From: Noah Goldstein <goldstein.w.n@gmail.com>
To: "H.J. Lu" <hjl.tools@gmail.com>
Cc: libc-alpha@sourceware.org
Subject: Re: [PATCH] x86-64: Add x87 fmod and remainder [BZ #30179]
Date: Thu, 9 Mar 2023 20:36:13 -0600 [thread overview]
Message-ID: <CAFUsyf+sF8iW5aHqaL3COqsMW8h2vMFZUcvRgXAoRLK66-HvYg@mail.gmail.com> (raw)
In-Reply-To: <20230309183312.205763-1-hjl.tools@gmail.com>
On Thu, Mar 9, 2023 at 12:33 PM H.J. Lu via Libc-alpha
<libc-alpha@sourceware.org> wrote:
>
> X87 (fprem/fprem1) implementations of fmod and remainder are much faster
> than generic fmod and remainder. Add e_fmod.S, e_fmodf.S, e_remainder.S
> and e_remainderf.S with fprem/fprem1. This fixes BZ #30179.
> ---
> sysdeps/x86_64/fpu/e_fmod.S | 22 ++++++++++++++++++++++
> sysdeps/x86_64/fpu/e_fmodf.S | 22 ++++++++++++++++++++++
> sysdeps/x86_64/fpu/e_remainder.S | 22 ++++++++++++++++++++++
> sysdeps/x86_64/fpu/e_remainderf.S | 22 ++++++++++++++++++++++
> 4 files changed, 88 insertions(+)
> create mode 100644 sysdeps/x86_64/fpu/e_fmod.S
> create mode 100644 sysdeps/x86_64/fpu/e_fmodf.S
> create mode 100644 sysdeps/x86_64/fpu/e_remainder.S
> create mode 100644 sysdeps/x86_64/fpu/e_remainderf.S
>
> diff --git a/sysdeps/x86_64/fpu/e_fmod.S b/sysdeps/x86_64/fpu/e_fmod.S
> new file mode 100644
> index 0000000000..4bdc8a1ab0
> --- /dev/null
> +++ b/sysdeps/x86_64/fpu/e_fmod.S
> @@ -0,0 +1,22 @@
> +/*
> + * Public domain.
> + */
> +
> +#include <machine/asm.h>
> +#include <libm-alias-finite.h>
> +
> +ENTRY(__ieee754_fmod)
> + movsd %xmm0, -16(%rsp)
> + movsd %xmm1, -8(%rsp)
> + fldl -8(%rsp)
> + fldl -16(%rsp)
> +1: fprem
> + fstsw %ax
> + sahf
> + jp 1b
For all functions can you replace `sahf; jp` with `testl $0x400, %eax; jnz`?
> + fstp %st(1)
> + fstpl -8(%rsp)
> + movsd -8(%rsp), %xmm0
> + ret
> +END (__ieee754_fmod)
> +libm_alias_finite (__ieee754_fmod, __fmod)
> diff --git a/sysdeps/x86_64/fpu/e_fmodf.S b/sysdeps/x86_64/fpu/e_fmodf.S
> new file mode 100644
> index 0000000000..6f76daff01
> --- /dev/null
> +++ b/sysdeps/x86_64/fpu/e_fmodf.S
> @@ -0,0 +1,22 @@
> +/*
> + * Public domain.
> + */
> +
> +#include <machine/asm.h>
> +#include <libm-alias-finite.h>
> +
> +ENTRY(__ieee754_fmodf)
> + movss %xmm0, -8(%rsp)
> + movss %xmm1, -4(%rsp)
> + flds -4(%rsp)
> + flds -8(%rsp)
> +1: fprem
> + fstsw %ax
> + sahf
> + jp 1b
> + fstp %st(1)
> + fstps -4(%rsp)
> + movss -4(%rsp), %xmm0
> + ret
> +END (__ieee754_fmodf)
> +libm_alias_finite (__ieee754_fmodf, __fmodf)
> diff --git a/sysdeps/x86_64/fpu/e_remainder.S b/sysdeps/x86_64/fpu/e_remainder.S
> new file mode 100644
> index 0000000000..be2184f25a
> --- /dev/null
> +++ b/sysdeps/x86_64/fpu/e_remainder.S
> @@ -0,0 +1,22 @@
> +/*
> + * Public domain.
> + */
> +
> +#include <machine/asm.h>
> +#include <libm-alias-finite.h>
> +
> +ENTRY(__ieee754_remainder)
> + movsd %xmm0, -16(%rsp)
> + movsd %xmm1, -8(%rsp)
> + fldl -8(%rsp)
> + fldl -16(%rsp)
> +1: fprem1
> + fstsw %ax
> + sahf
> + jp 1b
> + fstp %st(1)
> + fstpl -8(%rsp)
> + movsd -8(%rsp), %xmm0
> + ret
> +END (__ieee754_remainder)
> +libm_alias_finite (__ieee754_remainder, __remainder)
> diff --git a/sysdeps/x86_64/fpu/e_remainderf.S b/sysdeps/x86_64/fpu/e_remainderf.S
> new file mode 100644
> index 0000000000..42972d3f84
> --- /dev/null
> +++ b/sysdeps/x86_64/fpu/e_remainderf.S
> @@ -0,0 +1,22 @@
> +/*
> + * Public domain.
> + */
> +
> +#include <machine/asm.h>
> +#include <libm-alias-finite.h>
> +
> +ENTRY(__ieee754_remainderf)
> + movss %xmm0, -8(%rsp)
> + movss %xmm1, -4(%rsp)
> + flds -4(%rsp)
> + flds -8(%rsp)
> +1: fprem1
> + fstsw %ax
> + sahf
> + jp 1b
> + fstp %st(1)
> + fstps -4(%rsp)
> + movss -4(%rsp), %xmm0
> + ret
> +END (__ieee754_remainderf)
> +libm_alias_finite (__ieee754_remainderf, __remainderf)
> --
> 2.39.2
>
next prev parent reply other threads:[~2023-03-10 2:36 UTC|newest]
Thread overview: 3+ messages / expand[flat|nested] mbox.gz Atom feed top
2023-03-09 18:33 H.J. Lu
2023-03-10 2:36 ` Noah Goldstein [this message]
2023-03-10 16:42 ` H.J. Lu
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=CAFUsyf+sF8iW5aHqaL3COqsMW8h2vMFZUcvRgXAoRLK66-HvYg@mail.gmail.com \
--to=goldstein.w.n@gmail.com \
--cc=hjl.tools@gmail.com \
--cc=libc-alpha@sourceware.org \
/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).