This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Question on PR36873
- From: Jie Zhang <jie dot zhang at analog dot com>
- To: gcc at gcc dot gnu dot org
- Cc: rguenther at suse dot de, Mike Frysinger <vapier dot adi at gmail dot com>
- Date: Wed, 23 Dec 2009 14:43:23 +0800
- Subject: Question on PR36873
Hi,
We just got a similar problem on Blackfin GCC recently. Let me take the
test code from the bug as an example:
typedef unsigned short u16;
typedef unsigned int u32;
u32 a(volatile u16* off) {
return *off;
}
u32 b(u16* off) {
return *off;
}
compiled with
mingw32-gcc-4.3.0.exe -c -O2 -fomit-frame-pointer -mtune=core2 test.c
it produces:
00000000 <_a>:
0: 8b 44 24 04 mov 0x4(%esp),%eax
4: 0f b7 00 movzwl (%eax),%eax
7: 0f b7 c0 movzwl %ax,%eax <== The redundant insn
a: c3 ret
00000010 <_b>:
10: 8b 44 24 04 mov 0x4(%esp),%eax
14: 0f b7 00 movzwl (%eax),%eax
17: c3 ret
I don't understand Richard's comment. What do we not optimize volatile
accesses in this test case. I know we cannot do many optimizations on
volatile accesses, but I think it's OK to remove the redundant insn in
this case. Could someone provide me a case in which we cannot remove it.
Thanks,
Jie