[Bug c/41531] New: -O1 -fschedule-insns swscale error

[Bug c/41531] New: -O1 -fschedule-insns swscale error

[t66667 at gmail dot com]

x86_64-w64-mingw32-gcc -DHAVE_AV_CONFIG_H -I.. -D_ISOC99_SOURCE
-D_POSIX_C_SOURCE=200112 -O1 -fschedule-insns -MMD -MF swscale.d -MT swscale.o
-save-temps -c -o swscale.o swscale.c

Error on both 4_4-branch and trunk
The error message of 4_4-branch is different:
swscale.c: In function 'sws_scale':
swscale.c:3199: error: unable to find a register to spill in class 'CREG'
swscale.c:3199: error: this is the insn:
(insn 446 466 432 53 swscale.c:3186 (parallel [
            (set (reg:SI 39 r10 [254])
                (ashiftrt:SI (reg:SI 39 r10 [orig:251 <variable>.dstH ] [251])
                    (subreg:QI (reg:SI 43 r14 [orig:255
<variable>.chrDstVSubSample ] [255]) 0)))
            (clobber (reg:CC 17 flags))
        ]) 523 {*ashrsi3_1} (expr_list:REG_DEAD (reg:SI 43 r14 [orig:255
<variable>.chrDstVSubSample ] [255])
        (expr_list:REG_DEAD (reg:SI 39 r10 [orig:251 <variable>.dstH ] [251])
            (expr_list:REG_UNUSED (reg:CC 17 flags)
                (expr_list:REG_EQUAL (ashiftrt:SI (mem/s/j:SI (plus:DI
(reg/v/f:DI 3 bx [orig:103 c ] [103])
                                (const_int 24 [0x18])) [0 <variable>.dstH+0 S4
A64])
                        (subreg:QI (reg:SI 43 r14 [orig:255
<variable>.chrDstVSubSample ] [255]) 0))
                    (nil))))))
swscale.c:3199: confused by earlier errors, bailing out

trunk:
swscale.s: Assembler messages:
swscale.s:19090: Error: suffix or operands invalid for `cmp'
swscale.s:19151: Error: suffix or operands invalid for `cmp'
swscale.s:19320: Error: suffix or operands invalid for `cmp'
swscale.s:19355: Error: suffix or operands invalid for `cmp'


-- 
           Summary: -O1 -fschedule-insns swscale error
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: t66667 at gmail dot com
 GCC build triplet: i486-slackware-linux
  GCC host triplet: x86_64-w64-mingw32
GCC target triplet: x86_64-w64-mingw32


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41531

[Bug c/41531] -O1 -fschedule-insns swscale error

[t66667 at gmail dot com]

------- Comment #1 from t66667 at gmail dot com  2009-10-01 06:50 -------
Created an attachment (id=18684)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18684&action=view)
Preprocessed test case.

gcc version 4.5.0 20090928 Revision 152257


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41531

[Bug target/41531] -O1 -fschedule-insns swscale error

[pinskia at gcc dot gnu dot org]

------- Comment #2 from pinskia at gcc dot gnu dot org  2009-10-01 06:55 -------
x86_64 and x86 has issues with pre-RA scheduling pass because of register
requirements of the target.

*** This bug has been marked as a duplicate of 24319 ***


-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
          Component|c                           |target
         Resolution|                            |DUPLICATE


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41531

[Bug target/41531] -O1 -fschedule-insns swscale error

[lucier at math dot purdue dot edu]

------- Comment #3 from lucier at math dot purdue dot edu  2009-10-01 13:19 -------
This is not the same problem as 24319.  Vlad thinks he fixed 24319, and indeed
the problem in this bug report from 4.4 is gone.  The reported problem in 4.5
is different.

Don't turn 234319 into a grab bag of any problem that arises when using
-fschedule-insns.

And, again, I can't reopen this bug.

Brad


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41531

[Bug target/41531] -O1 -fschedule-insns swscale error

[ubizjak at gmail dot com]

------- Comment #4 from ubizjak at gmail dot com  2009-10-01 16:28 -------
(In reply to comment #3)
> This is not the same problem as 24319.  Vlad thinks he fixed 24319, and indeed
> the problem in this bug report from 4.4 is gone.  The reported problem in 4.5
> is different.
> 
> Don't turn 234319 into a grab bag of any problem that arises when using
> -fschedule-insns.
> 
> And, again, I can't reopen this bug.

OK, ok, lets' reopen this PR and leave to Vlad to decide.

BTW: The test doesn't fail for x86_64-linux with or without -m32. Annoying.


-- 

ubizjak at gmail dot com changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41531

[Bug target/41531] -O1 -fschedule-insns swscale error

[ubizjak at gmail dot com]

------- Comment #5 from ubizjak at gmail dot com  2009-10-01 16:59 -------
(In reply to comment #3)
> This is not the same problem as 24319.  Vlad thinks he fixed 24319, and indeed
> the problem in this bug report from 4.4 is gone.  The reported problem in 4.5
> is different.

Oh, I have noticed that 4.5 has different problem, involving assembler errors.
>From attached assembler:

/APP
 # 912 "swscale_template.c" 1
        lea                     11*8+4*4*256(%rdi), %rdx  
        <stuff deleted ...>
>>>	cmp                                 %r15d, %rax  
        <more stuff deleted ...>
        jb                                  1b             

 # 0 "" 2
/NO_APP

This simply means invalid asm constraints or operand prefixes in asm.

Thus, WONTFIX for 4.4 and INVALID for 4.5.


-- 

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|UNCONFIRMED                 |RESOLVED
         Resolution|                            |INVALID


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41531