[Bug c++/12612] New: ICE when compiling with -O3 -ffast-math

public inbox for gcc-bugs@sourceware.org
help / color / mirror / Atom feed

* [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math
@ 2003-10-14  7:51 martin at mpa-garching dot mpg dot de
  2003-10-14  7:53 ` [Bug c++/12612] " martin at mpa-garching dot mpg dot de
                   ` (13 more replies)
  0 siblings, 14 replies; 15+ messages in thread
From: martin at mpa-garching dot mpg dot de @ 2003-10-14  7:51 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612

           Summary: ICE when compiling with -O3 -ffast-math
           Product: gcc
           Version: 3.4
            Status: UNCONFIRMED
          Severity: normal
          Priority: P2
         Component: c++
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: martin at mpa-garching dot mpg dot de
                CC: gcc-bugs at gcc dot gnu dot org,martin at mpa-garching
                    dot mpg dot de
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: i686-pc-linux-gnu

gcc version 3.4 20031014 (experimental) crashes with an ICE when
compiling the attached code with -O3 -ffast-math:

~/data/planck/LevelS/Sampler>g++ -v -c -O3 -ffast-math sampler.ii
Reading specs from /afs/mpa/data/martin/ugcc/lib/gcc/i686-pc-linux-gnu/3.4/specs
Configured with: /scratch/gcc/configure --prefix=/afs/mpa/data/martin/ugcc
--enable-languages=c++ --disable-checking
Thread model: posix
gcc version 3.4 20031014 (experimental)
 /afs/mpa/data/martin/ugcc/libexec/gcc/i686-pc-linux-gnu/3.4/cc1plus
-fpreprocessed sampler.ii -quiet -dumpbase sampler.ii -mtune=pentiumpro -auxbase
sampler -O3 -version -ffast-math -o /tmp/ccycYGAB.s
GNU C++ version 3.4 20031014 (experimental) (i686-pc-linux-gnu)
        compiled by GNU C version 3.4 20031014 (experimental).
GGC heuristics: --param ggc-min-expand=63 --param ggc-min-heapsize=63442
sampler.cc: In function `int main(int, char**)':
sampler.cc:192: error: unrecognizable insn:
(insn:HI 6444 6442 6445 619 (parallel [
            (set (reg:DF 2256)
                (unspec:DF [
                        (reg:DF 2263)
                        (reg:XF 5025)
                    ] 66))
            (clobber (reg:XF 5412))
        ]) -1 (insn_list 6442 (nil))
    (expr_list:REG_DEAD (reg:DF 2263)
        (expr_list:REG_UNUSED (reg:XF 5412)
            (expr_list:REG_EQUAL (unspec:DF [
                        (reg:DF 2263)
                        (reg:XF 2264)
                    ] 66)
                (nil)))))
sampler.cc:192: internal compiler error: in extract_insn, at recog.c:2064
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.

The bug appears to be relatively new; it didn't trigger about two weeks ago.

Sorry for the long testcase; I currently don't have time to reduce it.


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug c++/12612] ICE when compiling with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
@ 2003-10-14  7:53 ` martin at mpa-garching dot mpg dot de
  2003-10-14 16:22 ` [Bug optimization/12612] [3.4 Regression] " pinskia at gcc dot gnu dot org
                   ` (12 subsequent siblings)
  13 siblings, 0 replies; 15+ messages in thread
From: martin at mpa-garching dot mpg dot de @ 2003-10-14  7:53 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612



------- Additional Comments From martin at mpa-garching dot mpg dot de  2003-10-14 07:53 -------
Created an attachment (id=4932)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=4932&action=view)
A (long) testcase to reproduce the bug


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when compiling with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
  2003-10-14  7:53 ` [Bug c++/12612] " martin at mpa-garching dot mpg dot de
@ 2003-10-14 16:22 ` pinskia at gcc dot gnu dot org
  2003-10-17 16:38 ` [Bug optimization/12612] [3.4 Regression] ICE when using log function " reichelt at gcc dot gnu dot org
                   ` (11 subsequent siblings)
  13 siblings, 0 replies; 15+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2003-10-14 16:22 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612


pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|c++                         |optimization
           Keywords|                            |ice-on-valid-code
            Summary|ICE when compiling with -O3 |[3.4 Regression] ICE when
                   |-ffast-math                 |compiling with -O3 -ffast-
                   |                            |math
   Target Milestone|---                         |3.4


------- Additional Comments From pinskia at gcc dot gnu dot org  2003-10-14 16:22 -------
It is a very new bug, as it does not occur in the mainline 20031009.


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
  2003-10-14  7:53 ` [Bug c++/12612] " martin at mpa-garching dot mpg dot de
  2003-10-14 16:22 ` [Bug optimization/12612] [3.4 Regression] " pinskia at gcc dot gnu dot org
@ 2003-10-17 16:38 ` reichelt at gcc dot gnu dot org
  2003-10-17 17:41 ` pinskia at gcc dot gnu dot org
                   ` (10 subsequent siblings)
  13 siblings, 0 replies; 15+ messages in thread
From: reichelt at gcc dot gnu dot org @ 2003-10-17 16:38 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612


reichelt at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |reichelt at gcc dot gnu dot
                   |                            |org
           Severity|normal                      |critical
             Status|UNCONFIRMED                 |NEW
     Ever Confirmed|                            |1
           Keywords|                            |monitored
   Last reconfirmed|0000-00-00 00:00:00         |2003-10-17 16:37:30
               date|                            |
            Summary|[3.4 Regression] ICE when   |[3.4 Regression] ICE when
                   |compiling with -O3 -ffast-  |using log function with -O3
                   |math                        |-ffast-math


------- Additional Comments From reichelt at gcc dot gnu dot org  2003-10-17 16:37 -------
Here's a short testcase that crashes cc1 and cc1plus when compiled with
"-O3 -ffast-math":

