diff mbox

[RFC] arm: use built-in byte swap function

Message ID 20130221203327.6558f89277468f7ffffa6506@freescale.com (mailing list archive)
State New, archived
Headers show

Commit Message

Kim Phillips Feb. 22, 2013, 2:33 a.m. UTC 
On Thu, 21 Feb 2013 11:40:54 -0500
Nicolas Pitre <nico@fluxnic.net> wrote:

> On Thu, 21 Feb 2013, Kim Phillips wrote:
> 
> > On Wed, 20 Feb 2013 23:29:58 -0500
> > Nicolas Pitre <nico@fluxnic.net> wrote:
> > 
> > > On Wed, 20 Feb 2013, Kim Phillips wrote:
> > > 
> > > > On Wed, 20 Feb 2013 10:43:18 -0500
> > > > Nicolas Pitre <nico@fluxnic.net> wrote:
> > > > 
> > > > > On Wed, 20 Feb 2013, Woodhouse, David wrote:
> > > > > > On Wed, 2013-02-20 at 09:06 -0500, Nicolas Pitre wrote:
> > > > > > > ... in which case there is no harm shipping a .c file and trivially 
> > > > > > > enforcing -O2, the rest being equal.
> > > > > > 
> > > > > > For today's compilers, unless the wind changes.
> > > > > 
> > > > > We'll adapt if necessary.  Going with -O2 should remain pretty safe anyway.
> > > > 
> > > > Alas, not so for gcc 4.4 - I had forgotten I had tested
> > > > Ubuntu/Linaro 4.4.7-1ubuntu2 here:
> > > > 
> > > > https://patchwork.kernel.org/patch/2101491/
> > > > 
> > > > add -O2 to that test script and gcc 4.4 *always* emits calls to
> > > > __bswap[sd]i2, even with -march=armv6k+.
> > 
> > argh, sorry - that script was testing support for 
> > __builtin_bswap{16,32,64} directly, which isn't the same as testing
> > code generation of a byte swap pattern in C.
> 
> Still, I'm not as confident as I was about this.

which part exactly?  Having -O2 as "protection"?  Yes, me neither.

> > I'll still try the assembly approach - gcc 4.4's armv6 output looks
> > worse than both the pre-armv6 and post-armv6 __arch_swab32
> > implementations currently in use:
> > 
> > mov     ip, sp
> > push    {fp, ip, lr, pc}
> > sub     fp, ip, #4
> 
> You should use -fomit-frame-pointer to compile this.  We don't need a 
> frame pointer here, especially for a leaf function that the compiler 
> decides to call on its own.
> 
> > and     r2, r0, #65280  ; 0xff00
> > lsl     ip, r0, #24
> > orr     r1, ip, r0, lsr #24
> > and     r0, r0, #16711680       ; 0xff0000
> > orr     r3, r1, r2, lsl #8
> > orr     r0, r3, r0, lsr #8
> 
> Other than that, it is true that the above is slightly suboptimal.

Here's the asm version I'm working on now, based on compiler
output of the C version.  Haven't tested beyond defconfig builds,
which pass ok.

Is there anything I have to do for thumb mode?  If so, how to test?


Thanks,

Kim

Comments

Nicolas Pitre Feb. 22, 2013, 3:40 a.m. UTC | #1 
On Thu, 21 Feb 2013, Kim Phillips wrote:

> Here's the asm version I'm working on now, based on compiler
> output of the C version.  Haven't tested beyond defconfig builds,
> which pass ok.
> 
> Is there anything I have to do for thumb mode?  If so, how to test?

You just need to pick a config that uses some ARMv7 processor, and 
enable CONFIG_THUMB2_KERNEL.  I don't see any problem with your patch 
wrt Thumb2.

Still, I have minor comments below.

> diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
> index dedf02b..e8a41d0 100644
> --- a/arch/arm/Kconfig
> +++ b/arch/arm/Kconfig
> @@ -59,6 +59,7 @@ config ARM
>  	select CLONE_BACKWARDS
>  	select OLD_SIGSUSPEND3
>  	select OLD_SIGACTION
> +	select ARCH_USE_BUILTIN_BSWAP
>  	help
>  	  The ARM series is a line of low-power-consumption RISC chip designs
>  	  licensed by ARM Ltd and targeted at embedded applications and
> diff --git a/arch/arm/boot/compressed/Makefile b/arch/arm/boot/compressed/Makefile
> index 5cad8a6..a277e97 100644
> --- a/arch/arm/boot/compressed/Makefile
> +++ b/arch/arm/boot/compressed/Makefile
> @@ -108,12 +108,12 @@ endif
>  
>  targets       := vmlinux vmlinux.lds \
>  		 piggy.$(suffix_y) piggy.$(suffix_y).o \
> -		 lib1funcs.o lib1funcs.S ashldi3.o ashldi3.S \
> +		 lib1funcs.o lib1funcs.S ashldi3.o ashldi3.S bswapsdi2.o \

Should be both bswapsdi2.o bswapsdi2.S

>  		 font.o font.c head.o misc.o $(OBJS)
>  
>  # Make sure files are removed during clean
>  extra-y       += piggy.gzip piggy.lzo piggy.lzma piggy.xzkern \
> -		 lib1funcs.S ashldi3.S $(libfdt) $(libfdt_hdrs)
> +		 lib1funcs.S ashldi3.S bswapsdi2.o $(libfdt) $(libfdt_hdrs)

Should be bswapsdi2.S.

>  ifeq ($(CONFIG_FUNCTION_TRACER),y)
>  ORIG_CFLAGS := $(KBUILD_CFLAGS)
> @@ -155,6 +155,12 @@ ashldi3 = $(obj)/ashldi3.o
>  $(obj)/ashldi3.S: $(srctree)/arch/$(SRCARCH)/lib/ashldi3.S
>  	$(call cmd,shipped)
>  
> +# For __bswapsi2, __bswapdi2
> +bswapsdi2 = $(obj)/bswapsdi2.o
> +
> +$(obj)/bswapsdi2.S: $(srctree)/arch/$(SRCARCH)/lib/bswapsdi2.S
> +	$(call cmd,shipped)
> +
>  # We need to prevent any GOTOFF relocs being used with references
>  # to symbols in the .bss section since we cannot relocate them
>  # independently from the rest at run time.  This can be achieved by
> @@ -176,7 +182,8 @@ if [ $(words $(ZRELADDR)) -gt 1 -a "$(CONFIG_AUTO_ZRELADDR)" = "" ]; then \
>  fi
>  
>  $(obj)/vmlinux: $(obj)/vmlinux.lds $(obj)/$(HEAD) $(obj)/piggy.$(suffix_y).o \
> -		$(addprefix $(obj)/, $(OBJS)) $(lib1funcs) $(ashldi3) FORCE
> +		$(addprefix $(obj)/, $(OBJS)) $(lib1funcs) $(ashldi3) \
> +		$(bswapsdi2) FORCE
>  	@$(check_for_multiple_zreladdr)
>  	$(call if_changed,ld)
>  	@$(check_for_bad_syms)
> diff --git a/arch/arm/kernel/armksyms.c b/arch/arm/kernel/armksyms.c
> index 60d3b73..ba578f7 100644
> --- a/arch/arm/kernel/armksyms.c
> +++ b/arch/arm/kernel/armksyms.c
> @@ -35,6 +35,8 @@ extern void __ucmpdi2(void);
>  extern void __udivsi3(void);
>  extern void __umodsi3(void);
>  extern void __do_div64(void);
> +extern void __bswapsi2(void);
> +extern void __bswapdi2(void);
>  
>  extern void __aeabi_idiv(void);
>  extern void __aeabi_idivmod(void);
> @@ -114,6 +116,8 @@ EXPORT_SYMBOL(__ucmpdi2);
>  EXPORT_SYMBOL(__udivsi3);
>  EXPORT_SYMBOL(__umodsi3);
>  EXPORT_SYMBOL(__do_div64);
> +EXPORT_SYMBOL(__bswapsi2);
> +EXPORT_SYMBOL(__bswapdi2);
>  
>  #ifdef CONFIG_AEABI
>  EXPORT_SYMBOL(__aeabi_idiv);
> diff --git a/arch/arm/lib/Makefile b/arch/arm/lib/Makefile
> index af72969..5383df7 100644
> --- a/arch/arm/lib/Makefile
> +++ b/arch/arm/lib/Makefile
> @@ -13,7 +13,7 @@ lib-y		:= backtrace.o changebit.o csumipv6.o csumpartial.o   \
>  		   ashldi3.o ashrdi3.o lshrdi3.o muldi3.o             \
>  		   ucmpdi2.o lib1funcs.o div64.o                      \
>  		   io-readsb.o io-writesb.o io-readsl.o io-writesl.o  \
> -		   call_with_stack.o
> +		   call_with_stack.o bswapsdi2.o
>  
>  mmu-y	:= clear_user.o copy_page.o getuser.o putuser.o
>  
> diff --git a/arch/arm/lib/bswapsdi2.S b/arch/arm/lib/bswapsdi2.S
> new file mode 100644
> index 0000000..e9c8ca7
> --- /dev/null
> +++ b/arch/arm/lib/bswapsdi2.S
> @@ -0,0 +1,36 @@
> +#include <linux/linkage.h>
> +
> +#if __LINUX_ARM_ARCH__ >= 6
> +ENTRY(__bswapsi2)
> +	rev	r0, r0
> +	bx	lr
> +ENDPROC(__bswapsi2)
> +
> +ENTRY(__bswapdi2)
> +	rev	r3, r0
> +	rev	r0, r1
> +	mov	r1, r3
> +	bx	lr
> +ENDPROC(__bswapdi2)
> +#else
> +ENTRY(__bswapsi2)
> +	eor     r3, r0, r0, ror #16
> +	lsr     r3, r3, #8

