Here is the test.c file and two objdump output from compiled with -O1 and
-O2 using gcc3.4.6 cross compiler(host is x86 linux and target is PPC440)
respectively. It's obvious that the output from -O2 inlines Goo().
BTW, I compiled with -O2 using gcc 3.3.2 cross compiler on the same file
which doesnt do inline for Goo().
=============test.c======================
static void Goo()
{
char a[1024];
a[0]=21;
}
void Foo(int cmd)
{
char b[32];
switch(cmd) {
case 1:
b[0] = 13;
break;
case 2:
Goo();
break;
}
}
===================use -O2=========================
00000000 <Foo>:
0: 2f 83 00 01 cmpwi cr7,r3,1
4: 94 21 fb d0 stwu r1,-1072(r1)
8: 2f 03 00 02 cmpwi cr6,r3,2
c: 41 9e 00 20 beq- cr7,2c <Foo+0x2c>
10: 41 9a 00 0c beq- cr6,1c <Foo+0x1c>
14: 38 21 04 30 addi r1,r1,1072
18: 4e 80 00 20 blr
1c: 38 00 00 15 li r0,21
20: 98 01 00 30 stb r0,48(r1)
24: 38 21 04 30 addi r1,r1,1072
28: 4e 80 00 20 blr
2c: 38 00 00 0d li r0,13
30: 98 01 00 10 stb r0,16(r1)
34: 38 21 04 30 addi r1,r1,1072
38: 4e 80 00 20 blr
===================use -O1=========================
00000000 <Goo>:
0: 94 21 fb f0 stwu r1,-1040(r1)
4: 38 21 04 10 addi r1,r1,1040
8: 4e 80 00 20 blr
0000000c <Foo>:
c: 94 21 ff d0 stwu r1,-48(r1)
10: 2f 83 00 01 cmpwi cr7,r3,1
14: 40 be 00 0c bne+ cr7,20 <Foo+0x14>
18: 38 00 00 0d li r0,13
1c: 98 01 00 10 stb r0,16(r1)
20: 38 21 00 30 addi r1,r1,48
24: 4e 80 00 20 blr
===============================================
========================================================
Re: gcc 3.4.6 on PPC why doing function inline with -O2?
From: Andrew Haley <aph at redhat dot com>
To: "Ming Lei" <mlei at Brocade dot COM>
Cc: <gcc-help at gcc dot gnu dot org>
Date: Sat, 17 Feb 2007 10:36:05 +0000
Subject: Re: gcc 3.4.6 on PPC why doing function inline with -O2?
References:
<A5443C1E48CD114B9B1A92EF04F6656601795EF3@xxxxxxxxxxxxxxxxxxxxxxxxxx>
--------------------------------------------------------------------------------
Ming Lei writes:
> I came across a problem while using gcc3.4.6 on PPC440. It turns out
> with -O2 flag the compiler does inline function expanding. I tested with
> 3.2 version it doesn't do it in O2. The 3.4.6 manual says that only with
> -O3 the compiler can do this inline function optimization. Even I used
> -fno-inline-functions it doesn't work, but used -fno-inline does prevent
> inline functions.
>
> Is it a bug in 3.4.6?
I don't think so. Do you have a test case that shows this problem?
Andrew.
_________________________________________________________________
The average US Credit Score is 675. The cost to see yours: $0 by Experian.
http://www.freecreditreport.com/pm/default.aspx?sc=660600&bcd=EMAILFOOTERAVERAGE