=============================
double log (double);

void foo (double* x)
{
    int i;
    for (i=0; i<1; ++i)
        x[0]=log(2.0);
}
=============================

The error message is:

PR12612.c: In function `foo':

PR12612.c:8: error: unrecognizable insn:
(insn:HI 66 57 33 0 (parallel [
            (set (reg:DF 60)
                (unspec:DF [
                        (reg:DF 64)
                        (reg:XF 63)
                    ] 66))
            (clobber (reg:XF 70))
        ]) -1 (insn_list 56 (insn_list 57 (nil)))
    (expr_list:REG_DEAD (reg:XF 63)
        (expr_list:REG_DEAD (reg:DF 64)
            (expr_list:REG_UNUSED (reg:XF 70)
                (nil)))))
PR12612.c:8: internal compiler error: in extract_insn, at recog.c:2064

If I replace the log function by sin, cos, exp, the code compiles.

According to Phil's regression hunter, the regression was introduced
between 2003-10-10-trunk (#422) and 2003-10-12-trunk (#423).


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
                   ` (2 preceding siblings ...)
  2003-10-17 16:38 ` [Bug optimization/12612] [3.4 Regression] ICE when using log function " reichelt at gcc dot gnu dot org
@ 2003-10-17 17:41 ` pinskia at gcc dot gnu dot org
  2003-10-17 17:41 ` reichelt at gcc dot gnu dot org
                   ` (9 subsequent siblings)
  13 siblings, 0 replies; 15+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2003-10-17 17:41 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612



------- Additional Comments From pinskia at gcc dot gnu dot org  2003-10-17 17:41 -------
Most likely another machine description problem like PR 12630 was.


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
                   ` (3 preceding siblings ...)
  2003-10-17 17:41 ` pinskia at gcc dot gnu dot org
@ 2003-10-17 17:41 ` reichelt at gcc dot gnu dot org
  2003-10-17 18:12 ` reichelt at gcc dot gnu dot org
                   ` (8 subsequent siblings)
  13 siblings, 0 replies; 15+ messages in thread
From: reichelt at gcc dot gnu dot org @ 2003-10-17 17:41 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612


reichelt at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |hubicka at ucw dot cz


------- Additional Comments From reichelt at gcc dot gnu dot org  2003-10-17 17:29 -------
Jan,

it looks like your patch

http://gcc.gnu.org/ml/gcc-cvs/2003-10/msg00441.html

is responsible for the regression.
Could you please have a look?

Thanks,
Volker


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
                   ` (4 preceding siblings ...)
  2003-10-17 17:41 ` reichelt at gcc dot gnu dot org
@ 2003-10-17 18:12 ` reichelt at gcc dot gnu dot org
  2003-10-17 18:13 ` hubicka at ucw dot cz
                   ` (7 subsequent siblings)
  13 siblings, 0 replies; 15+ messages in thread
From: reichelt at gcc dot gnu dot org @ 2003-10-17 18:12 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612

------- Additional Comments From reichelt at gcc dot gnu dot org  2003-10-17 17:54 -------
Something with the optimization -fweb seems to be wrong.
Because of Jan's patch it's turned on with -O3.
That's probably why the bug appears with the patch.

Btw, just specifying the option "-fweb" on the command
line causes an ICE:

cc1: internal compiler error: in common_handle_option, at opts.c:656
Please submit a full bug report, [etc.]

In addition the documentation is quite incomplete: "-fweb" is only
mentioned once in the manual:

Optimize yet more.  @option{-O3} turns on all optimizations specified by
@option{-O2} and also turns on the @option{-finline-functions},
@option{-fweb} and @option{-frename-registers} options.

But nowhere is an explanation what "-fweb actually" does and it doesn't
appear in the list of optimization options.

^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
                   ` (5 preceding siblings ...)
  2003-10-17 18:12 ` reichelt at gcc dot gnu dot org
@ 2003-10-17 18:13 ` hubicka at ucw dot cz
  2003-10-18 18:49 ` pinskia at gcc dot gnu dot org
                   ` (6 subsequent siblings)
  13 siblings, 0 replies; 15+ messages in thread
From: hubicka at ucw dot cz @ 2003-10-17 18:13 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612



------- Additional Comments From hubicka at ucw dot cz  2003-10-17 18:12 -------
Subject: Re:  [3.4 Regression] ICE when using log function with -O3 -ffast-math

> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612
> 
> 
> 
> ------- Additional Comments From reichelt at gcc dot gnu dot org  2003-10-17 17:54 -------
> Something with the optimization -fweb seems to be wrong.
> Because of Jan's patch it's turned on with -O3.
> That's probably why the bug appears with the patch.
> 
> Btw, just specifying the option "-fweb" on the command
> line causes an ICE:
> 
> cc1: internal compiler error: in common_handle_option, at opts.c:656
> Please submit a full bug report, [etc.]
> 
> In addition the documentation is quite incomplete: "-fweb" is only
> mentioned once in the manual:
> 
> Optimize yet more.  @option{-O3} turns on all optimizations specified by
> @option{-O2} and also turns on the @option{-finline-functions},
> @option{-fweb} and @option{-frename-registers} options.
> 
> But nowhere is an explanation what "-fweb actually" does and it doesn't
> appear in the list of optimization options.

This looks like some hunk of patch got lost.  I definitly wrote
documentation for that.  I will fix that.

Thanks,
Honza
> 
> 
> 
> 
> ------- You are receiving this mail because: -------
> You are on the CC list for the bug, or are watching someone who is.


