[Bug target/16691] New: arm-elf-gcc generates wrong-code

[Bug target/16691] New: arm-elf-gcc generates wrong-code

[gcc-bugzilla at gcc dot gnu dot org]

arm-elf-gcc generates wrong-code.

Why it use two 'strb', not one 'strh'?
And why it do 'ldrb'?
Because the 'glob_ptr' is the pointer to external device (16bit only,
volatile), both code could be a problem.

If I removed '__attribute__((__packed__))', gcc generates the code
what I expected.

The gcc (3.4.1, host compiler) and m6811-elf-gcc (3.4.1) generates
same assembly code with or without the '__attribute__((__packed__))'.

$ cat ex.c
struct a {
    volatile unsigned short v1;
} __attribute__((__packed__));

extern volatile void *glob_ptr;

void test(void)
{
    struct a *ptr = (struct a *) glob_ptr;
    ptr->v1 = 0xaa;
}

$ arm-elf-gcc -S -O2 ex.c
$ cat ex.s
	.file	"ex.c"
	.text
	.align	2
	.global	test
	.type	test, %function
test:
	@ args = 0, pretend = 0, frame = 0
	@ frame_needed = 0, uses_anonymous_args = 0
	@ link register save eliminated.
	ldr	r3, .L2
	ldr	r2, [r3, #0]
	mvn	r3, #85
	ldrb	r1, [r2, #0]	@ zero_extendqisi2
	strb	r3, [r2, #0]
	ldrb	r1, [r2, #1]	@ zero_extendqisi2
	add	r3, r3, #86
	@ lr needed for prologue
	strb	r3, [r2, #1]
	mov	pc, lr
.L3:
	.align	2
.L2:
	.word	glob_ptr
	.size	test, .-test
	.ident	"GCC: (GNU) 3.4.2 20040718 (prerelease)"

Environment:
System: Linux namsh 2.6.8-rc2 #91 Mon Jul 19 09:05:11 KST 2004 i686 athlon i386 GNU/Linux
Architecture: i686

	
host: i686-pc-linux-gnu
build: i686-pc-linux-gnu
target: arm-unknown-elf
configured with: ../configure --target=arm-elf --prefix=/usr/GNUARM --program-prefix=arm-elf- --with-gnu-ld --with-gnu-as --disable-shared --enable-languages=c,c++ --without-headers --with-newlib --disable-nls

How-To-Repeat:
	Just gerates the assembly code using the given ex.c.
------- Additional Comments From namsh at kldp dot org  2004-07-23 15:01 -------
Fix:
	Do not use '__attribute__((__packed__))'.
	Instead use '#pragma pack...'

-- 
           Summary: arm-elf-gcc generates wrong-code
           Product: gcc
           Version: 3.4.2
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: target
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: namsh at kldp dot org
                CC: gcc-bugs at gcc dot gnu dot org
 GCC build triplet: i686-pc-linux-gnu
  GCC host triplet: i686-pc-linux-gnu
GCC target triplet: arm-unknown-elf


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16691

[Bug target/16691] arm-elf-gcc generates wrong-code

[rearnsha at gcc dot gnu dot org]

------- Additional Comments From rearnsha at gcc dot gnu dot org  2004-07-23 16:12 -------
Invalid.  strh requires an aligned memory location.  If the object is unaligned
then two byte accesses are needed.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16691

[Bug target/16691] arm-elf-gcc generates wrong-code

[namsh at kldp dot org]

------- Additional Comments From namsh at kldp dot org  2004-07-24 02:35 -------
You mean arm-elf-gcc guesses its not aligned?
Then arm-elf-gcc's guess is wrong. Why it think memory is not aligned?

Actually memory is 2byte aligned. I already print the address.
And the addresses were 0xe0000aaa and 0xe0000554.

If this bug is really invalid, how I can avoid the problem with
__attribute__?


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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16691

[Bug target/16691] arm-elf-gcc generates wrong-code

[rearnsha at gcc dot gnu dot org]

------- Additional Comments From rearnsha at gcc dot gnu dot org  2004-07-24 12:50 -------
You wrote 'packed'.  Packed means it can be unaligned and the compiler must
treat it as so.  If it won't be unaligned, don't put packed on it.

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=16691