This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [v2 1/2] PPC64: Add libmvec SIMD single-precision natural exponent function.


Shawn Landden <shawn@git.icu> writes:

> 2019-05-11  Shawn Landden  <shawn@git.icu>
>
> 	[BZ #24209]
>         * NEWS: Noted the addition of PPC64 vector expf function.
>         * sysdeps/powerpc/bits/math-vector.h: Added entry for vector expf.
>         * sysdeps/powerpc/powerpc64/fpu/Versions: Added vector expf entry.
>         * sysdeps/powerpc/powerpc64/fpu/multiarch/Makefile:
>         (libmvec-sysdep_routines, CFLAGS-vec_s_expf4_vsx.c):
>         (CFLAGS-vec_s_exp2f_data.c): Added build of VSX SIMD expf
>         function and tests. Added vec_math_errf.c to build.
>         * sysdeps/powerpc/powerpc64/fpu/multiarch/math_config_flt.h: Modified for expf.
>         * sysdeps/powerpc/powerpc64/fpu/multiarch/test-float-vlen4-wrappers.c:
>         Added entry for vector expf.
>         * sysdeps/powerpc/powerpc64/fpu/multiarch/vec_s_expf4_vsx.c: New file.
>         * sysdeps/powerpc/powerpc64/fpu/multiarch/vec_s_exp2f_data.c: Likewise.
>         * sysdeps/powerpc/powerpc64/fpu/multiarch/vec_math_errf.c: Likewise.
>         * sysdeps/unix/sysv/linux/powerpc/powerpc64/libmvec.abilist: SIMD expf
>         added.

These entries are still indented with spaces.  Fixed.

> diff --git a/sysdeps/powerpc/powerpc64/fpu/multiarch/Makefile b/sysdeps/powerpc/powerpc64/fpu/multiarch/Makefile
> index a4f0d1694e..ad3c29b1ab 100644
> --- a/sysdeps/powerpc/powerpc64/fpu/multiarch/Makefile
> +++ b/sysdeps/powerpc/powerpc64/fpu/multiarch/Makefile
> @@ -46,10 +46,12 @@ endif
>  ifeq ($(subdir),mathvec)
>  libmvec-sysdep_routines += vec_d_cos2_vsx vec_s_cosf4_vsx \
>  			   vec_d_sin2_vsx vec_s_sinf4_vsx \
>  			   vec_d_log2_vsx vec_d_log_data \
>  			   vec_s_logf4_vsx vec_s_logf_data \
> +			   vec_s_expf4_vsx vec_s_exp2f_data \
> +			   vec_math_errf \
>  			   vec_d_sincos2_vsx vec_s_sincosf4_vsx
>  CFLAGS-vec_d_cos2_vsx.c += -mabi=altivec -maltivec -mvsx -mpower8-vector
>  CFLAGS-vec_d_log2_vsx.c += -mabi=altivec -maltivec -mvsx -mpower8-vector
>  CFLAGS-vec_d_log_data.c += -mabi=altivec -maltivec -mvsx -mpower8-vector
>  CFLAGS-vec_s_cosf4_vsx.c += -mabi=altivec -maltivec -mvsx
> @@ -57,19 +59,21 @@ CFLAGS-vec_s_logf4_vsx.c += -mabi=altivec -maltivec -mvsx
>  CFLAGS-vec_s_logf_data.c += -mabi=altivec -maltivec -mvsx
>  CFLAGS-vec_d_sin2_vsx.c += -mabi=altivec -maltivec -mvsx -mpower8-vector
>  CFLAGS-vec_s_sinf4_vsx.c += -mabi=altivec -maltivec -mvsx
>  CFLAGS-vec_d_sincos2_vsx.c += -mabi=altivec -maltivec -mvsx -mpower8-vector
>  CFLAGS-vec_s_sincosf4_vsx.c += -mabi=altivec -maltivec -mvsx
> +CFLAGS-vec_s_expf4_vsx.c += -mabi=altivec -maltivec -mvsx
> +CFLAGS-vec_s_exp2f_data.c += -mabi=altivec -maltivec -mvsx

vec_math_errf.c indirectly depends on altivec.h, which requires to use the same
flags.

Fixed locally.

> diff --git a/sysdeps/powerpc/powerpc64/fpu/multiarch/vec_math_errf.c b/sysdeps/powerpc/powerpc64/fpu/multiarch/vec_math_errf.c
> new file mode 100644
> index 0000000000..22b3a62fcf
> --- /dev/null
> +++ b/sysdeps/powerpc/powerpc64/fpu/multiarch/vec_math_errf.c
> @@ -0,0 +1,39 @@
> ...
> +attribute_hidden float
> +__math_oflowf (uint32_t sign)
> +{
> +  return xflowf (sign, 0x1p97f);
> +}
> +

Extra line at EOF here.  Fixed.

I'm merging this patch with the fixes I mentioned to branch tuliom/libmvec.

Thanks!

-- 
Tulio Magno


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]