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
                   ` (6 preceding siblings ...)
  2003-10-17 18:13 ` hubicka at ucw dot cz
@ 2003-10-18 18:49 ` pinskia at gcc dot gnu dot org
  2003-10-18 23:17 ` hubicka at ucw dot cz
                   ` (5 subsequent siblings)
  13 siblings, 0 replies; 15+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2003-10-18 18:49 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612



------- Additional Comments From pinskia at gcc dot gnu dot org  2003-10-18 18:44 -------
Yep it is a machine description problem just like PR 12630 was.
The idea is to use match_scratch: instead of match_dup.


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
                   ` (7 preceding siblings ...)
  2003-10-18 18:49 ` pinskia at gcc dot gnu dot org
@ 2003-10-18 23:17 ` hubicka at ucw dot cz
  2003-10-18 23:21 ` hubicka at ucw dot cz
                   ` (4 subsequent siblings)
  13 siblings, 0 replies; 15+ messages in thread
From: hubicka at ucw dot cz @ 2003-10-18 23:17 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612



------- Additional Comments From hubicka at ucw dot cz  2003-10-18 21:29 -------
Subject: Re:  [3.4 Regression] ICE when using log function with -O3 -ffast-math

> PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> 
> http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612
> 
> 
> 
> ------- Additional Comments From pinskia at gcc dot gnu dot org  2003-10-18 18:44 -------
> Yep it is a machine description problem just like PR 12630 was.
> The idea is to use match_scratch: instead of match_dup.
Right,
This is patch I made for that yesterday.
OK?
Sat Oct 18 23:14:37 CEST 2003  Jan Hubicka  <jh@suse.cz>
	PR optimization/12612
	* i386.md (fyl2x patterns and expanders): Use match_scratch instead of match_dup.
Index: i386.md
===================================================================
RCS file: /cvs/gcc/gcc/gcc/config/i386/i386.md,v
retrieving revision 1.485
diff -c -3 -p -r1.485 i386.md
*** i386.md	18 Oct 2003 07:24:59 -0000	1.485
--- i386.md	18 Oct 2003 21:07:30 -0000
***************
*** 15723,15729 ****
  		   (unspec:SF [(match_operand:SF 2 "register_operand" "0")
  			       (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)"
--- 15723,15729 ----
  		   (unspec:SF [(match_operand:SF 2 "register_operand" "0")
  			       (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
! 	      (clobber (match_scratch:SF 3 "1"))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)"
***************
*** 15736,15742 ****
  		   (unspec:DF [(match_operand:DF 2 "register_operand" "0")
  			       (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)"
--- 15736,15742 ----
  		   (unspec:DF [(match_operand:DF 2 "register_operand" "0")
  			       (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
! 	      (clobber (match_scratch:DF 3 "1"))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)"
***************
*** 15749,15755 ****
  		   (unspec:XF [(match_operand:XF 2 "register_operand" "0")
  			       (match_operand:XF 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
    "fyl2x"
--- 15749,15755 ----
  		   (unspec:XF [(match_operand:XF 2 "register_operand" "0")
  			       (match_operand:XF 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
! 	      (clobber (match_scratch:XF 3 "1"))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
    "fyl2x"
***************
*** 15761,15767 ****
  		   (unspec:TF [(match_operand:TF 2 "register_operand" "0")
  			       (match_operand:TF 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
    "fyl2x"
--- 15761,15767 ----
  		   (unspec:TF [(match_operand:TF 2 "register_operand" "0")
  			       (match_operand:TF 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
! 	      (clobber (match_scratch:TF 3 "1"))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
    "fyl2x"
***************
*** 15772,15778 ****
    [(parallel [(set (match_operand:SF 0 "register_operand" "")
  		   (unspec:SF [(match_operand:SF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_dup 2))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
--- 15772,15778 ----
    [(parallel [(set (match_operand:SF 0 "register_operand" "")
  		   (unspec:SF [(match_operand:SF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_scratch:SF 3 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
***************
*** 15787,15793 ****
    [(parallel [(set (match_operand:DF 0 "register_operand" "")
  		   (unspec:DF [(match_operand:DF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_dup 2))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
--- 15787,15793 ----
    [(parallel [(set (match_operand:DF 0 "register_operand" "")
  		   (unspec:DF [(match_operand:DF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_scratch:DF 3 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
***************
*** 15802,15808 ****
    [(parallel [(set (match_operand:XF 0 "register_operand" "")
  		   (unspec:XF [(match_operand:XF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_dup 2))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
  {
--- 15802,15808 ----
    [(parallel [(set (match_operand:XF 0 "register_operand" "")
  		   (unspec:XF [(match_operand:XF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_scratch:XF 3 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
  {
***************
*** 15817,15823 ****
    [(parallel [(set (match_operand:TF 0 "register_operand" "")
  		   (unspec:TF [(match_operand:TF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_dup 2))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
  {
--- 15817,15823 ----
    [(parallel [(set (match_operand:TF 0 "register_operand" "")
  		   (unspec:TF [(match_operand:TF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_scratch:TF 3 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
  {


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
                   ` (8 preceding siblings ...)
  2003-10-18 23:17 ` hubicka at ucw dot cz
@ 2003-10-18 23:21 ` hubicka at ucw dot cz
  2003-10-19 17:26 ` rth at redhat dot com
                   ` (3 subsequent siblings)
  13 siblings, 0 replies; 15+ messages in thread
From: hubicka at ucw dot cz @ 2003-10-18 23:21 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612



------- Additional Comments From hubicka at ucw dot cz  2003-10-18 23:17 -------
Subject: Re:  [3.4 Regression] ICE when using log function with -O3 -ffast-math

> > PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.
> > 
> > http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612
> > 
> > 
> > 
> > ------- Additional Comments From pinskia at gcc dot gnu dot org  2003-10-18 18:44 -------
> > Yep it is a machine description problem just like PR 12630 was.
> > The idea is to use match_scratch: instead of match_dup.
> Right,
> This is patch I made for that yesterday.
> OK?
> Sat Oct 18 23:14:37 CEST 2003  Jan Hubicka  <jh@suse.cz>
> 	PR optimization/12612
> 	* i386.md (fyl2x patterns and expanders): Use match_scratch instead of match_dup.

