[Bug target/19683] New: MIPS wrong-code for 64-bit multiply from java program.

[Bug target/19683] New: MIPS wrong-code for 64-bit multiply from java program.

[daney at gcc dot gnu dot org]

The testcase that I will attach when compiled with -O3 generates incorrect
results.  Good results are obtained at -O0

The testcase should be compiled like this:

$ mipsel-linux-gcj --main=TestMultSimple -O2 TestMultSimple.java -o
TestMultSimple-O2

$ mipsel-linux-gcj --main=TestMultSimple -O0 TestMultSimple.java -o
TestMultSimple-O0

When running I get this:
/junk # ./TestMultSimple-O2
test10a3
off    = -28800000      = 0800
soff   = 1106931389806  = 2005.01.28 08:56:29.806
syscur = 1106960189806  = 2005.01.28 16:56:29.806
d      = -5287802103562112000   = 7090.-33554411.-33554413 23:51:28.000
/junk # ./TestMultSimple-O0
test10a3
off    = -28800000      = 0800
soff   = 1106931392290  = 2005.01.28 08:56:32.290
syscur = 1106960192290  = 2005.01.28 16:56:32.290
d      = 1106870400000  = 2005.01.27 16:00:00.000

The second result is correct.  d should be a positive number of about the same
value as syscur.

When build with x86 targeted compiler the bug is not seen.
gcj (GCC) 3.4.2 20041017 (Red Hat 3.4.2-6.fc3) for example gives good results.

-- 
           Summary: MIPS wrong-code for 64-bit multiply from java program.
           Product: gcc
           Version: 3.4.3
            Status: UNCONFIRMED
          Keywords: wrong-code
          Severity: normal
          Priority: P2
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: daney at gcc dot gnu dot org
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: i686-pc-linux
  GCC host triplet: i686-pc-linux
GCC target triplet: mipsel-linux


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug target/19683] MIPS wrong-code for 64-bit multiply from java program.

[daney at gcc dot gnu dot org]

------- Additional Comments From daney at gcc dot gnu dot org  2005-01-29 01:00 -------
Created an attachment (id=8096)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=8096&action=view)
Testcase


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug target/19683] MIPS wrong-code for 64-bit multiply from java program.

[daney at gcc dot gnu dot org]

------- Additional Comments From daney at gcc dot gnu dot org  2005-01-29 01:16 -------
I should probably add the the compiler was configured with:

configure --target=mipsel-linux --with-sysroot=/usr/local/mipsel-linux
--with-arch=mips32 --with-float=soft --enable-languages=c,c++,java
--prefix=/usr/local/mipsel-linux --with-system-zlib --disable-java-awt
--without-x --disable-jvmpi

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug target/19683] MIPS wrong-code for 64-bit multiply from java program.

[daney at gcc dot gnu dot org]

------- Additional Comments From daney at gcc dot gnu dot org  2005-02-22 21:40 -------
Created an attachment (id=8257)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=8257&action=view)
C testcase


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug target/19683] MIPS wrong-code for 64-bit multiply from java program.

[daney at gcc dot gnu dot org]

------- Additional Comments From daney at gcc dot gnu dot org  2005-02-22 21:41 -------
Created an attachment (id=8258)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=8258&action=view)
C testcase driver.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug target/19683] MIPS wrong-code for 64-bit multiply.

[daney at gcc dot gnu dot org]

------- Additional Comments From daney at gcc dot gnu dot org  2005-02-22 22:02 -------
The new 'C' test case will reproduce the bug on:

$ mipsel-linux-gcc -v
Using built-in specs.
Target: mipsel-linux
Configured with: ../gcccvsmain/gcc/configure --target=mipsel-linux
--with-sysroot=/usr/local/mipsel-linux-test --with-arch=mips32 --with-float=soft
--prefix=/usr/local/mipsel-linux-test --with-system-zlib
--enable-languages=c,c++,java
Thread model: posix
gcc version 4.0.0 20050222 (experimental)

I have not been able to generate a 'C' test case that fails on 3.4.x.  A similar
java testcase failed on 3.3.1, but I have not tried the attached java case on 3.3.1.

