[x264-devel] [PATCH] Updated configure to detect MIPS platform and set flags.

[BugMaster]

On Thu, 16 Apr 2015 12:07:51 +0000, Kaustubh Raste wrote:
> Hi,

> May I request somebody from maintainers to have a look at this patch please?
> I am holding up my further patch submissions till this one gets accepted.

There is no point in holding back on patches. We can review them
simultaneous.

> Thanks
> Kaustubh

> -----Original Message-----
> From: x264-devel [mailto:x264-devel-bounces at videolan.org] On Behalf Of Kaustubh Raste
> Sent: Tuesday, April 14, 2015 6:37 PM
> To: x264-devel at videolan.org
> Subject: Re: [x264-devel] [PATCH] Updated configure to detect MIPS platform and set flags.

> Hi,

> Any review comments on below patch?

> -Kaustubh

> -----Original Message-----
> From: Kaustubh Raste 
> Sent: Friday, April 10, 2015 3:49 PM
> To: x264-devel at videolan.org
> Cc: Kaustubh Raste
> Subject: [PATCH] Updated configure to detect MIPS platform and set flags.

> The MIPS GCC compiler we are using has been configured with following flags.

> Target: mips-mti-linux-gnu
> Configured with: /user/mipsswt/git/gcc/configure
> --prefix=/user/mipsswt/install-mips-mti-linux-gnu --disable-libssp
> --disable-libgomp --disable-libmudflap --disable-decimal-float
> --with-mips-plt --target=mips-mti-linux-gnu --enable-languages=c,c++
> --enable-__cxa_atexit
> --with-sysroot=/user/mipsswt/install-mips-mti-linux-gnu/sysroot
> --with-build-sysroot=/user/mipsswt/install-mips-mti-linux-gnu/sysroot
> --with-bugurl=mips.support at imgtec.com --with-pkgversion='Codescape
> GNU Tools 2015.01-5 for MIPS MTI Linux'
> Thread model: posix

> The target triplet used in this configuration is
> mips-mti-linux-gnu. The supported architectures by this compiler
> (-march=XXX) contains mips64r2 mips64r3 mips64r5 mips64r6 etc.
> Thus we have used mips32r5 and mips64r6 etc. and not mipsisa*

I was asking not about gcc triplet but about OS triplet and
specifically about `uname -m` which can be used for automatic detection
of needed system without specifying --host. But if you want special
custom values than I am ok with that.

> Signed-off-by: Kaustubh Raste <kaustubh.raste at imgtec.com>
> ---
>  config.sub |  8 ++++++++
>  configure  | 35 +++++++++++++++++++++++++++++++++--
>  2 files changed, 41 insertions(+), 2 deletions(-)

> diff --git a/config.sub b/config.sub
> index 8df5511..9b67804 100755
> --- a/config.sub
> +++ b/config.sub
> @@ -294,6 +294,10 @@ case $basic_machine in
>         | mipsisa64sb1 | mipsisa64sb1el \
>         | mipsisa64sr71k | mipsisa64sr71kel \
>         | mipstx39 | mipstx39el \
> +       | mips32r5 | mips32r5el \
> +       | mips32r6 | mips32r6el \
> +       | mips64r5 | mips64r5el \
> +       | mips64r6 | mips64r6el \
>         | mn10200 | mn10300 \
>         | moxie \
>         | mt \
> @@ -411,6 +415,10 @@ case $basic_machine in
>         | mipsisa64sb1-* | mipsisa64sb1el-* \
>         | mipsisa64sr71k-* | mipsisa64sr71kel-* \
>         | mipstx39-* | mipstx39el-* \
> +       | mips32r5-* | mips32r5el-* \
> +       | mips32r6-* | mips32r6el-* \
> +       | mips64r5-* | mips64r5el-* \
> +       | mips64r6-* | mips64r6el-* \
>         | mmix-* \
>         | mt-* \
>         | msp430-* \
> diff --git a/configure b/configure
> index 15b1c91..56e2a11 100755
> --- a/configure
> +++ b/configure
> @@ -364,7 +364,8 @@ NL="
>  
>  # list of all preprocessor HAVE values we can define 
> CONFIG_HAVE="MALLOC_H ALTIVEC ALTIVEC_H MMX ARMV6 ARMV6T2 NEON
> BEOSTHREAD POSIXTHREAD WIN32THREAD THREAD LOG2F SWSCALE \
> -             LAVF FFMS GPAC AVS GPL VECTOREXT INTERLACED CPU_COUNT
> OPENCL THP LSMASH X86_INLINE_ASM AS_FUNC INTEL_DISPATCHER"
> +             LAVF FFMS GPAC AVS GPL VECTOREXT INTERLACED CPU_COUNT
> OPENCL THP LSMASH X86_INLINE_ASM AS_FUNC INTEL_DISPATCHER \
> +             MSA"
>  
>  # parse options
>  
> @@ -733,8 +734,25 @@ case $host_cpu in
>      sparc)
>          ARCH="SPARC"
>          ;;
> -    mips|mipsel|mips64|mips64el)
> +    mips*)
>          ARCH="MIPS"
> +        case $host_cpu in
> +            mips32r5|mips32r5el)
> +                CFLAGS="$CFLAGS -mips32r5 -msched-weight
> -funroll-loops -mload-store-pairs -mhard-float -mfp64"
> +                if [ $host_cpu = mips32r5el ] ; then
> +                    CFLAGS="$CFLAGS -EL"
> +                    LDFLAGS="$LDFLAGS -EL"
> +                fi
> +                ;;
> +            mips64r6|mips64r6el)
> +                CFLAGS="$CFLAGS -mips64r6 -mabi=64 -msched-weight
> -funroll-loops -mload-store-pairs -mhard-float -mfp64"
> +                LDFLAGS="$LDFLAGS -mabi=64"
> +                if [ $host_cpu = mips64r6el ] ; then
> +                    CFLAGS="$CFLAGS -EL"
> +                    LDFLAGS="$LDFLAGS -EL"
> +                fi
> +                ;;

No config for mips32r6|mips32r6el|mips64r5|mips64r5el ?

> +        esac
>          ;;
>      arm*)
>          ARCH="ARM"
> @@ -860,6 +878,19 @@ if [ $asm = auto -a \( $ARCH = ARM -o $ARCH = AARCH64 \) ] ; then
>      as_check ".func test${NL}.endfunc" && define HAVE_AS_FUNC 1  fi
>  
> +if [ $asm = auto -a $ARCH = MIPS ] ; then
> +    case $host_cpu in
> +        mips32r5|mips32r5el|mips64r6|mips64r6el)
> +            echo $CFLAGS | grep -Eq '(-mcpu|-march|-mfpu|-arch)' || CFLAGS="$CFLAGS -mmsa"
> +
> +            if cc_check '' '' '__asm__("move.v $w5, $w11");' ; then
> +                define HAVE_MSA
> +                LDFLAGS="$LDFLAGS -mmsa"
> +            fi
> +            ;;
> +    esac
> +fi
> +
>  [ $asm = no ] && AS=""
>  [ "x$AS" = x ] && asm="no" || asm="yes"
>  
> --
> 2.3.2

And in general, I don't really like idea of having config default
flags for every specific revision if MIPS cpu. Can we only make check
for MSA support with defining HAVE_MSA and ask to specify all other
rather specific options (targeting only gcc and no other compiler,
dunno if such exist for MIPS) through --extra-cflags --extra-ldflags?