It is actually more than that.  Several other patterns uses same scheme.
Additionally they use bogus extra parallel that confuses genrecog
handling of clobbers and finally reg-stack needs to be updated to
produce insns that still match.

I am still testing the attached patch, OK assuming it pass?

Honza

2003-10-19  Jan Hubicka  <jh@suse.cz>
	* reg-stack.c (subst_stack_regs_pat):  Use st(1) for clobbers.
	* i386.md (fpatan, fyl2x, fscale patterns and expanders): Use
	match_scratch; avoid bogus paralles.

Index: reg-stack.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/reg-stack.c,v
retrieving revision 1.134
diff -c -3 -p -r1.134 reg-stack.c
*** reg-stack.c	5 Oct 2003 19:50:55 -0000	1.134
--- reg-stack.c	18 Oct 2003 23:12:35 -0000
*************** subst_stack_regs_pat (rtx insn, stack re
*** 1430,1436 ****
  		      abort ();
  		  }
  		remove_note (insn, note);
! 		replace_reg (dest, LAST_STACK_REG);
  	      }
  	    else
  	      {
--- 1430,1436 ----
  		      abort ();
  		  }
  		remove_note (insn, note);
! 		replace_reg (dest, FIRST_STACK_REG + 1);
  	      }
  	    else
  	      {
Index: config/i386/i386.md
===================================================================
RCS file: /cvs/gcc/gcc/gcc/config/i386/i386.md,v
retrieving revision 1.485
diff -c -3 -p -r1.485 i386.md
*** config/i386/i386.md	18 Oct 2003 07:24:59 -0000	1.485
--- config/i386/i386.md	18 Oct 2003 23:12:38 -0000
***************
*** 15619,15629 ****
     (set_attr "mode" "XF")])
  
  (define_insn "atan2df3_1"
!   [(parallel [(set (match_operand:DF 0 "register_operand" "=f")
! 		   (unspec:DF [(match_operand:DF 2 "register_operand" "0")
! 			       (match_operand:DF 1 "register_operand" "u")]
! 		    UNSPEC_FPATAN))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
    "fpatan"
--- 15619,15629 ----
     (set_attr "mode" "XF")])
  
  (define_insn "atan2df3_1"
!   [(set (match_operand:DF 0 "register_operand" "=f")
! 	(unspec:DF [(match_operand:DF 2 "register_operand" "0")
! 		    (match_operand:DF 1 "register_operand" "u")]
! 		   UNSPEC_FPATAN))
!    (clobber (match_scratch:DF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
    "fpatan"
***************
*** 15644,15654 ****
  })
  
  (define_insn "atan2sf3_1"
!   [(parallel [(set (match_operand:SF 0 "register_operand" "=f")
! 		   (unspec:SF [(match_operand:SF 2 "register_operand" "0")
! 			       (match_operand:SF 1 "register_operand" "u")]
! 		    UNSPEC_FPATAN))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
    "fpatan"
--- 15644,15654 ----
  })
  
  (define_insn "atan2sf3_1"
!   [(set (match_operand:SF 0 "register_operand" "=f")
!         (unspec:SF [(match_operand:SF 2 "register_operand" "0")
! 		    (match_operand:SF 1 "register_operand" "u")]
! 		   UNSPEC_FPATAN))
!    (clobber (match_scratch:SF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
    "fpatan"
***************
*** 15669,15679 ****
  })
  
  (define_insn "atan2xf3_1"
!   [(parallel [(set (match_operand:XF 0 "register_operand" "=f")
! 		   (unspec:XF [(match_operand:XF 2 "register_operand" "0")
! 			       (match_operand:XF 1 "register_operand" "u")]
! 		    UNSPEC_FPATAN))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && ! TARGET_128BIT_LONG_DOUBLE"
    "fpatan"
--- 15669,15679 ----
  })
  
  (define_insn "atan2xf3_1"
!   [(set (match_operand:XF 0 "register_operand" "=f")
!         (unspec:XF [(match_operand:XF 2 "register_operand" "0")
! 	            (match_operand:XF 1 "register_operand" "u")]
! 	           UNSPEC_FPATAN))
!    (clobber (match_scratch:XF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && ! TARGET_128BIT_LONG_DOUBLE"
    "fpatan"
***************
*** 15694,15704 ****
  })
  
  (define_insn "atan2tf3_1"
!   [(parallel [(set (match_operand:TF 0 "register_operand" "=f")
! 		   (unspec:TF [(match_operand:TF 2 "register_operand" "0")
! 			       (match_operand:TF 1 "register_operand" "u")]
! 		    UNSPEC_FPATAN))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
    "fpatan"
--- 15694,15704 ----
  })
  
  (define_insn "atan2tf3_1"
!   [(set (match_operand:TF 0 "register_operand" "=f")
!         (unspec:TF [(match_operand:TF 2 "register_operand" "0")
! 		    (match_operand:TF 1 "register_operand" "u")]
! 		   UNSPEC_FPATAN))
!    (clobber (match_scratch:TF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
    "fpatan"
***************
*** 15719,15729 ****
  })
  
  (define_insn "*fyl2x_sfxf3"
!   [(parallel [(set (match_operand:SF 0 "register_operand" "=f")
! 		   (unspec:SF [(match_operand:SF 2 "register_operand" "0")
! 			       (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)"
--- 15719,15729 ----
  })
  