Here is the setup for failure:
$ mipsel-linux-gcc -O3 -c tm.c
$ mipsel-linux-gcc -O3 -c tm_driver.c
$ mipsel-linux-gcc -o tm tm_driver.o tm.o

Then on the host:
#./tm
The answer is: 3989327253662920704

Here is the setup for success:
$ mipsel-linux-gcc -O0 -c tm.c
$ mipsel-linux-gcc -O3 -c tm_driver.c
$ mipsel-linux-gcc -o tm tm_driver.o tm.o

Then on the host:
# ./tm
The answer is: 1109030400000


If we look at the relevent part of the disassembly we see:

        li      $6,86376448                     # 0x5260000
        ori     $6,$6,0x5c00
        multu   $2,$6
        lw      $28,16($sp)
        li      $5,92340224                     # 0x5810000
        ori     $5,$5,0x1392
        lw      $25,%call16(f4)($28)
        mflo    $4
        sw      $4,24($sp)
        lw      $4,%got($LC0)($28)
        madd    $6,$3

The problem is that before the madd, we need to copy the contents of hi -> low
and then clear hi to zero.  This is not happening.




-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|                            |4.0.0 3.4.3 3.3.1
            Summary|MIPS wrong-code for 64-bit  |MIPS wrong-code for 64-bit
                   |multiply from java program. |multiply.


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug target/19683] MIPS wrong-code for 64-bit multiply.

[daney at gcc dot gnu dot org]

------- Additional Comments From daney at gcc dot gnu dot org  2005-02-22 22:06 -------
I should comment that yes I have tried to simplify tm.c.  This is the best I
could do.  It seems that register pressure is important to force the compiler
into the failure mode, as in simpler cases the madd form of the multiply is
never generated.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug target/19683] MIPS wrong-code for 64-bit multiply.

[rsandifo at gcc dot gnu dot org]

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |rsandifo at gcc dot gnu dot
                   |dot org                     |org
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|                            |1
   Last reconfirmed|0000-00-00 00:00:00         |2005-02-22 22:40:58
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug target/19683] MIPS wrong-code for 64-bit multiply.

[rsandifo at redhat dot com]

------- Additional Comments From rsandifo at redhat dot com  2005-02-24 22:37 -------
Subject: Re:  MIPS wrong-code for 64-bit multiply.

David Daney <ddaney@avtrex.com> writes:
> Do you think that patch can/should be backported to 3.3.x and 3.4.x?

If it works. ;)  I think it's one of those that has to prove itself
on mainline for a week or so first.

Richard


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug rtl-optimization/19683] MIPS wrong-code for 64-bit multiply.

[rsandifo at gcc dot gnu dot org]

------- Additional Comments From rsandifo at gcc dot gnu dot org  2005-02-26 10:16 -------
Patch posted here:

    http://gcc.gnu.org/ml/gcc-patches/2005-02/msg01665.html


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
          Component|target                      |rtl-optimization
           Keywords|                            |patch


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug rtl-optimization/19683] MIPS wrong-code for 64-bit multiply.

[daney at gcc dot gnu dot org]

------- Additional Comments From daney at gcc dot gnu dot org  2005-03-03 21:19 -------
Thanks Richard,

The patch applies cleanly against 3.4.3 (with an offset of something like 9
lines).  I found no regressions with a top level make -k check for an i686-linux
-> mipsel-linux cross compiler with the patch applied to 3.4.3.  I have been
using the patch for about a week now on 3.4.3 with no problems found.

-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug rtl-optimization/19683] MIPS wrong-code for 64-bit multiply.

