From: caiyinyu <caiyinyu@loongson.cn>
To: Xiaolin Tang <tangxiaolin@loongson.cn>,
adhemerval.zanella@linaro.org, libc-alpha@sourceware.org,
i.swmail@xen0n.name
Cc: xry111@xry111.site, xuchenghua@loongson.cn, chenglulu@loongson.cn
Subject: Re: [PATCH v2 00/10] LoongArch: Hard Float Support Of Math Functions.
Date: Fri, 11 Nov 2022 12:02:16 +0800 [thread overview]
Message-ID: <44e65c33-a47d-3e67-bc5e-0b1d441e0ae0@loongson.cn> (raw)
In-Reply-To: <20221107140523.896735-1-tangxiaolin@loongson.cn>
Considering Xi Ruoyao's new patches of GCC about "LoongArch: Add some
floating-point operations[1]"
, I plan to include the following patches
02, 03, 06-10 temporarily excluding rint{,f} {l/ll}rint{,f} logb{,f} and
scalb{n,ln}{,f} related ones.
Also, I have noticed that there are some new functions added in Patch v2
03: s_isinf{f}.c
and s_isnan{f}.c.
All these patches were tested on LoongArch machine with GCC 12.1.0,
binutils 2.38.50.20220519, and
Linux kernel 5.19 and no new fails were caused.
[1] https://gcc.gnu.org/pipermail/gcc-patches/2022-November/605454.html
在 2022/11/7 下午10:05, Xiaolin Tang 写道:
> Hi, reviewers.
> Those patches contains hardware floating-point support of math fuctions
> for the LoongArch ISA. LoongArch ISA has many hardware floating-point
> instructions that can be used in the corresponding math functions,
> and the performance of those newly implemented functions is good.
>
> Changes from v1 to v2:
> 1. Delete copysign{f/ } to implement it on compiler.
> (https://gcc.gnu.org/pipermail/gcc-patches/2022-November/605071.html)
> 2. Replace __builtin_expect with __glibc_likely or __glibc_unlikely.
> 3. Add a brief description on each file header.
> 4. Split one patch to ten patches for similar functions implementation.
>
> instructions functions
> fma.{s/d} __fma{f/ }
> f{max/min}.{s/d}, fclass.{s/d} __f{max/min}i{mum/mum_mum}{f/ }
> f{maxa/mina}.{s/d} __f{max,min}mag{f/ }
> f{maxa/mina}.{s/d}, fclass.{s/d __f{max/min}imum{_mag/_mag_num}{f/ }
> frint.{s/d} __{ /l/ll}rint{f/ }
> fscaleb.{s/d}, fclass.{s/d} __scalbn{f/ }, __ieee754__scalb{f/ }
> flogb.{s/d}, fclass.{s/d} __logb{f/ }, __ieee754_ilogb{f/ }
> fclass.{s/d} __fpclassify{f/ }, __issignaling{f/ }
> __finite{f/ }, __isnan{f/ }, isinff{f/ }
>
> Xiaolin Tang (10):
> LoongArch: Hard Float Support for functions {ll/l/ }rint{f/ }.
> LoongArch: Use __builtin_{fma, fmaf} to implement function {fma,
> fmaf}.
> LoongArch: Hard Float Support for float-point classification
> functions.
> LoongArch: Hard Float Support for functions logb{f/ },
> __ieee754_ilogb{f/ }.
> LoongArch: Hard Float Support for functions scalb{f/ }, scalbn{f/ }.
> LoongArch: Hard Float Support for fmaximum{f/ }, fminimum{f/ }.
> LoongArch: Hard Float Support for fmaximum_num{f/ }, fminimum_num{f/
> }.
> LoongArch: Hard Float Support for fmaxmag{f/ }, fminmag{f/ }.
> LoongArch: Hard Float Support for fmaximum_mag{f/ }, fminimum_mag{f/
> }.
> LoongArch: Hard Float Support for fmaximum_mag_num{f/ },
> fminimum_mag_num{f/ }.