  (define_insn "*fyl2x_sfxf3"
!   [(set (match_operand:SF 0 "register_operand" "=f")
!          (unspec:SF [(match_operand:SF 2 "register_operand" "0")
! 		     (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
!    (clobber (match_scratch:SF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)"
***************
*** 15732,15742 ****
     (set_attr "mode" "SF")])
  
  (define_insn "*fyl2x_dfxf3"
!   [(parallel [(set (match_operand:DF 0 "register_operand" "=f")
! 		   (unspec:DF [(match_operand:DF 2 "register_operand" "0")
! 			       (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)"
--- 15732,15742 ----
     (set_attr "mode" "SF")])
  
  (define_insn "*fyl2x_dfxf3"
!   [(set (match_operand:DF 0 "register_operand" "=f")
!          (unspec:DF [(match_operand:DF 2 "register_operand" "0")
! 		     (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
!    (clobber (match_scratch:DF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)"
***************
*** 15745,15755 ****
     (set_attr "mode" "DF")])
  
  (define_insn "*fyl2x_xf3"
!   [(parallel [(set (match_operand:XF 0 "register_operand" "=f")
! 		   (unspec:XF [(match_operand:XF 2 "register_operand" "0")
! 			       (match_operand:XF 1 "register_operand" "u")]
! 		    UNSPEC_FYL2X))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
    "fyl2x"
--- 15745,15755 ----
     (set_attr "mode" "DF")])
  
  (define_insn "*fyl2x_xf3"
!   [(set (match_operand:XF 0 "register_operand" "=f")
!         (unspec:XF [(match_operand:XF 2 "register_operand" "0")
! 		    (match_operand:XF 1 "register_operand" "u")]
! 	           UNSPEC_FYL2X))
!    (clobber (match_scratch:XF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
    "fyl2x"
***************
*** 15757,15767 ****
     (set_attr "mode" "XF")])
  
  (define_insn "*fyl2x_tfxf3"
!   [(parallel [(set (match_operand:TF 0 "register_operand" "=f")
! 		   (unspec:TF [(match_operand:TF 2 "register_operand" "0")
! 			       (match_operand:TF 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
    "fyl2x"
--- 15757,15767 ----
     (set_attr "mode" "XF")])
  
  (define_insn "*fyl2x_tfxf3"
!   [(set (match_operand:TF 0 "register_operand" "=f")
! 	(unspec:TF [(match_operand:TF 2 "register_operand" "0")
! 		    (match_operand:TF 1 "register_operand" "u")]
  		    UNSPEC_FYL2X))
!    (clobber (match_scratch:TF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
    "fyl2x"
***************
*** 15772,15778 ****
    [(parallel [(set (match_operand:SF 0 "register_operand" "")
  		   (unspec:SF [(match_operand:SF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_dup 2))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
--- 15772,15778 ----
    [(parallel [(set (match_operand:SF 0 "register_operand" "")
  		   (unspec:SF [(match_operand:SF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_scratch:SF 3 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
***************
*** 15787,15793 ****
    [(parallel [(set (match_operand:DF 0 "register_operand" "")
  		   (unspec:DF [(match_operand:DF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_dup 2))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
--- 15787,15793 ----
    [(parallel [(set (match_operand:DF 0 "register_operand" "")
  		   (unspec:DF [(match_operand:DF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_scratch:DF 3 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
***************
*** 15802,15808 ****
    [(parallel [(set (match_operand:XF 0 "register_operand" "")
  		   (unspec:XF [(match_operand:XF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_dup 2))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
  {
--- 15802,15808 ----
    [(parallel [(set (match_operand:XF 0 "register_operand" "")
  		   (unspec:XF [(match_operand:XF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_scratch:XF 3 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
  {
***************
*** 15817,15823 ****
    [(parallel [(set (match_operand:TF 0 "register_operand" "")
  		   (unspec:TF [(match_operand:TF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_dup 2))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
  {
--- 15817,15823 ----
    [(parallel [(set (match_operand:TF 0 "register_operand" "")
  		   (unspec:TF [(match_operand:TF 1 "register_operand" "")
  			       (match_dup 2)] UNSPEC_FYL2X))
! 	      (clobber (match_scratch:TF 3 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
  {
***************
*** 15829,15839 ****
  })
  
  (define_insn "*fscale_sfxf3"
!   [(parallel [(set (match_operand:SF 0 "register_operand" "=f")
! 		   (unspec:SF [(match_operand 2 "register_operand" "0")
! 			       (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FSCALE))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)
--- 15829,15839 ----
  })
  
  (define_insn "*fscale_sfxf3"
!   [(set (match_operand:SF 0 "register_operand" "=f")
! 	 (unspec:SF [(match_operand 2 "register_operand" "0")
! 		     (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FSCALE))
!    (clobber (match_scratch:SF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)
***************
*** 15843,15853 ****
     (set_attr "mode" "SF")])
  
  (define_insn "*fscale_dfxf3"
!   [(parallel [(set (match_operand:DF 0 "register_operand" "=f")
! 		   (unspec:DF [(match_operand 2 "register_operand" "0")
! 			       (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FSCALE))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)
--- 15843,15853 ----
     (set_attr "mode" "SF")])
  
  (define_insn "*fscale_dfxf3"
!   [(set (match_operand:DF 0 "register_operand" "=f")
! 	 (unspec:DF [(match_operand 2 "register_operand" "0")
! 		     (match_operand 1 "register_operand" "u")]
  		    UNSPEC_FSCALE))
!    (clobber (match_scratch:DF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations
     && GET_MODE (operands[1]) == (TARGET_128BIT_LONG_DOUBLE ? TFmode : XFmode)
***************
*** 15857,15867 ****
     (set_attr "mode" "DF")])
  