Some older binutils used with pre ARMv6 platforms don't understand the 
latest unified syntax.  So in this case it is better to use:

	mov	r3, r3, lsr #8

> +	bic     r3, r3, #65280  @ 0xff00

Please use #0xff00 directly rather than keeping it as a comment.

> +	eor     r0, r3, r0, ror #8
> +	mov     pc, lr
> +ENDPROC(__bswapsi2)
> +
> +ENTRY(__bswapdi2)
> +	mov     ip, r1
> +	eor     r3, ip, ip, ror #16
> +	eor     r1, r0, r0, ror #16
> +	lsr     r1, r1, #8
> +	lsr     r3, r3, #8
> +	bic     r3, r3, #65280  @ 0xff00
> +	bic     r1, r1, #65280  @ 0xff00

Same comments for the 4 instructions above.

> +	eor     r1, r1, r0, ror #8
> +	eor     r0, r3, ip, ror #8
> +	mov     pc, lr
> +ENDPROC(__bswapdi2)
> +#endif
> 
> Thanks,
> 
> Kim
>
diff mbox

Patch

diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
index dedf02b..e8a41d0 100644
--- a/arch/arm/Kconfig
+++ b/arch/arm/Kconfig
@@ -59,6 +59,7 @@  config ARM
 	select CLONE_BACKWARDS
 	select OLD_SIGSUSPEND3
 	select OLD_SIGACTION
+	select ARCH_USE_BUILTIN_BSWAP
 	help
 	  The ARM series is a line of low-power-consumption RISC chip designs
 	  licensed by ARM Ltd and targeted at embedded applications and
diff --git a/arch/arm/boot/compressed/Makefile b/arch/arm/boot/compressed/Makefile
index 5cad8a6..a277e97 100644
--- a/arch/arm/boot/compressed/Makefile
+++ b/arch/arm/boot/compressed/Makefile
@@ -108,12 +108,12 @@  endif
 
 targets       := vmlinux vmlinux.lds \
 		 piggy.$(suffix_y) piggy.$(suffix_y).o \
-		 lib1funcs.o lib1funcs.S ashldi3.o ashldi3.S \
+		 lib1funcs.o lib1funcs.S ashldi3.o ashldi3.S bswapsdi2.o \
 		 font.o font.c head.o misc.o $(OBJS)
 
 # Make sure files are removed during clean
 extra-y       += piggy.gzip piggy.lzo piggy.lzma piggy.xzkern \
-		 lib1funcs.S ashldi3.S $(libfdt) $(libfdt_hdrs)
+		 lib1funcs.S ashldi3.S bswapsdi2.o $(libfdt) $(libfdt_hdrs)
 
 ifeq ($(CONFIG_FUNCTION_TRACER),y)
 ORIG_CFLAGS := $(KBUILD_CFLAGS)
@@ -155,6 +155,12 @@  ashldi3 = $(obj)/ashldi3.o
 $(obj)/ashldi3.S: $(srctree)/arch/$(SRCARCH)/lib/ashldi3.S
 	$(call cmd,shipped)
 
