005-01-03 09:55 -------
Fixed.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19236

[Bug target/19236] [4.0 regression] ICE compiling logp(x) with -ffast-math

[bangerth at dealii dot org]

------- Additional Comments From bangerth at dealii dot org  2005-01-03 14:34 -------
Uros, would you mind adding the testcase as well? I know it's small, but 
who knows what it's good for -- we got this case wrong once already, let's 
make sure we don't again. 
 
Thanks 
  Wolfgang 

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19236

[Bug target/19236] [4.0 regression] ICE compiling logp(x) with -ffast-math

[uros at kss-loka dot si]

------- Additional Comments From uros at kss-loka dot si  2005-01-03 15:37 -------
(In reply to comment #7)

No problem, this one should do the trick:

--cut here--
/* PR target/19236 */
/* { dg-do compile { target i?86-*-* } } */
/* { dg-options "-ffast-math" } */

extern float log1pf (float);
extern double log1p (double);  

float testf (float __x) { 
  return log1pf(1.0); 
}

double test (double __x) { 
  return log1p(1.0); 
}
--cut here--


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19236