[4.2] MIPS: missed optimization

[Manuel Lauss <mano at roarinelk dot homelinux dot net>]

Hello,

Please consider this little snippett of code:

----- 8< ---------------------- 8< --------------

#define AU1000_INTC0_INT_BASE 8
#define IC0_FALLINGCLR	0xb0400078
#define IC0_RISINGCLR	0xb040007c

static inline void au_writel(unsigned long d, unsigned long a)
{
  *(unsigned long *)(a) = d;
}

void au1x_ic0_ack(unsigned int irq_nr)
{
  unsigned int bit = irq_nr - AU1000_INTC0_INT_BASE;
  au_writel(1 << bit, IC0_FALLINGCLR);
  au_writel(1 << bit, IC0_RISINGCLR);
}

----- 8< ---------------------- 8< --------------

cross-compiled for a MIPS32 environment (linux-kernel source), the
following output is generated:

Dump of assembler code for function au1x_ic0_ack:
0x80101140 <au1x_ic0_ack+0>:    addiu   a0,a0,-8
0x80101144 <au1x_ic0_ack+4>:    lui     v0,0xb040
0x80101148 <au1x_ic0_ack+8>:    li      v1,1
0x8010114c <au1x_ic0_ack+12>:   sllv    v1,v1,a0
0x80101150 <au1x_ic0_ack+16>:   ori     a0,v0,0x78
0x80101154 <au1x_ic0_ack+20>:   ori     v0,v0,0x7c
0x80101158 <au1x_ic0_ack+24>:   sw      v1,0(v0)
0x8010115c <au1x_ic0_ack+28>:   sw      v1,0(a0)
0x80101160 <au1x_ic0_ack+32>:   jr      ra
0x80101164 <au1x_ic0_ack+36>:   nop

My question is why does it generate the two ori instructions at +16 and +20
when it could just as easily add the offsets to the sw instructions?

Versions 4.3.2, 4.2.4 and even 3.4.6 generate this sequence;  there's no
difference between -O2 and -Os.

Admittedly my understanding of mips assembly is not yet very advanced, am I
missing something or is this a bug?


For reference, the full commandline is:

mipsel-softfloat-linux-gnu-gcc -Wp,-MD,arch/mips/alchemy/common/.irq.o.d
-nostdinc -isystem /usr/lib/gcc/mipsel-softfloat-linux-gnu/4.3.2/include
-D__KERNEL__ -Iinclude -Iinclude2
-I/mnt/work/au1200/kernel/linux-2.6.git/include
-I/mnt/work/au1200/kernel/linux-2.6.git/arch/mips/include -include include/linux/autoconf.h
-I/mnt/work/au1200/kernel/linux-2.6.git/arch/mips/alchemy/common
-Iarch/mips/alchemy/common -Wall -Wundef -Wstrict-prototypes -Wno-trigraphs
-fno-strict-aliasing -fno-common -Werror-implicit-function-declaration -Os
-mabi=32 -G 0 -mno-abicalls -fno-pic -pipe -msoft-float -ffreestanding
-march=mips32 -Wa,-mips32 -Wa,--trap
-I/mnt/work/au1200/kernel/linux-2.6.git/arch/mips/include/asm/mach-au1x00
-I/mnt/work/au1200/kernel/linux-2.6.git/arch/mips/include/asm/mach-exm1200
-Iarch/mips/include/asm/mach-exm1200
-I/mnt/work/au1200/kernel/linux-2.6.git/arch/mips/include/asm/mach-generic
-D"VMLINUX_LOAD_ADDRESS=0xffffffff80100000" -fno-stack-protector
-fomit-frame-pointer -g -Wdeclaration-after-statement -Wno-pointer-sign
-Werror  -D"KBUILD_STR(s)=#s" -D"KBUILD_BASENAME=KBUILD_STR(irq)"
-D"KBUILD_MODNAME=KBUILD_STR(irq)"  -c -o arch/mips/alchemy/common/irq.o
/mnt/work/au1200/kernel/linux-2.6.git/arch/mips/alchemy/common/irq.c

Target: mipsel-softfloat-linux-gnu
Configured with:
/tmp-ram/portage/cross-mipsel-softfloat-linux-gnu/gcc-4.3.2/work/gcc-4.3.2/configure
--prefix=/usr/
--bindir=/usr/x86_64-pc-linux-gnu/mipsel-softfloat-linux-gnu/gcc-bin/4.3.2
--includedir=/usr/lib/gcc/mipsel-softfloat-linux-gnu/4.3.2/include/
--datadir=/usr/share/gcc-data/mipsel-softfloat-linux-gnu/4.3.2
--mandir=/usr/share/gcc-data/mipsel-softfloat-linux-gnu/4.3.2/man/
--infodir=/usr/share/gcc-data/mipsel-softfloat-linux-gnu/4.3.2/info
--with-gxx-include-dir=/usr/lib/gcc/mipsel-softfloat-linux-gnu/4.3.2/include/g++-v4/
--host=x86_64-pc-linux-gnu --target=mipsel-softfloat-linux-gnu
--build=x86_64-pc-linux-gnu --disable-altivec --with-float=soft/
--disable-nls --with-system-zlib --disable-checking --disable-werror
--enable-secureplt --disable-multilib --disable-libmudflap --disable-libssp/
--enable-cld --disable-libgcj --enable-languages=c,c++
--with-sysroot=/usr/mipsel-softfloat-linux-gnu --disable-bootstrap/
--disable-libgomp --enable-__cxa_atexit --enable-clocale=gnu
--with-bugurl=http://bugs.gentoo.org/ --with-pkgversion='Gentoo 4.3.2 p1.0'
Thread model: posix
gcc version 4.3.2 (Gentoo 4.3.2 p1.0)


Thanks,
	Manuel Lauss