From: Richard Guenther <richard.guenther@gmail.com>
To: "H.J. Lu" <hjl.tools@gmail.com>
Cc: gcc-patches@gcc.gnu.org, Uros Bizjak <ubizjak@gmail.com>,
Maxim Kuvyrkov <maxim@codesourcery.com>
Subject: Re: PATCH: Enable TARGET_HAS_SINCOS if x87 FPU fsincos is available
Date: Mon, 30 Aug 2010 15:24:00 -0000 [thread overview]
Message-ID: <AANLkTimA+x53SYyDxKZnkezuivXWiRZ_k-4XTz3O_Q8H@mail.gmail.com> (raw)
In-Reply-To: <20100830144016.GA31228@intel.com>
On Mon, Aug 30, 2010 at 4:40 PM, H.J. Lu <hongjiu.lu@intel.com> wrote:
> Hi,
>
> On x86, sincos is always available if x87 FPU fsincos is available.
> This patch enables TARGET_HAS_SINCOS for -ffast-math and x87 FPU. Also
> x86 Bionic C library doesn't provide sincos and we shouldn't enable
> TARGET_HAS_SINCOS with OPTION_BIONIC on x86. OK for trunk?
Hm. Shouldn't it be conditional on USE_FANCY_MATH_387
&& !NO_FANCY_MATH_387 as well?
Also a flag dependency breaks with the target/option attributes,
so I'm not sure it is a good idea. Does anyone care for FP
execution performance on 32bit anyway?
Richard.
> Thanks.
>
>
> H.J.
> ---
> 2010-08-30 H.J. Lu <hongjiu.lu@intel.com>
>
> * config/linux.h (TARGET_HAS_SINCOS): Define only if not
> defined.
>
> * config/i386/i386.h (TARGET_HAS_SINCOS): New. Enabled if x87
> FPU fsincos is available.
> * config/i386/x86-64.h (TARGET_HAS_SINCOS): Likewise.
> * config/i386/linux.h (TARGET_HAS_SINCOS): Likewise.
> * config/i386/linux64.h (TARGET_HAS_SINCOS): Likewise.
>
> diff --git a/gcc/config/i386/i386.h b/gcc/config/i386/i386.h
> index 5bae99d..493b5d2 100644
> --- a/gcc/config/i386/i386.h
> +++ b/gcc/config/i386/i386.h
> @@ -2492,6 +2492,12 @@ struct GTY(()) machine_function {
> #undef TARG_COND_NOT_TAKEN_BRANCH_COST
> #define TARG_COND_NOT_TAKEN_BRANCH_COST ix86_cost->cond_not_taken_branch_cost
>
> +/* Use x87 FPU fsincos if it is available. */
> +#undef TARGET_HAS_SINCOS
> +#define TARGET_HAS_SINCOS \
> + (flag_unsafe_math_optimizations \
> + && (!TARGET_SSE_MATH || TARGET_MIX_SSE_I387))
> +
> /*
> Local variables:
> version-control: t
> diff --git a/gcc/config/i386/linux.h b/gcc/config/i386/linux.h
> index 2a31880..f466d6d 100644
> --- a/gcc/config/i386/linux.h
> +++ b/gcc/config/i386/linux.h
> @@ -35,6 +35,15 @@ along with GCC; see the file COPYING3. If not see
> #undef TARGET_TLS_DIRECT_SEG_REFS_DEFAULT
> #define TARGET_TLS_DIRECT_SEG_REFS_DEFAULT MASK_TLS_DIRECT_SEG_REFS
>
> +/* Whether we have sincos that follows the GNU extension. There is no
> + sincos in Bionic C library. We can only use x87 FPU fsincos if it
> + is available. */
> +#undef TARGET_HAS_SINCOS
> +#define TARGET_HAS_SINCOS \
> + (OPTION_GLIBC \
> + || (flag_unsafe_math_optimizations \
> + && (!TARGET_SSE_MATH || TARGET_MIX_SSE_I387)))
> +
> #undef ASM_COMMENT_START
> #define ASM_COMMENT_START "#"
>
> diff --git a/gcc/config/i386/linux64.h b/gcc/config/i386/linux64.h
> index 867de59..9f38a57 100644
> --- a/gcc/config/i386/linux64.h
> +++ b/gcc/config/i386/linux64.h
> @@ -50,6 +50,15 @@ see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
> #undef TARGET_TLS_DIRECT_SEG_REFS_DEFAULT
> #define TARGET_TLS_DIRECT_SEG_REFS_DEFAULT MASK_TLS_DIRECT_SEG_REFS
>
> +/* Whether we have sincos that follows the GNU extension. There is no
> + sincos in Bionic C library. We can only use x87 FPU fsincos if it
> + is available. */
> +#undef TARGET_HAS_SINCOS
> +#define TARGET_HAS_SINCOS \
> + (OPTION_GLIBC \
> + || (flag_unsafe_math_optimizations \
> + && (!TARGET_SSE_MATH || TARGET_MIX_SSE_I387)))
> +
> /* Provide a LINK_SPEC. Here we provide support for the special GCC
> options -static and -shared, which allow us to link things in one
> of these three modes by applying the appropriate combinations of
> diff --git a/gcc/config/i386/x86-64.h b/gcc/config/i386/x86-64.h
> index 1b64813..d18ec23 100644
> --- a/gcc/config/i386/x86-64.h
> +++ b/gcc/config/i386/x86-64.h
> @@ -107,3 +107,9 @@ see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
>
> #undef TARGET_ASM_UNIQUE_SECTION
> #define TARGET_ASM_UNIQUE_SECTION x86_64_elf_unique_section
> +
> +/* Use x87 FPU fsincos if it is available. */
> +#undef TARGET_HAS_SINCOS
> +#define TARGET_HAS_SINCOS \
> + (flag_unsafe_math_optimizations \
> + && (!TARGET_SSE_MATH || TARGET_MIX_SSE_I387))
> diff --git a/gcc/config/linux.h b/gcc/config/linux.h
> index e283a9a..576a2ac 100644
> --- a/gcc/config/linux.h
> +++ b/gcc/config/linux.h
> @@ -159,7 +159,9 @@ see the files COPYING3 and COPYING.RUNTIME respectively. If not, see
> is present in the runtime library. */
> #define TARGET_C99_FUNCTIONS (OPTION_GLIBC)
>
> +#ifndef TARGET_HAS_SINCOS
> /* Whether we have sincos that follows the GNU extension. */
> #define TARGET_HAS_SINCOS (OPTION_GLIBC || OPTION_BIONIC)
> +#endif
>
> #define TARGET_POSIX_IO
>
next prev parent reply other threads:[~2010-08-30 14:57 UTC|newest]
Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-08-30 14:55 H.J. Lu
2010-08-30 15:24 ` Richard Guenther [this message]
2010-08-30 16:17 ` H.J. Lu
2010-09-05 7:21 ` t66667
2010-09-05 8:21 ` H.J. Lu
2010-08-30 16:32 ` Maxim Kuvyrkov
2010-10-30 1:06 Jing Yu
[not found] ` <AANLkTinGLSX0U2PR0oM753NhFM8Poygx6qf3E-iSgM=a@mail.gmail.com>
2010-10-31 23:03 ` Maxim Kuvyrkov
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=AANLkTimA+x53SYyDxKZnkezuivXWiRZ_k-4XTz3O_Q8H@mail.gmail.com \
--to=richard.guenther@gmail.com \
--cc=gcc-patches@gcc.gnu.org \
--cc=hjl.tools@gmail.com \
--cc=maxim@codesourcery.com \
--cc=ubizjak@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).