[Re: Unaligned access to packed structs on ppc405]

[Re: Unaligned access to packed structs on ppc405]

[Yaro Pollak]

> Yaro Pollak writes:

> I am cross-compiling on GCC 3.4.1 for the PowerPC 405, which can handle 
> unaligned accesses in hardware.
> Whenever I am accessing members of a packed bit-struct by pointer, the 
> compiler produces byte accesses, instead of 4-byte accesses.
> I've tried to use -mstrict-align but to no avail.

Hi everyone,
I still didn't resolve this problem, and frankly, I am lost.
I was using GCC 3.4.1, so I've compiled GCC 4.0.1 and tried to compile the code with it, and the results were exactly the same.
I've tried to use -mno-strict-align and -mstrict-align and there was no effect on the assembly produced.

The code that I'm trying to compile is 

struct Test {
	unsigned	b	: 4;
	unsigned	c	: 8;
	unsigned	d	: 6;
	unsigned	e	: 3;
	unsigned	f	: 7;
	unsigned	g	: 6;
	unsigned	h	: 2;
}__attribute__((__packed__));

int main (void) {
	Test* pt = (Test*)1234;
	return pt->b + pt->c + pt->d;
}

and the resulting assembly from g++ -mcpu=405 -O3 - S -c -mno-strict-align looks like:
	li 9,1234
	lbz 11,0(9)
	lbz 10,1(9)
	rlwinm 8,11,4,24,27
	srwi 0,10,4
	lbz 3,2(9)
	or 0,0,8
	rlwinm 11,11,28,28,31
	add 11,11,0
	rlwinm 10,10,2,26,29
	lwz 0,12(1)
	srwi 3,3,6
	or 3,3,10
	add 3,11,3
	mtlr 0
	addi 1,1,8
	blr 


Note the 3 lbz.
If on the other hand I compile:

	Test pt;
	return pt.b + pt.c + pt.d;

The assembly looks like this:
	lwz 3,8(1)
	rlwinm 0,3,4,28,31
	rlwinm 9,3,12,24,31
	add 0,0,9
	rlwinm 3,3,18,26,31
	add 3,0,3
	lwz 0,28(1)
	addi 1,1,24
	mtlr 0
	blr 


Does anyone have any idea about this? This is a rather trivial problem, I MUST be doing something wrong, please help me.

Thanks,
Yaro

Re: Unaligned access to packed structs on ppc405

[David Edelsohn]

>>>>> Yaro Pollak writes:

> If on the other hand I compile:
> 
>    Test pt;
>    return pt.b + pt.c + pt.d;
> 
>    lwz 3,8(1)

	That is an artifact of the example.  If I declare "pt" as a global
variable, GCC 4.1 prerelease produces three lbz.

	The lbz has to do with the size and the packed alignment.  With
the packed structure, GCC chooses the smallest memory access that covers
the bitfield.  Once GCC has chosen bytes, it cannot merge the loads
together.  If the structure were not declared packed, GCC would use wider
loads with masking, and then determine that the loads refer to the same
object. 

David