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-03-07 17:48 -------
Subject: Bug 19683

CVSROOT:	/cvs/gcc
Module name:	gcc
Changes by:	rsandifo@gcc.gnu.org	2005-03-07 17:48:47

Modified files:
	gcc            : ChangeLog reload1.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.dg/torture: pr19683-1.c 

Log message:
	PR rtl-optimization/19683
	* reload1.c (choose_reload_regs): Pass the number of bits, not the
	number of bytes, to smallest_int_for_mode.  Fix arguments to
	REG_CANNOT_CHANGE_MODE_P.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&r1=2.7711&r2=2.7712
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/reload1.c.diff?cvsroot=gcc&r1=1.462&r2=1.463
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&r1=1.5118&r2=1.5119
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/torture/pr19683-1.c.diff?cvsroot=gcc&r1=NONE&r2=1.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug rtl-optimization/19683] MIPS wrong-code for 64-bit multiply.

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-03-07 17:51 -------
Subject: Bug 19683

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-4_0-branch
Changes by:	rsandifo@gcc.gnu.org	2005-03-07 17:51:36

Modified files:
	gcc            : ChangeLog reload1.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.dg/torture: pr19683-1.c 

Log message:
	PR rtl-optimization/19683
	* reload1.c (choose_reload_regs): Pass the number of bits, not the
	number of bytes, to smallest_int_for_mode.  Fix arguments to
	REG_CANNOT_CHANGE_MODE_P.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=2.7592.2.30&r2=2.7592.2.31
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/reload1.c.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.461&r2=1.461.6.1
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=1.5084.2.22&r2=1.5084.2.23
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/torture/pr19683-1.c.diff?cvsroot=gcc&only_with_tag=gcc-4_0-branch&r1=NONE&r2=1.1.2.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug rtl-optimization/19683] MIPS wrong-code for 64-bit multiply.

[rsandifo at gcc dot gnu dot org]

------- Additional Comments From rsandifo at gcc dot gnu dot org  2005-03-07 17:54 -------
Patch applied to mainline and 4.0.  Will apply to 3.4 branch
in a week or so if nothing goes wrong.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
      Known to fail|4.0.0 3.4.3 3.3.1           |3.4.3 3.3.1
      Known to work|                            |4.1.0 4.0.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug rtl-optimization/19683] MIPS wrong-code for 64-bit multiply.

[cvs-commit at gcc dot gnu dot org]

------- Additional Comments From cvs-commit at gcc dot gnu dot org  2005-03-17 21:12 -------
Subject: Bug 19683

CVSROOT:	/cvs/gcc
Module name:	gcc
Branch: 	gcc-3_4-branch
Changes by:	rsandifo@gcc.gnu.org	2005-03-17 21:11:49

Modified files:
	gcc            : ChangeLog reload1.c 
	gcc/testsuite  : ChangeLog 
Added files:
	gcc/testsuite/gcc.dg/torture: pr19683-1.c 

Log message:
	PR rtl-optimization/19683
	* reload1.c (choose_reload_regs): Pass the number of bits, not the
	number of bytes, to smallest_int_for_mode.  Fix arguments to
	REG_CANNOT_CHANGE_MODE_P.

Patches:
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=2.2326.2.820&r2=2.2326.2.821
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/reload1.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.419.4.7&r2=1.419.4.8
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/ChangeLog.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=1.3389.2.371&r2=1.3389.2.372
http://gcc.gnu.org/cgi-bin/cvsweb.cgi/gcc/gcc/testsuite/gcc.dg/torture/pr19683-1.c.diff?cvsroot=gcc&only_with_tag=gcc-3_4-branch&r1=NONE&r2=1.1.6.1



-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683

[Bug rtl-optimization/19683] MIPS wrong-code for 64-bit multiply.

[rsandifo at gcc dot gnu dot org]

------- Additional Comments From rsandifo at gcc dot gnu dot org  2005-03-17 21:13 -------
Committed to 3.4 as promised, after bootstrapping & regression
testing on mips-sgi-irix6.5.


-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |RESOLVED
      Known to fail|3.4.3 3.3.1                 |3.3.1
      Known to work|4.1.0 4.0.0                 |4.1.0 4.0.0 3.4.4
         Resolution|                            |FIXED
   Target Milestone|---                         |3.4.4


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=19683