Re: [RFC PATCH 3/4] sh: define __BIG_ENDIAN for math-emu

From: Geert Uytterhoeven
Date: Fri Jun 04 2021 - 03:24:40 EST


Hi Randy,

On Fri, Jun 4, 2021 at 1:19 AM Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote:
> On 6/3/21 12:54 AM, Geert Uytterhoeven wrote:
> > On Thu, Jun 3, 2021 at 1:17 AM Randy Dunlap <rdunlap@xxxxxxxxxxxxx> wrote:
> >> The headers in include/math-emu/ test for __BYTE_ORDER == __BIG_ENDIAN
> >> without checking to see if these macros are defined, so add
> >> a define for __BIG_ENDIAN before pulling in these headers.
> >>
> >> This placates these build warnings:
> >>
> >> In file included from ../arch/sh/math-emu/math.c:23:
> >> ../include/math-emu/single.h:50:21: warning: "__BIG_ENDIAN" is not defined, evaluates to 0 [-Wundef]
> >> 50 | #if __BYTE_ORDER == __BIG_ENDIAN
> >> In file included from ../arch/sh/math-emu/math.c:24:
> >> ../include/math-emu/double.h:59:21: warning: "__BIG_ENDIAN" is not defined, evaluates to 0 [-Wundef]
> >> 59 | #if __BYTE_ORDER == __BIG_ENDIAN
> >>
> >> Fixes: 4b565680d163 ("sh: math-emu support")
> >> Signed-off-by: Randy Dunlap <rdunlap@xxxxxxxxxxxxx>
> >
> > Thanks for your patch!
> >
> >> --- linux-next-20210528.orig/arch/sh/math-emu/sfp-util.h
> >> +++ linux-next-20210528/arch/sh/math-emu/sfp-util.h
> >> @@ -70,4 +70,4 @@
> >>
> >> #define __BYTE_ORDER __LITTLE_ENDIAN
> >>
> >> -
> >> +#define __BIG_ENDIAN 0
> >
> > I don't think this is the right fix.
> >
> > I think the right values should be picked up from:
> >
> > include/uapi/linux/byteorder/big_endian.h:#define __BIG_ENDIAN 4321
> > include/uapi/linux/byteorder/little_endian.h:#define __LITTLE_ENDIAN 1234
> >
> > How is this picked up on other architectures using <math-emu/single.h>?
>
> There isn't very much to compare to in other arch/.
> I've made a v2 patch that is done like arch/nds32/ does.
> What do you think about this one?
>
> thanks.
> ---
> From: Randy Dunlap <rdunlap@xxxxxxxxxxxxx>
> Subject: [RFC PATCH 2/3 v2] sh: define __BIG_ENDIAN for math-emu
>
> Fix this by defining both ENDIAN macros in
> <asm/sfp-machine.h> so that they can be utilized in
> <math-emu/soft-fp.h> according to the latter's comment:
> /* Allow sfp-machine to have its own byte order definitions. */
>
> (This is what is done in arch/nds32/include/asm/sfp-machine.h.)
>
> This placates these build warnings:
>
> In file included from ../arch/sh/math-emu/math.c:23:
> ../include/math-emu/single.h:50:21: warning: "__BIG_ENDIAN" is not defined, evaluates to 0 [-Wundef]
> 50 | #if __BYTE_ORDER == __BIG_ENDIAN
> In file included from ../arch/sh/math-emu/math.c:24:
> ../include/math-emu/double.h:59:21: warning: "__BIG_ENDIAN" is not defined, evaluates to 0 [-Wundef]
> 59 | #if __BYTE_ORDER == __BIG_ENDIAN
>
> Fixes: 4b565680d163 ("sh: math-emu support")
> Signed-off-by: Randy Dunlap <rdunlap@xxxxxxxxxxxxx>
> Cc: Yoshinori Sato <ysato@xxxxxxxxxxxxxxxxxxxx>
> Cc: Rich Felker <dalias@xxxxxxxx>
> Cc: linux-sh@xxxxxxxxxxxxxxx
> Cc: Geert Uytterhoeven <geert@xxxxxxxxxxxxxx>
> Cc: John Paul Adrian Glaubitz <glaubitz@xxxxxxxxxxxxxxxxxxx>
> ---
> arch/sh/include/asm/sfp-machine.h | 8 ++++++++
> 1 file changed, 8 insertions(+)
>
> --- linux-next-20210528.orig/arch/sh/include/asm/sfp-machine.h
> +++ linux-next-20210528/arch/sh/include/asm/sfp-machine.h
> @@ -13,6 +13,14 @@
> #ifndef _SFP_MACHINE_H
> #define _SFP_MACHINE_H
>
> +#ifdef __BIG_ENDIAN__
> +#define __BYTE_ORDER __BIG_ENDIAN
> +#define __LITTLE_ENDIAN 0
> +#else
> +#define __BYTE_ORDER __LITTLE_ENDIAN
> +#define __BIG_ENDIAN 0
> +#endif
> +
> #define _FP_W_TYPE_SIZE 32
> #define _FP_W_TYPE unsigned long
> #define _FP_WS_TYPE signed long

These checks match with what is set by my sh cross-compiler (gcc
8.1.0):

diff <(sh4-linux-gcc-8.1.0 -ml -dM -E - < /dev/null | grep -E
"(BYTE_ORDER|ENDIAN)") <(sh4-linux-gcc-8.1.0 -mb -dM -E - < /dev/null
| grep -E "(BYTE_ORDER|ENDIAN)")
--- /dev/fd/63 2021-06-04 09:15:50.689928352 +0200
+++ /dev/fd/62 2021-06-04 09:15:50.689928352 +0200
@@ -1,6 +1,6 @@
#define __ORDER_LITTLE_ENDIAN__ 1234
-#define __FLOAT_WORD_ORDER__ __ORDER_LITTLE_ENDIAN__
+#define __BIG_ENDIAN__ 1
+#define __FLOAT_WORD_ORDER__ __ORDER_BIG_ENDIAN__
#define __ORDER_PDP_ENDIAN__ 3412
-#define __LITTLE_ENDIAN__ 1
#define __ORDER_BIG_ENDIAN__ 4321
-#define __BYTE_ORDER__ __ORDER_LITTLE_ENDIAN__
+#define __BYTE_ORDER__ __ORDER_BIG_ENDIAN__

So
Reviewed-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>

Note that powerpc checks on _BIG_ENDIAN, which works as my powerpc
cross-compiler (gcc 9.3.0) defines both _BIG_ENDIAN and _BIG_ENDIAN__.

Gr{oetje,eeting}s,

Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert@xxxxxxxxxxxxxx

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
-- Linus Torvalds