  (define_insn "*fscale_xf3"
!   [(parallel [(set (match_operand:XF 0 "register_operand" "=f")
! 		   (unspec:XF [(match_operand:XF 2 "register_operand" "0")
! 			       (match_operand:XF 1 "register_operand" "u")]
! 		    UNSPEC_FSCALE))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
    "fscale\;fstp\t%y1"
--- 15857,15867 ----
     (set_attr "mode" "DF")])
  
  (define_insn "*fscale_xf3"
!   [(set (match_operand:XF 0 "register_operand" "=f")
! 	(unspec:XF [(match_operand:XF 2 "register_operand" "0")
! 		    (match_operand:XF 1 "register_operand" "u")]
! 	           UNSPEC_FSCALE))
!    (clobber (match_scratch:XF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
    "fscale\;fstp\t%y1"
***************
*** 15869,15879 ****
     (set_attr "mode" "XF")])
  
  (define_insn "*fscale_tf3"
!   [(parallel [(set (match_operand:TF 0 "register_operand" "=f")
! 		   (unspec:TF [(match_operand:TF 2 "register_operand" "0")
! 			       (match_operand:TF 1 "register_operand" "u")]
! 		    UNSPEC_FSCALE))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
    "fscale\;fstp\t%y1"
--- 15869,15879 ----
     (set_attr "mode" "XF")])
  
  (define_insn "*fscale_tf3"
!   [(set (match_operand:TF 0 "register_operand" "=f")
! 	(unspec:TF [(match_operand:TF 2 "register_operand" "0")
! 		    (match_operand:TF 1 "register_operand" "u")]
! 		   UNSPEC_FSCALE))
!    (clobber (match_scratch:TF 3 "=1"))]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
    "fscale\;fstp\t%y1"
***************
*** 15930,15936 ****
     (set (match_dup 9) (plus:XF (match_dup 7) (match_dup 8)))
     (parallel [(set (match_operand:SF 0 "register_operand" "")
  		   (unspec:SF [(match_dup 9) (match_dup 5)] UNSPEC_FSCALE))
! 	      (clobber (match_dup 5))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
--- 15930,15936 ----
     (set (match_dup 9) (plus:XF (match_dup 7) (match_dup 8)))
     (parallel [(set (match_operand:SF 0 "register_operand" "")
  		   (unspec:SF [(match_dup 9) (match_dup 5)] UNSPEC_FSCALE))
! 	      (clobber (match_scratch:SF 5 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
***************
*** 15960,15966 ****
     (set (match_dup 9) (plus:TF (match_dup 7) (match_dup 8)))
     (parallel [(set (match_operand:SF 0 "register_operand" "")
  		   (unspec:SF [(match_dup 9) (match_dup 5)] UNSPEC_FSCALE))
! 	      (clobber (match_dup 5))])]
    ""
  {
    rtx temp;
--- 15960,15966 ----
     (set (match_dup 9) (plus:TF (match_dup 7) (match_dup 8)))
     (parallel [(set (match_operand:SF 0 "register_operand" "")
  		   (unspec:SF [(match_dup 9) (match_dup 5)] UNSPEC_FSCALE))
! 	      (clobber (match_scratch:SF 5 ""))])]
    ""
  {
    rtx temp;
***************
*** 15983,15989 ****
     (set (match_dup 9) (plus:XF (match_dup 7) (match_dup 8)))
     (parallel [(set (match_operand:DF 0 "register_operand" "")
  		   (unspec:DF [(match_dup 9) (match_dup 5)] UNSPEC_FSCALE))
! 	      (clobber (match_dup 5))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
--- 15983,15989 ----
     (set (match_dup 9) (plus:XF (match_dup 7) (match_dup 8)))
     (parallel [(set (match_operand:DF 0 "register_operand" "")
  		   (unspec:DF [(match_dup 9) (match_dup 5)] UNSPEC_FSCALE))
! 	      (clobber (match_scratch:DF 5 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
***************
*** 16014,16020 ****
     (set (match_dup 9) (plus:TF (match_dup 7) (match_dup 8)))
     (parallel [(set (match_operand:DF 0 "register_operand" "")
  		   (unspec:DF [(match_dup 9) (match_dup 5)] UNSPEC_FSCALE))
! 	      (clobber (match_dup 5))])]
     ""
  {
    rtx temp;
--- 16014,16020 ----
     (set (match_dup 9) (plus:TF (match_dup 7) (match_dup 8)))
     (parallel [(set (match_operand:DF 0 "register_operand" "")
  		   (unspec:DF [(match_dup 9) (match_dup 5)] UNSPEC_FSCALE))
! 	      (clobber (match_scratch:DF 5 ""))])]
     ""
  {
    rtx temp;
***************
*** 16036,16042 ****
     (set (match_dup 8) (plus:XF (match_dup 6) (match_dup 7)))
     (parallel [(set (match_operand:XF 0 "register_operand" "")
  		   (unspec:XF [(match_dup 8) (match_dup 4)] UNSPEC_FSCALE))
! 	      (clobber (match_dup 4))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
  {
--- 16036,16042 ----
     (set (match_dup 8) (plus:XF (match_dup 6) (match_dup 7)))
     (parallel [(set (match_operand:XF 0 "register_operand" "")
  		   (unspec:XF [(match_dup 8) (match_dup 4)] UNSPEC_FSCALE))
! 	      (clobber (match_scratch:XF 5 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
  {
***************
*** 16072,16078 ****
     (set (match_dup 8) (plus:TF (match_dup 6) (match_dup 7)))
     (parallel [(set (match_operand:TF 0 "register_operand" "")
  		   (unspec:TF [(match_dup 8) (match_dup 4)] UNSPEC_FSCALE))
! 	      (clobber (match_dup 4))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
  {
--- 16072,16078 ----
     (set (match_dup 8) (plus:TF (match_dup 6) (match_dup 7)))
     (parallel [(set (match_operand:TF 0 "register_operand" "")
  		   (unspec:TF [(match_dup 8) (match_dup 4)] UNSPEC_FSCALE))
! 	      (clobber (match_scratch:TF 5 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
  {
***************
*** 16091,16097 ****
  		   (unspec:DF [(match_dup 2)
  			       (match_operand:DF 1 "register_operand" "")]
  		    UNSPEC_FPATAN))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
--- 16091,16097 ----
  		   (unspec:DF [(match_dup 2)
  			       (match_operand:DF 1 "register_operand" "")]
  		    UNSPEC_FPATAN))
! 	      (clobber (match_scratch:DF 3 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations"
  {
***************
*** 16104,16110 ****
  		   (unspec:XF [(match_dup 2)
  			       (match_operand:XF 1 "register_operand" "")]
  		    UNSPEC_FPATAN))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
  {
--- 16104,16110 ----
  		   (unspec:XF [(match_dup 2)
  			       (match_operand:XF 1 "register_operand" "")]
  		    UNSPEC_FPATAN))
! 	      (clobber (match_scratch:XF 3 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && !TARGET_128BIT_LONG_DOUBLE"
  {
***************
*** 16117,16123 ****
  		   (unspec:TF [(match_dup 2)
  			       (match_operand:TF 1 "register_operand" "")]
  		    UNSPEC_FPATAN))
! 	      (clobber (match_dup 1))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
  {
--- 16117,16123 ----
  		   (unspec:TF [(match_dup 2)
  			       (match_operand:TF 1 "register_operand" "")]
  		    UNSPEC_FPATAN))
! 	      (clobber (match_scratch:TF 3 ""))])]
    "! TARGET_NO_FANCY_MATH_387 && TARGET_80387
     && flag_unsafe_math_optimizations && TARGET_128BIT_LONG_DOUBLE"
  {
Index: config/pa/pa.md
===================================================================
RCS file: /cvs/gcc/gcc/gcc/config/pa/pa.md,v
retrieving revision 1.132
diff -c -3 -p -r1.132 pa.md
*** config/pa/pa.md	17 Oct 2003 03:32:21 -0000	1.132
--- config/pa/pa.md	18 Oct 2003 23:12:39 -0000
***************
*** 3016,3025 ****
--- 3016,3032 ----
  ;; therefore it is forced to operand 2.  If the count is compile-time
  ;; determined, we need two scratch registers for the unrolled code.
  (define_insn "movstrsi_internal"
+ <<<<<<< pa.md
+   [(set (mem:BLK (match_operand:SI 0 "register_operand" "r,r"))
+ 	(mem:BLK (match_operand:SI 1 "register_operand" "r,r")))
+    (clobber (match_scratch:SI 7 "=0,0")))
+    (clobber (match_scratch:SI 8 "=1,1"))
+ =======
    [(set (mem:BLK (match_operand:SI 0 "register_operand" "r,r"))
  	(mem:BLK (match_operand:SI 1 "register_operand" "r,r")))
     (clobber (match_scratch:SI 7 "=0,0"))
     (clobber (match_scratch:SI 8 "=1,1"))
+ >>>>>>> 1.132
     (clobber (match_operand:SI 2 "register_operand" "=r,r"))	;loop cnt/tmp
     (clobber (match_operand:SI 3 "register_operand" "=&r,&r"))	;item tmp
     (clobber (match_operand:SI 6 "register_operand" "=&r,&r"))	;item tmp2
Index: doc/invoke.texi
===================================================================
RCS file: /cvs/gcc/gcc/gcc/doc/invoke.texi,v
retrieving revision 1.349
diff -c -3 -p -r1.349 invoke.texi
*** doc/invoke.texi	18 Oct 2003 00:22:22 -0000	1.349
--- doc/invoke.texi	18 Oct 2003 23:12:40 -0000
*************** larger binaries.  Very high values are n
*** 4773,4794 ****
  binaries may adversely affect runtime performance.
  The default value is 200.
  
