From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-pg1-x535.google.com (mail-pg1-x535.google.com [IPv6:2607:f8b0:4864:20::535]) by sourceware.org (Postfix) with ESMTPS id 4DFDD3858D39 for ; Wed, 29 Dec 2021 21:25:31 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 4DFDD3858D39 Received: by mail-pg1-x535.google.com with SMTP id s1so17971034pga.5 for ; Wed, 29 Dec 2021 13:25:31 -0800 (PST) X-Google-DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=1e100.net; s=20210112; h=x-gm-message-state:date:from:to:cc:subject:message-id:references :mime-version:content-disposition:in-reply-to; bh=VZuMqLsGD29rkrpUT0YfgcahP0NTr8xFoghkuCB46Ls=; b=kTOGeAyV8xaRkvysuA2mUQUPdddm+L+rWkli/OGHhhpd46+FC6CPrV4JEgm0AWmCaZ aE05MjhcrJ/a2K0Kbzlk9Hx/5Y6Rk5JknwIX1AdqBl8TFo2iCtCWPcwiWHRZTnsRK5Rh BHBdgaJAj80OXaJMVjghWKvuxvX+t/UoaZslfppeURkSw4fxitaKqf/QqG7it380qIuK I2MZLPb75GS0fnxrLLO0IdRoXld8qVuhx5Syk2hCcFvGWO1XXz6W7UN1Nx1l2aSSjplt yvASoS2llVerzaZErv/CSwV17KB05zsNHSsosNIoVQmUyO2AWsPSXytLGp9tLUpfwtm8 XxeQ== X-Gm-Message-State: AOAM533rTPP1ve0HGJvHEJ9Gx2vL7pg3bQy4VIWTQ5zkaGNggHAdrzDx OXltwRH+GSmEsmSWL4olO9w= X-Google-Smtp-Source: ABdhPJz232ofxQHChMAI213hTrcgKKMzIlK3QRT0PzrGgskoWrkDlVosjuN+enPHgM5X+dmBxuKqpA== X-Received: by 2002:a63:954c:: with SMTP id t12mr20331874pgn.97.1640813129260; Wed, 29 Dec 2021 13:25:29 -0800 (PST) Received: from gnu-tgl-3.localdomain ([172.58.35.133]) by smtp.gmail.com with ESMTPSA id d1sm12526137pfj.51.2021.12.29.13.25.28 (version=TLS1_3 cipher=TLS_AES_256_GCM_SHA384 bits=256/256); Wed, 29 Dec 2021 13:25:28 -0800 (PST) Received: by gnu-tgl-3.localdomain (Postfix, from userid 1000) id CB379C08AD; Wed, 29 Dec 2021 13:25:27 -0800 (PST) Date: Wed, 29 Dec 2021 13:25:27 -0800 From: "H.J. Lu" To: Sunil K Pandey Cc: libc-alpha@sourceware.org, hjl.tools@gmail.com, andrey.kolesov@intel.com, marius.cornea@intel.com Subject: Re: [PATCH v5 02/18] x86-64: Add vector asin/asinf implementation to libmvec Message-ID: References: <20211229064000.1465621-1-skpgkp2@gmail.com> <20211229064000.1465621-3-skpgkp2@gmail.com> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20211229064000.1465621-3-skpgkp2@gmail.com> X-Spam-Status: No, score=-3028.1 required=5.0 tests=BAYES_00, DKIM_SIGNED, DKIM_VALID, DKIM_VALID_AU, DKIM_VALID_EF, FREEMAIL_FROM, GIT_PATCH_0, KAM_SHORT, KAM_STOCKGEN, RCVD_IN_BARRACUDACENTRAL, RCVD_IN_DNSWL_NONE, SPF_HELO_NONE, SPF_PASS, TXREP autolearn=ham autolearn_force=no version=3.4.4 X-Spam-Checker-Version: SpamAssassin 3.4.4 (2020-01-24) on server2.sourceware.org X-BeenThere: libc-alpha@sourceware.org X-Mailman-Version: 2.1.29 Precedence: list List-Id: Libc-alpha mailing list List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , X-List-Received-Date: Wed, 29 Dec 2021 21:25:36 -0000 On Tue, Dec 28, 2021 at 10:39:44PM -0800, Sunil K Pandey wrote: > Implement vectorized asin/asinf containing SSE, AVX, AVX2 and > AVX512 versions for libmvec as per vector ABI. It also contains > accuracy and ABI tests for vector asin/asinf with regenerated ulps. > --- > bits/libm-simd-decl-stubs.h | 11 + > math/bits/mathcalls.h | 2 +- > .../unix/sysv/linux/x86_64/libmvec.abilist | 8 + > sysdeps/x86/fpu/bits/math-vector.h | 4 + > .../x86/fpu/finclude/math-vector-fortran.h | 4 + > sysdeps/x86_64/fpu/Makeconfig | 1 + > sysdeps/x86_64/fpu/Versions | 2 + > sysdeps/x86_64/fpu/libm-test-ulps | 20 ++ > .../fpu/multiarch/svml_d_asin2_core-sse2.S | 20 ++ > .../x86_64/fpu/multiarch/svml_d_asin2_core.c | 27 ++ > .../fpu/multiarch/svml_d_asin2_core_sse4.S | 288 +++++++++++++++++ > .../fpu/multiarch/svml_d_asin4_core-sse.S | 20 ++ > .../x86_64/fpu/multiarch/svml_d_asin4_core.c | 27 ++ > .../fpu/multiarch/svml_d_asin4_core_avx2.S | 273 ++++++++++++++++ > .../fpu/multiarch/svml_d_asin8_core-avx2.S | 20 ++ > .../x86_64/fpu/multiarch/svml_d_asin8_core.c | 27 ++ > .../fpu/multiarch/svml_d_asin8_core_avx512.S | 295 ++++++++++++++++++ > .../fpu/multiarch/svml_s_asinf16_core-avx2.S | 20 ++ > .../fpu/multiarch/svml_s_asinf16_core.c | 28 ++ > .../multiarch/svml_s_asinf16_core_avx512.S | 260 +++++++++++++++ > .../fpu/multiarch/svml_s_asinf4_core-sse2.S | 20 ++ > .../x86_64/fpu/multiarch/svml_s_asinf4_core.c | 28 ++ > .../fpu/multiarch/svml_s_asinf4_core_sse4.S | 252 +++++++++++++++ > .../fpu/multiarch/svml_s_asinf8_core-sse.S | 20 ++ > .../x86_64/fpu/multiarch/svml_s_asinf8_core.c | 28 ++ > .../fpu/multiarch/svml_s_asinf8_core_avx2.S | 249 +++++++++++++++ > sysdeps/x86_64/fpu/svml_d_asin2_core.S | 29 ++ > sysdeps/x86_64/fpu/svml_d_asin4_core.S | 29 ++ > sysdeps/x86_64/fpu/svml_d_asin4_core_avx.S | 25 ++ > sysdeps/x86_64/fpu/svml_d_asin8_core.S | 25 ++ > sysdeps/x86_64/fpu/svml_s_asinf16_core.S | 25 ++ > sysdeps/x86_64/fpu/svml_s_asinf4_core.S | 29 ++ > sysdeps/x86_64/fpu/svml_s_asinf8_core.S | 29 ++ > sysdeps/x86_64/fpu/svml_s_asinf8_core_avx.S | 25 ++ > .../x86_64/fpu/test-double-libmvec-asin-avx.c | 1 + > .../fpu/test-double-libmvec-asin-avx2.c | 1 + > .../fpu/test-double-libmvec-asin-avx512f.c | 1 + > sysdeps/x86_64/fpu/test-double-libmvec-asin.c | 3 + > .../x86_64/fpu/test-double-vlen2-wrappers.c | 1 + > .../fpu/test-double-vlen4-avx2-wrappers.c | 1 + > .../x86_64/fpu/test-double-vlen4-wrappers.c | 1 + > .../x86_64/fpu/test-double-vlen8-wrappers.c | 1 + > .../x86_64/fpu/test-float-libmvec-asinf-avx.c | 1 + > .../fpu/test-float-libmvec-asinf-avx2.c | 1 + > .../fpu/test-float-libmvec-asinf-avx512f.c | 1 + > sysdeps/x86_64/fpu/test-float-libmvec-asinf.c | 3 + > .../x86_64/fpu/test-float-vlen16-wrappers.c | 1 + > .../x86_64/fpu/test-float-vlen4-wrappers.c | 1 + > .../fpu/test-float-vlen8-avx2-wrappers.c | 1 + > .../x86_64/fpu/test-float-vlen8-wrappers.c | 1 + > 50 files changed, 2189 insertions(+), 1 deletion(-) > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core-sse2.S > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core.c > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core_sse4.S > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core-sse.S > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core.c > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core_avx2.S > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core-avx2.S > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core.c > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core_avx512.S > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core-avx2.S > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core.c > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core_avx512.S > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core-sse2.S > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core.c > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core_sse4.S > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core-sse.S > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core.c > create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core_avx2.S > create mode 100644 sysdeps/x86_64/fpu/svml_d_asin2_core.S > create mode 100644 sysdeps/x86_64/fpu/svml_d_asin4_core.S > create mode 100644 sysdeps/x86_64/fpu/svml_d_asin4_core_avx.S > create mode 100644 sysdeps/x86_64/fpu/svml_d_asin8_core.S > create mode 100644 sysdeps/x86_64/fpu/svml_s_asinf16_core.S > create mode 100644 sysdeps/x86_64/fpu/svml_s_asinf4_core.S > create mode 100644 sysdeps/x86_64/fpu/svml_s_asinf8_core.S > create mode 100644 sysdeps/x86_64/fpu/svml_s_asinf8_core_avx.S > create mode 100644 sysdeps/x86_64/fpu/test-double-libmvec-asin-avx.c > create mode 100644 sysdeps/x86_64/fpu/test-double-libmvec-asin-avx2.c > create mode 100644 sysdeps/x86_64/fpu/test-double-libmvec-asin-avx512f.c > create mode 100644 sysdeps/x86_64/fpu/test-double-libmvec-asin.c > create mode 100644 sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx.c > create mode 100644 sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx2.c > create mode 100644 sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx512f.c > create mode 100644 sysdeps/x86_64/fpu/test-float-libmvec-asinf.c > > diff --git a/bits/libm-simd-decl-stubs.h b/bits/libm-simd-decl-stubs.h > index b4647ca918..ae8ee882d0 100644 > --- a/bits/libm-simd-decl-stubs.h > +++ b/bits/libm-simd-decl-stubs.h > @@ -120,4 +120,15 @@ > #define __DECL_SIMD_atanf32x > #define __DECL_SIMD_atanf64x > #define __DECL_SIMD_atanf128x > + > +#define __DECL_SIMD_asin > +#define __DECL_SIMD_asinf > +#define __DECL_SIMD_asinl > +#define __DECL_SIMD_asinf16 > +#define __DECL_SIMD_asinf32 > +#define __DECL_SIMD_asinf64 > +#define __DECL_SIMD_asinf128 > +#define __DECL_SIMD_asinf32x > +#define __DECL_SIMD_asinf64x > +#define __DECL_SIMD_asinf128x > #endif > diff --git a/math/bits/mathcalls.h b/math/bits/mathcalls.h > index 3e27c21f21..bb53b7021e 100644 > --- a/math/bits/mathcalls.h > +++ b/math/bits/mathcalls.h > @@ -52,7 +52,7 @@ > /* Arc cosine of X. */ > __MATHCALL_VEC (acos,, (_Mdouble_ __x)); > /* Arc sine of X. */ > -__MATHCALL (asin,, (_Mdouble_ __x)); > +__MATHCALL_VEC (asin,, (_Mdouble_ __x)); > /* Arc tangent of X. */ > __MATHCALL_VEC (atan,, (_Mdouble_ __x)); > /* Arc tangent of Y/X. */ > diff --git a/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist b/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist > index a93258db6f..ab03a07f92 100644 > --- a/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist > +++ b/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist > @@ -47,18 +47,26 @@ GLIBC_2.22 _ZGVeN8v_sin F > GLIBC_2.22 _ZGVeN8vv_pow F > GLIBC_2.22 _ZGVeN8vvv_sincos F > GLIBC_2.35 _ZGVbN2v_acos F > +GLIBC_2.35 _ZGVbN2v_asin F > GLIBC_2.35 _ZGVbN2v_atan F > GLIBC_2.35 _ZGVbN4v_acosf F > +GLIBC_2.35 _ZGVbN4v_asinf F > GLIBC_2.35 _ZGVbN4v_atanf F > GLIBC_2.35 _ZGVcN4v_acos F > +GLIBC_2.35 _ZGVcN4v_asin F > GLIBC_2.35 _ZGVcN4v_atan F > GLIBC_2.35 _ZGVcN8v_acosf F > +GLIBC_2.35 _ZGVcN8v_asinf F > GLIBC_2.35 _ZGVcN8v_atanf F > GLIBC_2.35 _ZGVdN4v_acos F > +GLIBC_2.35 _ZGVdN4v_asin F > GLIBC_2.35 _ZGVdN4v_atan F > GLIBC_2.35 _ZGVdN8v_acosf F > +GLIBC_2.35 _ZGVdN8v_asinf F > GLIBC_2.35 _ZGVdN8v_atanf F > GLIBC_2.35 _ZGVeN16v_acosf F > +GLIBC_2.35 _ZGVeN16v_asinf F > GLIBC_2.35 _ZGVeN16v_atanf F > GLIBC_2.35 _ZGVeN8v_acos F > +GLIBC_2.35 _ZGVeN8v_asin F > GLIBC_2.35 _ZGVeN8v_atan F > diff --git a/sysdeps/x86/fpu/bits/math-vector.h b/sysdeps/x86/fpu/bits/math-vector.h > index 1c0e5c5e35..73cb8849ff 100644 > --- a/sysdeps/x86/fpu/bits/math-vector.h > +++ b/sysdeps/x86/fpu/bits/math-vector.h > @@ -66,6 +66,10 @@ > # define __DECL_SIMD_atan __DECL_SIMD_x86_64 > # undef __DECL_SIMD_atanf > # define __DECL_SIMD_atanf __DECL_SIMD_x86_64 > +# undef __DECL_SIMD_asin > +# define __DECL_SIMD_asin __DECL_SIMD_x86_64 > +# undef __DECL_SIMD_asinf > +# define __DECL_SIMD_asinf __DECL_SIMD_x86_64 > > # endif > #endif > diff --git a/sysdeps/x86/fpu/finclude/math-vector-fortran.h b/sysdeps/x86/fpu/finclude/math-vector-fortran.h > index ddcccb11d7..4552c2bdfa 100644 > --- a/sysdeps/x86/fpu/finclude/math-vector-fortran.h > +++ b/sysdeps/x86/fpu/finclude/math-vector-fortran.h > @@ -32,6 +32,8 @@ > !GCC$ builtin (acosf) attributes simd (notinbranch) if('x86_64') > !GCC$ builtin (atan) attributes simd (notinbranch) if('x86_64') > !GCC$ builtin (atanf) attributes simd (notinbranch) if('x86_64') > +!GCC$ builtin (asin) attributes simd (notinbranch) if('x86_64') > +!GCC$ builtin (asinf) attributes simd (notinbranch) if('x86_64') > > !GCC$ builtin (cos) attributes simd (notinbranch) if('x32') > !GCC$ builtin (cosf) attributes simd (notinbranch) if('x32') > @@ -49,3 +51,5 @@ > !GCC$ builtin (acosf) attributes simd (notinbranch) if('x32') > !GCC$ builtin (atan) attributes simd (notinbranch) if('x32') > !GCC$ builtin (atanf) attributes simd (notinbranch) if('x32') > +!GCC$ builtin (asin) attributes simd (notinbranch) if('x32') > +!GCC$ builtin (asinf) attributes simd (notinbranch) if('x32') > diff --git a/sysdeps/x86_64/fpu/Makeconfig b/sysdeps/x86_64/fpu/Makeconfig > index dae0887f13..e0eae0b196 100644 > --- a/sysdeps/x86_64/fpu/Makeconfig > +++ b/sysdeps/x86_64/fpu/Makeconfig > @@ -23,6 +23,7 @@ postclean-generated += libmvec.mk > # Define for both math and mathvec directories. > libmvec-funcs = \ > acos \ > + asin \ > atan \ > cos \ > exp \ > diff --git a/sysdeps/x86_64/fpu/Versions b/sysdeps/x86_64/fpu/Versions > index 424f6d526e..10baf869a5 100644 > --- a/sysdeps/x86_64/fpu/Versions > +++ b/sysdeps/x86_64/fpu/Versions > @@ -15,8 +15,10 @@ libmvec { > } > GLIBC_2.35 { > _ZGVbN2v_acos; _ZGVcN4v_acos; _ZGVdN4v_acos; _ZGVeN8v_acos; > + _ZGVbN2v_asin; _ZGVcN4v_asin; _ZGVdN4v_asin; _ZGVeN8v_asin; > _ZGVbN2v_atan; _ZGVcN4v_atan; _ZGVdN4v_atan; _ZGVeN8v_atan; > _ZGVbN4v_acosf; _ZGVcN8v_acosf; _ZGVdN8v_acosf; _ZGVeN16v_acosf; > + _ZGVbN4v_asinf; _ZGVcN8v_asinf; _ZGVdN8v_asinf; _ZGVeN16v_asinf; > _ZGVbN4v_atanf; _ZGVcN8v_atanf; _ZGVdN8v_atanf; _ZGVeN16v_atanf; > } > } > diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps > index 2e64e59803..ea0f833381 100644 > --- a/sysdeps/x86_64/fpu/libm-test-ulps > +++ b/sysdeps/x86_64/fpu/libm-test-ulps > @@ -93,6 +93,26 @@ float: 1 > float128: 2 > ldouble: 1 > > +Function: "asin_vlen16": > +float: 1 > + > +Function: "asin_vlen2": > +double: 1 > + > +Function: "asin_vlen4": > +double: 1 > +float: 1 > + > +Function: "asin_vlen4_avx2": > +double: 1 > + > +Function: "asin_vlen8": > +double: 1 > +float: 1 > + > +Function: "asin_vlen8_avx2": > +float: 1 > + > Function: "asinh": > double: 2 > float: 2 > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core-sse2.S b/sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core-sse2.S > new file mode 100644 > index 0000000000..57e1d41a7b > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core-sse2.S > @@ -0,0 +1,20 @@ > +/* SSE2 version of vectorized asin, vector length is 2. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define _ZGVbN2v_asin _ZGVbN2v_asin_sse2 > +#include "../svml_d_asin2_core.S" > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core.c b/sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core.c > new file mode 100644 > index 0000000000..e46c3af81e > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core.c > @@ -0,0 +1,27 @@ > +/* Multiple versions of vectorized asin, vector length is 2. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define SYMBOL_NAME _ZGVbN2v_asin > +#include "ifunc-mathvec-sse4_1.h" > + > +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); > + > +#ifdef SHARED > +__hidden_ver1 (_ZGVbN2v_asin, __GI__ZGVbN2v_asin, __redirect__ZGVbN2v_asin) > + __attribute__ ((visibility ("hidden"))); > +#endif > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core_sse4.S b/sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core_sse4.S > new file mode 100644 > index 0000000000..a6f7a41623 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_asin2_core_sse4.S > @@ -0,0 +1,288 @@ > +/* Function asin vectorized with SSE4. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + https://www.gnu.org/licenses/. */ > + > +/* > + * ALGORITHM DESCRIPTION: > + * > + * SelMask = (|x| >= 0.5) ? 1 : 0; > + * R = SelMask ? sqrt(0.5 - 0.5*|x|) : |x| > + * asin(x) = (SelMask ? (Pi/2 - 2*Poly(R)) : Poly(R))*(-1)^sign(x) > + * > + */ > + > +/* Offsets for data table __svml_dasin_data_internal > + */ > +#define AbsMask 0 > +#define OneHalf 16 > +#define SmallNorm 32 > +#define One 48 > +#define Two 64 > +#define sqrt_coeff 80 > +#define poly_coeff 144 > +#define Pi2H 336 > + > +#include > + > + .text > + .section .text.sse4,"ax",@progbits > +ENTRY(_ZGVbN2v_asin_sse4) > + subq $72, %rsp > + cfi_def_cfa_offset(80) > + movaps %xmm0, %xmm5 > + movups __svml_dasin_data_internal(%rip), %xmm3 > + movups OneHalf+__svml_dasin_data_internal(%rip), %xmm8 > + > +/* x = |arg| */ > + movaps %xmm3, %xmm4 > + andps %xmm5, %xmm4 > + > +/* Y = 0.5 - 0.5*x */ > + movaps %xmm8, %xmm6 > + mulpd %xmm4, %xmm6 > + movaps %xmm8, %xmm14 > + > +/* x^2 */ > + movaps %xmm4, %xmm2 > + subpd %xmm6, %xmm14 > + mulpd %xmm4, %xmm2 > + > +/* S ~ -2*sqrt(Y) */ > + cvtpd2ps %xmm14, %xmm9 > + minpd %xmm14, %xmm2 > + movlhps %xmm9, %xmm9 > + movaps %xmm14, %xmm15 > + rsqrtps %xmm9, %xmm10 > + cmpltpd SmallNorm+__svml_dasin_data_internal(%rip), %xmm15 > + addpd %xmm14, %xmm14 > + cvtps2pd %xmm10, %xmm11 > + andnps %xmm11, %xmm15 > + movaps %xmm4, %xmm1 > + movaps %xmm15, %xmm12 > + andnps %xmm5, %xmm3 > + mulpd %xmm15, %xmm12 > + mulpd %xmm14, %xmm15 > + mulpd %xmm12, %xmm14 > + cmpnltpd %xmm8, %xmm1 > + subpd Two+__svml_dasin_data_internal(%rip), %xmm14 > + > +/* polynomial */ > + movups poly_coeff+__svml_dasin_data_internal(%rip), %xmm6 > + movaps %xmm2, %xmm12 > + mulpd %xmm2, %xmm6 > + mulpd %xmm2, %xmm12 > + addpd poly_coeff+16+__svml_dasin_data_internal(%rip), %xmm6 > + movups One+__svml_dasin_data_internal(%rip), %xmm7 > + movaps %xmm12, %xmm8 > + cmpltpd %xmm4, %xmm7 > + mulpd %xmm12, %xmm6 > + movmskpd %xmm7, %edx > + movups poly_coeff+32+__svml_dasin_data_internal(%rip), %xmm9 > + movaps %xmm14, %xmm0 > + movups poly_coeff+64+__svml_dasin_data_internal(%rip), %xmm7 > + mulpd %xmm2, %xmm9 > + mulpd %xmm2, %xmm7 > + addpd poly_coeff+48+__svml_dasin_data_internal(%rip), %xmm9 > + addpd poly_coeff+80+__svml_dasin_data_internal(%rip), %xmm7 > + mulpd %xmm12, %xmm8 > + mulpd %xmm12, %xmm7 > + addpd %xmm6, %xmm9 > + mulpd %xmm15, %xmm0 > + mulpd %xmm8, %xmm9 > + movups poly_coeff+96+__svml_dasin_data_internal(%rip), %xmm10 > + mulpd %xmm2, %xmm10 > + movups sqrt_coeff+__svml_dasin_data_internal(%rip), %xmm13 > + mulpd %xmm14, %xmm13 > + addpd poly_coeff+112+__svml_dasin_data_internal(%rip), %xmm10 > + addpd sqrt_coeff+16+__svml_dasin_data_internal(%rip), %xmm13 > + addpd %xmm7, %xmm10 > + mulpd %xmm14, %xmm13 > + addpd %xmm9, %xmm10 > + addpd sqrt_coeff+32+__svml_dasin_data_internal(%rip), %xmm13 > + mulpd %xmm12, %xmm10 > + mulpd %xmm13, %xmm14 > + movups poly_coeff+128+__svml_dasin_data_internal(%rip), %xmm11 > + mulpd %xmm2, %xmm11 > + addpd sqrt_coeff+48+__svml_dasin_data_internal(%rip), %xmm14 > + addpd poly_coeff+144+__svml_dasin_data_internal(%rip), %xmm11 > + mulpd %xmm14, %xmm0 > + addpd %xmm10, %xmm11 > + subpd %xmm15, %xmm0 > + mulpd %xmm11, %xmm12 > + movups poly_coeff+160+__svml_dasin_data_internal(%rip), %xmm13 > + movaps %xmm1, %xmm14 > + mulpd %xmm2, %xmm13 > + addpd poly_coeff+176+__svml_dasin_data_internal(%rip), %xmm13 > + addpd %xmm12, %xmm13 > + mulpd %xmm13, %xmm2 > + andnps %xmm4, %xmm14 > + andps %xmm1, %xmm0 > + orps %xmm0, %xmm14 > + mulpd %xmm14, %xmm2 > + addpd %xmm2, %xmm14 > + movups Pi2H+__svml_dasin_data_internal(%rip), %xmm0 > + andps %xmm1, %xmm0 > + addpd %xmm14, %xmm0 > + pxor %xmm3, %xmm0 > + testl %edx, %edx > + > +/* Go to special inputs processing branch */ > + jne L(SPECIAL_VALUES_BRANCH) > + # LOE rbx rbp r12 r13 r14 r15 edx xmm0 xmm5 > + > +/* Restore registers > + * and exit the function > + */ > + > +L(EXIT): > + addq $72, %rsp > + cfi_def_cfa_offset(8) > + ret > + cfi_def_cfa_offset(80) > + > +/* Branch to process > + * special inputs > + */ > + > +L(SPECIAL_VALUES_BRANCH): > + movups %xmm5, 32(%rsp) > + movups %xmm0, 48(%rsp) > + # LOE rbx rbp r12 r13 r14 r15 edx > + > + xorl %eax, %eax > + movq %r12, 16(%rsp) > + cfi_offset(12, -64) > + movl %eax, %r12d > + movq %r13, 8(%rsp) > + cfi_offset(13, -72) > + movl %edx, %r13d > + movq %r14, (%rsp) > + cfi_offset(14, -80) > + # LOE rbx rbp r15 r12d r13d > + > +/* Range mask > + * bits check > + */ > + > +L(RANGEMASK_CHECK): > + btl %r12d, %r13d > + > +/* Call scalar math function */ > + jc L(SCALAR_MATH_CALL) > + # LOE rbx rbp r15 r12d r13d > + > +/* Special inputs > + * processing loop > + */ > + > +L(SPECIAL_VALUES_LOOP): > + incl %r12d > + cmpl $2, %r12d > + > +/* Check bits in range mask */ > + jl L(RANGEMASK_CHECK) > + # LOE rbx rbp r15 r12d r13d > + > + movq 16(%rsp), %r12 > + cfi_restore(12) > + movq 8(%rsp), %r13 > + cfi_restore(13) > + movq (%rsp), %r14 > + cfi_restore(14) > + movups 48(%rsp), %xmm0 > + > +/* Go to exit */ > + jmp L(EXIT) > + cfi_offset(12, -64) > + cfi_offset(13, -72) > + cfi_offset(14, -80) > + # LOE rbx rbp r12 r13 r14 r15 xmm0 > + > +/* Scalar math fucntion call > + * to process special input > + */ > + > +L(SCALAR_MATH_CALL): > + movl %r12d, %r14d > + movsd 32(%rsp,%r14,8), %xmm0 > + call asin@PLT > + # LOE rbx rbp r14 r15 r12d r13d xmm0 > + > + movsd %xmm0, 48(%rsp,%r14,8) > + > +/* Process special inputs in loop */ > + jmp L(SPECIAL_VALUES_LOOP) > + # LOE rbx rbp r15 r12d r13d > +END(_ZGVbN2v_asin_sse4) > + > + .section .rodata, "a" > + .align 16 > + > +#ifdef __svml_dasin_data_internal_typedef > +typedef unsigned int VUINT32; > +typedef struct { > + __declspec(align(16)) VUINT32 AbsMask[2][2]; > + __declspec(align(16)) VUINT32 OneHalf[2][2]; > + __declspec(align(16)) VUINT32 SmallNorm[2][2]; > + __declspec(align(16)) VUINT32 One[2][2]; > + __declspec(align(16)) VUINT32 Two[2][2]; > + __declspec(align(16)) VUINT32 sqrt_coeff[4][2][2]; > + __declspec(align(16)) VUINT32 poly_coeff[12][2][2]; > + __declspec(align(16)) VUINT32 Pi2H[2][2]; > +} __svml_dasin_data_internal; > +#endif > +__svml_dasin_data_internal: > + /*== AbsMask ==*/ > + .quad 0x7fffffffffffffff, 0x7fffffffffffffff > + /*== OneHalf ==*/ > + .align 16 > + .quad 0x3fe0000000000000, 0x3fe0000000000000 > + /*== SmallNorm ==*/ > + .align 16 > + .quad 0x3000000000000000, 0x3000000000000000 > + /*== One ==*/ > + .align 16 > + .quad 0x3ff0000000000000, 0x3ff0000000000000 > + /*== Two ==*/ > + .align 16 > + .quad 0x4000000000000000, 0x4000000000000000 > + /*== sqrt_coeff[4] ==*/ > + .align 16 > + .quad 0xbf918000993B24C3, 0xbf918000993B24C3 /* sqrt_coeff4 */ > + .quad 0x3fa400006F70D42D, 0x3fa400006F70D42D /* sqrt_coeff3 */ > + .quad 0xbfb7FFFFFFFFFE97, 0xbfb7FFFFFFFFFE97 /* sqrt_coeff2 */ > + .quad 0x3fcFFFFFFFFFFF9D, 0x3fcFFFFFFFFFFF9D /* sqrt_coeff1 */ > + /*== poly_coeff[12] ==*/ > + .align 16 > + .quad 0x3fa07520C70EB909, 0x3fa07520C70EB909 /* poly_coeff12 */ > + .quad 0xbf90FB17F7DBB0ED, 0xbf90FB17F7DBB0ED /* poly_coeff11 */ > + .quad 0x3f943F44BFBC3BAE, 0x3f943F44BFBC3BAE /* poly_coeff10 */ > + .quad 0x3f7A583395D45ED5, 0x3f7A583395D45ED5 /* poly_coeff9 */ > + .quad 0x3f88F8DC2AFCCAD6, 0x3f88F8DC2AFCCAD6 /* poly_coeff8 */ > + .quad 0x3f8C6DBBCB88BD57, 0x3f8C6DBBCB88BD57 /* poly_coeff7 */ > + .quad 0x3f91C6DCF538AD2E, 0x3f91C6DCF538AD2E /* poly_coeff6 */ > + .quad 0x3f96E89CEBDEFadd, 0x3f96E89CEBDEFadd /* poly_coeff5 */ > + .quad 0x3f9F1C72E13AD8BE, 0x3f9F1C72E13AD8BE /* poly_coeff4 */ > + .quad 0x3fa6DB6DB3B445F8, 0x3fa6DB6DB3B445F8 /* poly_coeff3 */ > + .quad 0x3fb333333337E0DE, 0x3fb333333337E0DE /* poly_coeff2 */ > + .quad 0x3fc555555555529C, 0x3fc555555555529C /* poly_coeff1 */ > + /*== Pi2H ==*/ > + .align 16 > + .quad 0x3ff921fb54442d18, 0x3ff921fb54442d18 > + .align 16 > + .type __svml_dasin_data_internal,@object > + .size __svml_dasin_data_internal,.-__svml_dasin_data_internal > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core-sse.S b/sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core-sse.S > new file mode 100644 > index 0000000000..1006fddc59 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core-sse.S > @@ -0,0 +1,20 @@ > +/* SSE version of vectorized asin, vector length is 4. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define _ZGVdN4v_asin _ZGVdN4v_asin_sse_wrapper > +#include "../svml_d_asin4_core.S" > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core.c b/sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core.c > new file mode 100644 > index 0000000000..b896516f5e > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core.c > @@ -0,0 +1,27 @@ > +/* Multiple versions of vectorized asin, vector length is 4. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define SYMBOL_NAME _ZGVdN4v_asin > +#include "ifunc-mathvec-avx2.h" > + > +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); > + > +#ifdef SHARED > +__hidden_ver1 (_ZGVdN4v_asin, __GI__ZGVdN4v_asin, __redirect__ZGVdN4v_asin) > + __attribute__ ((visibility ("hidden"))); > +#endif > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core_avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core_avx2.S > new file mode 100644 > index 0000000000..80467b616f > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_asin4_core_avx2.S > @@ -0,0 +1,273 @@ > +/* Function asin vectorized with AVX2. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + https://www.gnu.org/licenses/. */ > + > +/* > + * ALGORITHM DESCRIPTION: > + * > + * SelMask = (|x| >= 0.5) ? 1 : 0; > + * R = SelMask ? sqrt(0.5 - 0.5*|x|) : |x| > + * asin(x) = (SelMask ? (Pi/2 - 2*Poly(R)) : Poly(R))*(-1)^sign(x) > + * > + */ > + > +/* Offsets for data table __svml_dasin_data_internal > + */ > +#define AbsMask 0 > +#define OneHalf 32 > +#define SmallNorm 64 > +#define One 96 > +#define Two 128 > +#define sqrt_coeff 160 > +#define poly_coeff 288 > +#define Pi2H 672 > + > +#include > + > + .text > + .section .text.avx2,"ax",@progbits > +ENTRY(_ZGVdN4v_asin_avx2) > + pushq %rbp > + cfi_def_cfa_offset(16) > + movq %rsp, %rbp > + cfi_def_cfa(6, 16) > + cfi_offset(6, -16) > + andq $-32, %rsp > + subq $96, %rsp > + vmovupd __svml_dasin_data_internal(%rip), %ymm6 > + vmovupd OneHalf+__svml_dasin_data_internal(%rip), %ymm10 > + vmovupd One+__svml_dasin_data_internal(%rip), %ymm8 > + vmovapd %ymm0, %ymm5 > + > +/* x = |arg| */ > + vandpd %ymm5, %ymm6, %ymm4 > + > +/* Y = 0.5 - 0.5*x */ > + vmovapd %ymm10, %ymm15 > + vfnmadd231pd %ymm4, %ymm10, %ymm15 > + > +/* x^2 */ > + vmulpd %ymm4, %ymm4, %ymm7 > + vcmplt_oqpd %ymm4, %ymm8, %ymm9 > + > +/* S ~ -2*sqrt(Y) */ > + vcmplt_oqpd SmallNorm+__svml_dasin_data_internal(%rip), %ymm15, %ymm13 > + vminpd %ymm15, %ymm7, %ymm2 > + vaddpd %ymm15, %ymm15, %ymm7 > + vcmpnlt_uqpd %ymm10, %ymm4, %ymm1 > + vcvtpd2ps %ymm15, %xmm11 > + vmovupd poly_coeff+64+__svml_dasin_data_internal(%rip), %ymm10 > + vmulpd %ymm2, %ymm2, %ymm15 > + vrsqrtps %xmm11, %xmm12 > + vmovupd poly_coeff+192+__svml_dasin_data_internal(%rip), %ymm11 > + vfmadd213pd poly_coeff+96+__svml_dasin_data_internal(%rip), %ymm2, %ymm10 > + vcvtps2pd %xmm12, %ymm14 > + vmulpd %ymm15, %ymm15, %ymm12 > + vfmadd213pd poly_coeff+224+__svml_dasin_data_internal(%rip), %ymm2, %ymm11 > + vandnpd %ymm14, %ymm13, %ymm0 > + vandnpd %ymm5, %ymm6, %ymm3 > + vmulpd %ymm0, %ymm0, %ymm6 > + vmovupd poly_coeff+128+__svml_dasin_data_internal(%rip), %ymm13 > + vmovupd poly_coeff+256+__svml_dasin_data_internal(%rip), %ymm14 > + vfmadd213pd poly_coeff+160+__svml_dasin_data_internal(%rip), %ymm2, %ymm13 > + vfmadd213pd poly_coeff+288+__svml_dasin_data_internal(%rip), %ymm2, %ymm14 > + vfmadd213pd %ymm11, %ymm15, %ymm13 > + vmovmskpd %ymm9, %edx > + vmulpd %ymm7, %ymm0, %ymm9 > + vfmsub213pd Two+__svml_dasin_data_internal(%rip), %ymm6, %ymm7 > + > +/* polynomial */ > + vmovupd poly_coeff+__svml_dasin_data_internal(%rip), %ymm6 > + vmovupd sqrt_coeff+__svml_dasin_data_internal(%rip), %ymm0 > + vmulpd %ymm7, %ymm9, %ymm8 > + vfmadd213pd poly_coeff+32+__svml_dasin_data_internal(%rip), %ymm2, %ymm6 > + vfmadd213pd sqrt_coeff+32+__svml_dasin_data_internal(%rip), %ymm7, %ymm0 > + vfmadd213pd %ymm10, %ymm15, %ymm6 > + vmovupd poly_coeff+320+__svml_dasin_data_internal(%rip), %ymm10 > + vfmadd213pd sqrt_coeff+64+__svml_dasin_data_internal(%rip), %ymm7, %ymm0 > + vfmadd213pd %ymm13, %ymm12, %ymm6 > + vfmadd213pd poly_coeff+352+__svml_dasin_data_internal(%rip), %ymm2, %ymm10 > + vfmadd213pd sqrt_coeff+96+__svml_dasin_data_internal(%rip), %ymm7, %ymm0 > + vfmadd213pd %ymm14, %ymm15, %ymm6 > + vfmsub213pd %ymm9, %ymm8, %ymm0 > + vfmadd213pd %ymm10, %ymm15, %ymm6 > + vblendvpd %ymm1, %ymm0, %ymm4, %ymm4 > + vmulpd %ymm6, %ymm2, %ymm2 > + vfmadd213pd %ymm4, %ymm4, %ymm2 > + vandpd Pi2H+__svml_dasin_data_internal(%rip), %ymm1, %ymm1 > + vaddpd %ymm2, %ymm1, %ymm0 > + vxorpd %ymm3, %ymm0, %ymm0 > + testl %edx, %edx > + > +/* Go to special inputs processing branch */ > + jne L(SPECIAL_VALUES_BRANCH) > + # LOE rbx r12 r13 r14 r15 edx ymm0 ymm5 > + > +/* Restore registers > + * and exit the function > + */ > + > +L(EXIT): > + movq %rbp, %rsp > + popq %rbp > + cfi_def_cfa(7, 8) > + cfi_restore(6) > + ret > + cfi_def_cfa(6, 16) > + cfi_offset(6, -16) > + > +/* Branch to process > + * special inputs > + */ > + > +L(SPECIAL_VALUES_BRANCH): > + vmovupd %ymm5, 32(%rsp) > + vmovupd %ymm0, 64(%rsp) > + # LOE rbx r12 r13 r14 r15 edx ymm0 > + > + xorl %eax, %eax > + # LOE rbx r12 r13 r14 r15 eax edx > + > + vzeroupper > + movq %r12, 16(%rsp) > + /* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -80; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xb0, 0xff, 0xff, 0xff, 0x22 > + movl %eax, %r12d > + movq %r13, 8(%rsp) > + /* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -88; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa8, 0xff, 0xff, 0xff, 0x22 > + movl %edx, %r13d > + movq %r14, (%rsp) > + /* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -96; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa0, 0xff, 0xff, 0xff, 0x22 > + # LOE rbx r15 r12d r13d > + > +/* Range mask > + * bits check > + */ > + > +L(RANGEMASK_CHECK): > + btl %r12d, %r13d > + > +/* Call scalar math function */ > + jc L(SCALAR_MATH_CALL) > + # LOE rbx r15 r12d r13d > + > +/* Special inputs > + * processing loop > + */ > + > +L(SPECIAL_VALUES_LOOP): > + incl %r12d > + cmpl $4, %r12d > + > +/* Check bits in range mask */ > + jl L(RANGEMASK_CHECK) > + # LOE rbx r15 r12d r13d > + > + movq 16(%rsp), %r12 > + cfi_restore(12) > + movq 8(%rsp), %r13 > + cfi_restore(13) > + movq (%rsp), %r14 > + cfi_restore(14) > + vmovupd 64(%rsp), %ymm0 > + > +/* Go to exit */ > + jmp L(EXIT) > + /* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -80; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xb0, 0xff, 0xff, 0xff, 0x22 > + /* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -88; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa8, 0xff, 0xff, 0xff, 0x22 > + /* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -96; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa0, 0xff, 0xff, 0xff, 0x22 > + # LOE rbx r12 r13 r14 r15 ymm0 > + > +/* Scalar math fucntion call > + * to process special input > + */ > + > +L(SCALAR_MATH_CALL): > + movl %r12d, %r14d > + movsd 32(%rsp,%r14,8), %xmm0 > + call asin@PLT > + # LOE rbx r14 r15 r12d r13d xmm0 > + > + movsd %xmm0, 64(%rsp,%r14,8) > + > +/* Process special inputs in loop */ > + jmp L(SPECIAL_VALUES_LOOP) > + # LOE rbx r15 r12d r13d > +END(_ZGVdN4v_asin_avx2) > + > + .section .rodata, "a" > + .align 32 > + > +#ifdef __svml_dasin_data_internal_typedef > +typedef unsigned int VUINT32; > +typedef struct { > + __declspec(align(32)) VUINT32 AbsMask[4][2]; > + __declspec(align(32)) VUINT32 OneHalf[4][2]; > + __declspec(align(32)) VUINT32 SmallNorm[4][2]; > + __declspec(align(32)) VUINT32 One[4][2]; > + __declspec(align(32)) VUINT32 Two[4][2]; > + __declspec(align(32)) VUINT32 sqrt_coeff[4][4][2]; > + __declspec(align(32)) VUINT32 poly_coeff[12][4][2]; > + __declspec(align(32)) VUINT32 Pi2H[4][2]; > +} __svml_dasin_data_internal; > +#endif > +__svml_dasin_data_internal: > + /*== AbsMask ==*/ > + .quad 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff > + /*== OneHalf ==*/ > + .align 32 > + .quad 0x3fe0000000000000, 0x3fe0000000000000, 0x3fe0000000000000, 0x3fe0000000000000 > + /*== SmallNorm ==*/ > + .align 32 > + .quad 0x3000000000000000, 0x3000000000000000, 0x3000000000000000, 0x3000000000000000 > + /*== One ==*/ > + .align 32 > + .quad 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000 > + /*== Two ==*/ > + .align 32 > + .quad 0x4000000000000000, 0x4000000000000000, 0x4000000000000000, 0x4000000000000000 > + /*== sqrt_coeff[4] ==*/ > + .align 32 > + .quad 0xbf918000993B24C3, 0xbf918000993B24C3, 0xbf918000993B24C3, 0xbf918000993B24C3 /* sqrt_coeff4 */ > + .quad 0x3fa400006F70D42D, 0x3fa400006F70D42D, 0x3fa400006F70D42D, 0x3fa400006F70D42D /* sqrt_coeff3 */ > + .quad 0xbfb7FFFFFFFFFE97, 0xbfb7FFFFFFFFFE97, 0xbfb7FFFFFFFFFE97, 0xbfb7FFFFFFFFFE97 /* sqrt_coeff2 */ > + .quad 0x3fcFFFFFFFFFFF9D, 0x3fcFFFFFFFFFFF9D, 0x3fcFFFFFFFFFFF9D, 0x3fcFFFFFFFFFFF9D /* sqrt_coeff1 */ > + /*== poly_coeff[12] ==*/ > + .align 32 > + .quad 0x3fa07520C70EB909, 0x3fa07520C70EB909, 0x3fa07520C70EB909, 0x3fa07520C70EB909 /* poly_coeff12 */ > + .quad 0xbf90FB17F7DBB0ED, 0xbf90FB17F7DBB0ED, 0xbf90FB17F7DBB0ED, 0xbf90FB17F7DBB0ED /* poly_coeff11 */ > + .quad 0x3f943F44BFBC3BAE, 0x3f943F44BFBC3BAE, 0x3f943F44BFBC3BAE, 