+# For __bswapsi2, __bswapdi2
+bswapsdi2 = $(obj)/bswapsdi2.o
+
+$(obj)/bswapsdi2.S: $(srctree)/arch/$(SRCARCH)/lib/bswapsdi2.S
+	$(call cmd,shipped)
+
 # We need to prevent any GOTOFF relocs being used with references
 # to symbols in the .bss section since we cannot relocate them
 # independently from the rest at run time.  This can be achieved by
@@ -176,7 +182,8 @@  if [ $(words $(ZRELADDR)) -gt 1 -a "$(CONFIG_AUTO_ZRELADDR)" = "" ]; then \
 fi
 
 $(obj)/vmlinux: $(obj)/vmlinux.lds $(obj)/$(HEAD) $(obj)/piggy.$(suffix_y).o \
-		$(addprefix $(obj)/, $(OBJS)) $(lib1funcs) $(ashldi3) FORCE
+		$(addprefix $(obj)/, $(OBJS)) $(lib1funcs) $(ashldi3) \
+		$(bswapsdi2) FORCE
 	@$(check_for_multiple_zreladdr)
 	$(call if_changed,ld)
 	@$(check_for_bad_syms)
diff --git a/arch/arm/kernel/armksyms.c b/arch/arm/kernel/armksyms.c
index 60d3b73..ba578f7 100644
--- a/arch/arm/kernel/armksyms.c
+++ b/arch/arm/kernel/armksyms.c
@@ -35,6 +35,8 @@  extern void __ucmpdi2(void);
 extern void __udivsi3(void);
 extern void __umodsi3(void);
 extern void __do_div64(void);
+extern void __bswapsi2(void);
+extern void __bswapdi2(void);
 
 extern void __aeabi_idiv(void);
 extern void __aeabi_idivmod(void);
@@ -114,6 +116,8 @@  EXPORT_SYMBOL(__ucmpdi2);
 EXPORT_SYMBOL(__udivsi3);
 EXPORT_SYMBOL(__umodsi3);
 EXPORT_SYMBOL(__do_div64);
+EXPORT_SYMBOL(__bswapsi2);
+EXPORT_SYMBOL(__bswapdi2);
 
 #ifdef CONFIG_AEABI
 EXPORT_SYMBOL(__aeabi_idiv);
diff --git a/arch/arm/lib/Makefile b/arch/arm/lib/Makefile
index af72969..5383df7 100644
--- a/arch/arm/lib/Makefile
+++ b/arch/arm/lib/Makefile
@@ -13,7 +13,7 @@  lib-y		:= backtrace.o changebit.o csumipv6.o csumpartial.o   \
 		   ashldi3.o ashrdi3.o lshrdi3.o muldi3.o             \
 		   ucmpdi2.o lib1funcs.o div64.o                      \
 		   io-readsb.o io-writesb.o io-readsl.o io-writesl.o  \
-		   call_with_stack.o
+		   call_with_stack.o bswapsdi2.o
 
 mmu-y	:= clear_user.o copy_page.o getuser.o putuser.o
 
diff --git a/arch/arm/lib/bswapsdi2.S b/arch/arm/lib/bswapsdi2.S
new file mode 100644
index 0000000..e9c8ca7
--- /dev/null
+++ b/arch/arm/lib/bswapsdi2.S
@@ -0,0 +1,36 @@ 
+#include <linux/linkage.h>
+
+#if __LINUX_ARM_ARCH__ >= 6
+ENTRY(__bswapsi2)
+	rev	r0, r0
+	bx	lr
+ENDPROC(__bswapsi2)
+
+ENTRY(__bswapdi2)
+	rev	r3, r0
+	rev	r0, r1
+	mov	r1, r3
+	bx	lr
+ENDPROC(__bswapdi2)
+#else
+ENTRY(__bswapsi2)
+	eor     r3, r0, r0, ror #16
+	lsr     r3, r3, #8
+	bic     r3, r3, #65280  @ 0xff00
+	eor     r0, r3, r0, ror #8
+	mov     pc, lr
+ENDPROC(__bswapsi2)
+
+ENTRY(__bswapdi2)
+	mov     ip, r1
+	eor     r3, ip, ip, ror #16
+	eor     r1, r0, r0, ror #16
+	lsr     r1, r1, #8
+	lsr     r3, r3, #8
+	bic     r3, r3, #65280  @ 0xff00
+	bic     r1, r1, #65280  @ 0xff00
+	eor     r1, r1, r0, ror #8
+	eor     r0, r3, ip, ror #8
+	mov     pc, lr
+ENDPROC(__bswapdi2)
+#endif