- @item max-inline-slope
- After exceeding the maximum number of inlined instructions by repeated
- inlining, a linear function is used to decrease the allowable size
- for single functions.  The slope of that function is the negative
- reciprocal of the number specified here.
- This parameter is ignored when @option{-funit-at-a-time} is used.
- The default value is 32.
- 
- @item min-inline-insns
- The repeated inlining is throttled more and more by the linear function
- after exceeding the limit.  To avoid too much throttling, a minimum for
- this function is specified here to allow repeated inlining for very small
- functions even when a lot of repeated inlining already has been done.
- This parameter is ignored when @option{-funit-at-a-time} is used.
- The default value is 10.
- 
  @item large-function-insns
  The limit specifying really large functions.  For functions greater than this
  limit inlining is constrained by @option{--param large-function-growth}.
--- 4773,4778 ----
Index: testsuite/g++.dg/opt/inline4.C
===================================================================
RCS file: /cvs/gcc/gcc/gcc/testsuite/g++.dg/opt/inline4.C,v
retrieving revision 1.2
diff -c -3 -p -r1.2 inline4.C
*** testsuite/g++.dg/opt/inline4.C	23 Jul 2003 16:45:15 -0000	1.2
--- testsuite/g++.dg/opt/inline4.C	18 Oct 2003 23:12:40 -0000
***************
*** 1,4 ****
! // { dg-options "-O2 -ftemplate-depth-20000 --param min-inline-insns=100 --param max-inline-insns=3" }
  
  template <int I>
  inline void g() { g<I-1>(); return; }
--- 1,4 ----
! // { dg-options "-O2 -ftemplate-depth-20000" }
  
  template <int I>
  inline void g() { g<I-1>(); return; }
Index: testsuite/gcc.dg/inline-2.c
===================================================================
RCS file: /cvs/gcc/gcc/gcc/testsuite/gcc.dg/inline-2.c,v
retrieving revision 1.4
diff -c -3 -p -r1.4 inline-2.c
*** testsuite/gcc.dg/inline-2.c	14 Mar 2003 17:49:12 -0000	1.4
--- testsuite/gcc.dg/inline-2.c	18 Oct 2003 23:12:42 -0000
*************** static int foo(void)
*** 11,17 ****
  
  int bar(void)
  {
!   return foo() + 1;
  }
  
  /* { dg-final { scan-assembler-not "jsr" { target alpha*-*-* } } } */
--- 11,18 ----
  
  int bar(void)
  {
!   /* Call twice to avoid bypassing the limit for functions called once.  */
!   return foo() + foo() + 1;
  }
  
