[Bug c/45127] New: Out-of-order execution

[Bug c/45127] New: Out-of-order execution

[aleksazr at gmail dot com]

This program is written for AT91SAM9260.
It is on compiled with yagarto with GCC 4.5.0.
on win xp sp2.

The prog reads 10 dwords from
address 0 and sends them through uart.

Adress 0 (on '9260) can either be ROM or SRAM,
depending on REMAP settings.

The prog first does a REMAP, then reads 10 dwords.

However, generated code first reads ptr[0],
does REMAP, then reads ptr[1-9], which means
that the first dword will be read from ROM,
and all others from SRAM.

"MATRIX_MRCR" is volatile, so it should be
executed first, IMHO.

It can be fixed by:

1. add volatile to ptr definition (line 31).

2. instead of linking with libarm.a, 
   change lines 7 (#if 1) and 40 (SendDword2).
   This method, however, produces less readable LSS file.

I've included two LSS files: original.lss and method2.lss
Original.lss is generated with the "as is" src.c.
Methot2.lss  is generated after changing lines 7 & 40.


-- 
           Summary: Out-of-order execution
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: c
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: aleksazr at gmail dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45127

[Bug c/45127] Out-of-order execution

[aleksazr at gmail dot com]

------- Comment #1 from aleksazr at gmail dot com  2010-07-29 13:13 -------
Created an attachment (id=21348)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=21348&action=view)
all sources

unpack to c:\ so that you have c:\00 folder


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45127

[Bug c/45127] Out-of-order execution

[rguenth at gcc dot gnu dot org]

------- Comment #2 from rguenth at gcc dot gnu dot org  2010-07-29 13:30 -------
rar is not a portable archive format.  Please use tar instead together with
gzip (or zip).

You might want to try -fno-delete-null-pointer-checks.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45127

[Bug target/45127] Out-of-order execution

[aleksazr at gmail dot com]

------- Comment #3 from aleksazr at gmail dot com  2010-07-29 14:56 -------
Created an attachment (id=21351)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=21351&action=view)
all sources


-- 

aleksazr at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #21348|0                           |1
        is obsolete|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45127

[Bug target/45127] Out-of-order execution

[rearnsha at gcc dot gnu dot org]

------- Comment #4 from rearnsha at gcc dot gnu dot org  2010-07-29 16:29 -------
Volatile alone won't prevent re-ordering of non-volatile memory operations. 
The volatile keyword only applies to that particular location (requiring the
compiler not to remove it, or change the number of accesses).

In this case you need to emit a memory barrier to prevent the compiler from
re-arranging the code.

A simple way to do this is to create an empty asm block that clobbers memory,
for example:

   asm  ("" : : : "memory");

this will cause the compiler to force out all pending memory operations before
the barrier and not to move later operations earlier.


-- 

rearnsha at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45127

[Bug target/45127] Out-of-order execution

[aleksazr at gmail dot com]

------- Comment #5 from aleksazr at gmail dot com  2010-07-29 19:19 -------
Thank you for a good explanation. Cheers!


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45127

[Bug target/45127] Out-of-order execution

[aleksazr at gmail dot com]

------- Comment #6 from aleksazr at gmail dot com  2010-07-30 13:56 -------
Anyway, the files can be used to generate poor listings,
and that is also a bug. See method2.lss


-- 

aleksazr at gmail dot com changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45127

[Bug target/45127] Out-of-order execution

[aleksazr at gmail dot com]

------- Comment #7 from aleksazr at gmail dot com  2010-07-30 14:00 -------
Poor listings = listings without debug info.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=45127