>
> sysdeps/loongarch/fpu/e_ilogb.c | 39 ++++++++++++
> sysdeps/loongarch/fpu/e_ilogbf.c | 39 ++++++++++++
> sysdeps/loongarch/fpu/e_scalb.c | 61 +++++++++++++++++++
> sysdeps/loongarch/fpu/e_scalbf.c | 61 +++++++++++++++++++
> sysdeps/loongarch/fpu/math-use-builtins-fma.h | 4 ++
> sysdeps/loongarch/fpu/s_finite.c | 30 +++++++++
> sysdeps/loongarch/fpu/s_finitef.c | 30 +++++++++
> sysdeps/loongarch/fpu/s_fmaximum.c | 40 ++++++++++++
> sysdeps/loongarch/fpu/s_fmaximum_mag.c | 40 ++++++++++++
> sysdeps/loongarch/fpu/s_fmaximum_mag_num.c | 48 +++++++++++++++
> sysdeps/loongarch/fpu/s_fmaximum_mag_numf.c | 48 +++++++++++++++
> sysdeps/loongarch/fpu/s_fmaximum_magf.c | 40 ++++++++++++
> sysdeps/loongarch/fpu/s_fmaximum_num.c | 48 +++++++++++++++
> sysdeps/loongarch/fpu/s_fmaximum_numf.c | 49 +++++++++++++++
> sysdeps/loongarch/fpu/s_fmaximumf.c | 40 ++++++++++++
> sysdeps/loongarch/fpu/s_fmaxmag.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_fmaxmagf.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_fminimum.c | 40 ++++++++++++
> sysdeps/loongarch/fpu/s_fminimum_mag.c | 40 ++++++++++++
> sysdeps/loongarch/fpu/s_fminimum_mag_num.c | 48 +++++++++++++++
> sysdeps/loongarch/fpu/s_fminimum_mag_numf.c | 48 +++++++++++++++
> sysdeps/loongarch/fpu/s_fminimum_magf.c | 40 ++++++++++++
> sysdeps/loongarch/fpu/s_fminimum_num.c | 48 +++++++++++++++
> sysdeps/loongarch/fpu/s_fminimum_numf.c | 48 +++++++++++++++
> sysdeps/loongarch/fpu/s_fminimumf.c | 40 ++++++++++++
> sysdeps/loongarch/fpu/s_fminmag.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_fminmagf.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_fpclassify.c | 38 ++++++++++++
> sysdeps/loongarch/fpu/s_fpclassifyf.c | 38 ++++++++++++
> sysdeps/loongarch/fpu/s_isinf.c | 30 +++++++++
> sysdeps/loongarch/fpu/s_isinff.c | 30 +++++++++
> sysdeps/loongarch/fpu/s_isnan.c | 31 ++++++++++
> sysdeps/loongarch/fpu/s_isnanf.c | 31 ++++++++++
> sysdeps/loongarch/fpu/s_issignaling.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_issignalingf.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_llrint.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_llrintf.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_logb.c | 30 +++++++++
> sysdeps/loongarch/fpu/s_logbf.c | 30 +++++++++
> sysdeps/loongarch/fpu/s_lrint.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_lrintf.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_rint.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_rintf.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_scalbn.c | 29 +++++++++
> sysdeps/loongarch/fpu/s_scalbnf.c | 29 +++++++++
> sysdeps/loongarch/fpu_control.h | 17 ++++++
> 46 files changed, 1650 insertions(+)
> create mode 100644 sysdeps/loongarch/fpu/e_ilogb.c
> create mode 100644 sysdeps/loongarch/fpu/e_ilogbf.c
> create mode 100644 sysdeps/loongarch/fpu/e_scalb.c
> create mode 100644 sysdeps/loongarch/fpu/e_scalbf.c
> create mode 100644 sysdeps/loongarch/fpu/math-use-builtins-fma.h
> create mode 100644 sysdeps/loongarch/fpu/s_finite.c
> create mode 100644 sysdeps/loongarch/fpu/s_finitef.c
> create mode 100644 sysdeps/loongarch/fpu/s_fmaximum.c
> create mode 100644 sysdeps/loongarch/fpu/s_fmaximum_mag.c
> create mode 100644 sysdeps/loongarch/fpu/s_fmaximum_mag_num.c
> create mode 100644 sysdeps/loongarch/fpu/s_fmaximum_mag_numf.c