  /* { dg-final { scan-assembler-not "jsr" { target alpha*-*-* } } } */


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
                   ` (9 preceding siblings ...)
  2003-10-18 23:21 ` hubicka at ucw dot cz
@ 2003-10-19 17:26 ` rth at redhat dot com
  2003-10-19 18:00 ` cvs-commit at gcc dot gnu dot org
                   ` (2 subsequent siblings)
  13 siblings, 0 replies; 15+ messages in thread
From: rth at redhat dot com @ 2003-10-19 17:26 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612



------- Additional Comments From rth at redhat dot com  2003-10-19 17:23 -------
Subject: Re:  [3.4 Regression] ICE when using log function with -O3 -ffast-math

Ok.



r~


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
                   ` (10 preceding siblings ...)
  2003-10-19 17:26 ` rth at redhat dot com
@ 2003-10-19 18:00 ` cvs-commit at gcc dot gnu dot org
  2003-10-19 18:04 ` pinskia at gcc dot gnu dot org
  2003-10-19 20:58 ` martin at mpa-garching dot mpg dot de
  13 siblings, 0 replies; 15+ messages in thread
From: cvs-commit at gcc dot gnu dot org @ 2003-10-19 18:00 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612



------- Additional Comments From cvs-commit at gcc dot gnu dot org  2003-10-19 18:00 -------
Subject: Bug 12612

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	hubicka@gcc.gnu.org	2003-10-19 18:00:13

Modified files:
	gcc            : ChangeLog reg-stack.c 
	gcc/config/i386: i386.c i386.md 

Log message:
	* i386.c (print_reg): Do not abort on certain registers.
	
	PR optimization/12612
	* reg-stack.c (subst_stack_regs_pat):  Use st(1) for clobbers.
	* i386.md (fpatan, fyl2x, fscale patterns and expanders): Use
	match_scratch; avoid bogus paralles.
	
	PR target/12674
	* i386.c (ix86_function_regparm): Disable implicit register passing
	conventions when profiling.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.1478&r2=2.1479
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/reg-stack.c.diff?cvsroot=gcc&r1=1.134&r2=1.135
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/i386/i386.c.diff?cvsroot=gcc&r1=1.608&r2=1.609
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/config/i386/i386.md.diff?cvsroot=gcc&r1=1.486&r2=1.487


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
                   ` (11 preceding siblings ...)
  2003-10-19 18:00 ` cvs-commit at gcc dot gnu dot org
@ 2003-10-19 18:04 ` pinskia at gcc dot gnu dot org
  2003-10-19 20:58 ` martin at mpa-garching dot mpg dot de
  13 siblings, 0 replies; 15+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2003-10-19 18:04 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612


pinskia at gcc dot gnu dot org changed:

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


------- Additional Comments From pinskia at gcc dot gnu dot org  2003-10-19 18:04 -------
Fixed for 3.4.


^ permalink raw reply	[flat|nested] 15+ messages in thread

* [Bug optimization/12612] [3.4 Regression] ICE when using log function with -O3 -ffast-math
  2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
                   ` (12 preceding siblings ...)
  2003-10-19 18:04 ` pinskia at gcc dot gnu dot org
@ 2003-10-19 20:58 ` martin at mpa-garching dot mpg dot de
  13 siblings, 0 replies; 15+ messages in thread
From: martin at mpa-garching dot mpg dot de @ 2003-10-19 20:58 UTC (permalink / raw)
  To: gcc-bugs

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=12612



------- Additional Comments From martin at mpa-garching dot mpg dot de  2003-10-19 20:34 -------
Hi Andrew, hi Jan,

my code runs fine again; thanks for the patch!


^ permalink raw reply	[flat|nested] 15+ messages in thread

end of thread, other threads:[~2003-10-19 20:34 UTC | newest]

Thread overview: 15+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2003-10-14  7:51 [Bug c++/12612] New: ICE when compiling with -O3 -ffast-math martin at mpa-garching dot mpg dot de
2003-10-14  7:53 ` [Bug c++/12612] " martin at mpa-garching dot mpg dot de
2003-10-14 16:22 ` [Bug optimization/12612] [3.4 Regression] " pinskia at gcc dot gnu dot org
2003-10-17 16:38 ` [Bug optimization/12612] [3.4 Regression] ICE when using log function " reichelt at gcc dot gnu dot org
2003-10-17 17:41 ` pinskia at gcc dot gnu dot org
2003-10-17 17:41 ` reichelt at gcc dot gnu dot org
2003-10-17 18:12 ` reichelt at gcc dot gnu dot org
2003-10-17 18:13 ` hubicka at ucw dot cz
2003-10-18 18:49 ` pinskia at gcc dot gnu dot org
2003-10-18 23:17 ` hubicka at ucw dot cz
2003-10-18 23:21 ` hubicka at ucw dot cz
2003-10-19 17:26 ` rth at redhat dot com
2003-10-19 18:00 ` cvs-commit at gcc dot gnu dot org
2003-10-19 18:04 ` pinskia at gcc dot gnu dot org
2003-10-19 20:58 ` martin at mpa-garching dot mpg dot de

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).