> create mode 100644 sysdeps/loongarch/fpu/s_fmaximum_magf.c
> create mode 100644 sysdeps/loongarch/fpu/s_fmaximum_num.c
> create mode 100644 sysdeps/loongarch/fpu/s_fmaximum_numf.c
> create mode 100644 sysdeps/loongarch/fpu/s_fmaximumf.c
> create mode 100644 sysdeps/loongarch/fpu/s_fmaxmag.c
> create mode 100644 sysdeps/loongarch/fpu/s_fmaxmagf.c
> create mode 100644 sysdeps/loongarch/fpu/s_fminimum.c
> create mode 100644 sysdeps/loongarch/fpu/s_fminimum_mag.c
> create mode 100644 sysdeps/loongarch/fpu/s_fminimum_mag_num.c
> create mode 100644 sysdeps/loongarch/fpu/s_fminimum_mag_numf.c
> create mode 100644 sysdeps/loongarch/fpu/s_fminimum_magf.c
> create mode 100644 sysdeps/loongarch/fpu/s_fminimum_num.c
> create mode 100644 sysdeps/loongarch/fpu/s_fminimum_numf.c
> create mode 100644 sysdeps/loongarch/fpu/s_fminimumf.c
> create mode 100644 sysdeps/loongarch/fpu/s_fminmag.c
> create mode 100644 sysdeps/loongarch/fpu/s_fminmagf.c
> create mode 100644 sysdeps/loongarch/fpu/s_fpclassify.c
> create mode 100644 sysdeps/loongarch/fpu/s_fpclassifyf.c
> create mode 100644 sysdeps/loongarch/fpu/s_isinf.c
> create mode 100644 sysdeps/loongarch/fpu/s_isinff.c
> create mode 100644 sysdeps/loongarch/fpu/s_isnan.c
> create mode 100644 sysdeps/loongarch/fpu/s_isnanf.c
> create mode 100644 sysdeps/loongarch/fpu/s_issignaling.c
> create mode 100644 sysdeps/loongarch/fpu/s_issignalingf.c
> create mode 100644 sysdeps/loongarch/fpu/s_llrint.c
> create mode 100644 sysdeps/loongarch/fpu/s_llrintf.c
> create mode 100644 sysdeps/loongarch/fpu/s_logb.c
> create mode 100644 sysdeps/loongarch/fpu/s_logbf.c
> create mode 100644 sysdeps/loongarch/fpu/s_lrint.c
> create mode 100644 sysdeps/loongarch/fpu/s_lrintf.c
> create mode 100644 sysdeps/loongarch/fpu/s_rint.c
> create mode 100644 sysdeps/loongarch/fpu/s_rintf.c
> create mode 100644 sysdeps/loongarch/fpu/s_scalbn.c
> create mode 100644 sysdeps/loongarch/fpu/s_scalbnf.c
>
prev parent reply other threads:[~2022-11-11 4:02 UTC|newest]
Thread overview: 11+ messages / expand[flat|nested] mbox.gz Atom feed top
2022-11-07 14:05 Xiaolin Tang
2022-11-07 14:05 ` [PATCH v2 01/10] LoongArch: Hard Float Support for functions {ll/l/ }rint{f/ } Xiaolin Tang
2022-11-07 14:38 ` Xi Ruoyao
2022-11-07 16:36 ` Adhemerval Zanella Netto
2022-11-07 14:05 ` [PATCH v2 02/10] LoongArch: Use __builtin_{fma, fmaf} to implement function {fma, fmaf} Xiaolin Tang
2022-11-07 14:05 ` [PATCH v2 03/10] LoongArch: Hard Float Support for float-point classification functions Xiaolin Tang
2022-11-07 14:05 ` [PATCH v2 04/10] LoongArch: Hard Float Support for functions logb{f/ }, __ieee754_ilogb{f/ } Xiaolin Tang
2022-11-07 14:05 ` [PATCH v2 05/10] LoongArch: Hard Float Support for functions scalb{f/ }, scalbn{f/ } Xiaolin Tang
2022-11-08 3:46 ` Xi Ruoyao
2022-11-07 14:05 ` [PATCH v2 06/10] LoongArch: Hard Float Support for fmaximum{f/ }, fminimum{f/ } Xiaolin Tang
2022-11-11 4:02 ` caiyinyu [this message]
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=44e65c33-a47d-3e67-bc5e-0b1d441e0ae0@loongson.cn \
--to=caiyinyu@loongson.cn \
--cc=adhemerval.zanella@linaro.org \
--cc=chenglulu@loongson.cn \
--cc=i.swmail@xen0n.name \
--cc=libc-alpha@sourceware.org \
--cc=tangxiaolin@loongson.cn \
--cc=xry111@xry111.site \
--cc=xuchenghua@loongson.cn \
/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).