0x3f943F44BFBC3BAE /* poly_coeff10 */ > + .quad 0x3f7A583395D45ED5, 0x3f7A583395D45ED5, 0x3f7A583395D45ED5, 0x3f7A583395D45ED5 /* poly_coeff9 */ > + .quad 0x3f88F8DC2AFCCAD6, 0x3f88F8DC2AFCCAD6, 0x3f88F8DC2AFCCAD6, 0x3f88F8DC2AFCCAD6 /* poly_coeff8 */ > + .quad 0x3f8C6DBBCB88BD57, 0x3f8C6DBBCB88BD57, 0x3f8C6DBBCB88BD57, 0x3f8C6DBBCB88BD57 /* poly_coeff7 */ > + .quad 0x3f91C6DCF538AD2E, 0x3f91C6DCF538AD2E, 0x3f91C6DCF538AD2E, 0x3f91C6DCF538AD2E /* poly_coeff6 */ > + .quad 0x3f96E89CEBDEFadd, 0x3f96E89CEBDEFadd, 0x3f96E89CEBDEFadd, 0x3f96E89CEBDEFadd /* poly_coeff5 */ > + .quad 0x3f9F1C72E13AD8BE, 0x3f9F1C72E13AD8BE, 0x3f9F1C72E13AD8BE, 0x3f9F1C72E13AD8BE /* poly_coeff4 */ > + .quad 0x3fa6DB6DB3B445F8, 0x3fa6DB6DB3B445F8, 0x3fa6DB6DB3B445F8, 0x3fa6DB6DB3B445F8 /* poly_coeff3 */ > + .quad 0x3fb333333337E0DE, 0x3fb333333337E0DE, 0x3fb333333337E0DE, 0x3fb333333337E0DE /* poly_coeff2 */ > + .quad 0x3fc555555555529C, 0x3fc555555555529C, 0x3fc555555555529C, 0x3fc555555555529C /* poly_coeff1 */ > + /*== Pi2H ==*/ > + .align 32 > + .quad 0x3ff921fb54442d18, 0x3ff921fb54442d18, 0x3ff921fb54442d18, 0x3ff921fb54442d18 > + .align 32 > + .type __svml_dasin_data_internal,@object > + .size __svml_dasin_data_internal,.-__svml_dasin_data_internal > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core-avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core-avx2.S > new file mode 100644 > index 0000000000..354a55dfaa > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core-avx2.S > @@ -0,0 +1,20 @@ > +/* AVX2 version of vectorized asin, vector length is 8. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define _ZGVeN8v_asin _ZGVeN8v_asin_avx2_wrapper > +#include "../svml_d_asin8_core.S" > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core.c b/sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core.c > new file mode 100644 > index 0000000000..b03e4a2b9c > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core.c > @@ -0,0 +1,27 @@ > +/* Multiple versions of vectorized asin, vector length is 8. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define SYMBOL_NAME _ZGVeN8v_asin > +#include "ifunc-mathvec-avx512-skx.h" > + > +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); > + > +#ifdef SHARED > +__hidden_ver1 (_ZGVeN8v_asin, __GI__ZGVeN8v_asin, __redirect__ZGVeN8v_asin) > + __attribute__ ((visibility ("hidden"))); > +#endif > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core_avx512.S b/sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core_avx512.S > new file mode 100644 > index 0000000000..b2fd8edb13 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_asin8_core_avx512.S > @@ -0,0 +1,295 @@ > +/* Function asin vectorized with AVX-512. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + https://www.gnu.org/licenses/. */ > + > +/* > + * ALGORITHM DESCRIPTION: > + * > + * SelMask = (|x| >= 0.5) ? 1 : 0; > + * R = SelMask ? sqrt(0.5 - 0.5*|x|) : |x| > + * asin(x) = (SelMask ? (Pi/2 - 2*Poly(R)) : Poly(R))*(-1)^sign(x) > + * > + */ > + > +/* Offsets for data table __svml_dasin_data_internal > + */ > +#define AbsMask 0 > +#define OneHalf 64 > +#define SmallNorm 128 > +#define One 192 > +#define Two 256 > +#define sqrt_coeff_1 320 > +#define sqrt_coeff_2 384 > +#define sqrt_coeff_3 448 > +#define sqrt_coeff_4 512 > +#define poly_coeff_1 576 > +#define poly_coeff_2 640 > +#define poly_coeff_3 704 > +#define poly_coeff_4 768 > +#define poly_coeff_5 832 > +#define poly_coeff_6 896 > +#define poly_coeff_7 960 > +#define poly_coeff_8 1024 > +#define poly_coeff_9 1088 > +#define poly_coeff_10 1152 > +#define poly_coeff_11 1216 > +#define poly_coeff_12 1280 > +#define Pi2H 1344 > + > +#include > + > + .text > + .section .text.evex512,"ax",@progbits > +ENTRY(_ZGVeN8v_asin_skx) > + pushq %rbp > + cfi_def_cfa_offset(16) > + movq %rsp, %rbp > + cfi_def_cfa(6, 16) > + cfi_offset(6, -16) > + andq $-64, %rsp > + subq $192, %rsp > + vmovups OneHalf+__svml_dasin_data_internal(%rip), %zmm8 > + > +/* S ~ -2*sqrt(Y) */ > + vmovups SmallNorm+__svml_dasin_data_internal(%rip), %zmm10 > + vmovups Two+__svml_dasin_data_internal(%rip), %zmm14 > + vmovups sqrt_coeff_1+__svml_dasin_data_internal(%rip), %zmm15 > + vmovups sqrt_coeff_2+__svml_dasin_data_internal(%rip), %zmm2 > + vmovups sqrt_coeff_3+__svml_dasin_data_internal(%rip), %zmm1 > + vmovups One+__svml_dasin_data_internal(%rip), %zmm9 > + vmovaps %zmm0, %zmm6 > + > +/* x = |arg| */ > + vandpd __svml_dasin_data_internal(%rip), %zmm6, %zmm4 > + > +/* Y = 0.5 - 0.5*x */ > + vmovaps %zmm8, %zmm11 > + vfnmadd231pd {rn-sae}, %zmm4, %zmm8, %zmm11 > + > +/* x^2 */ > + vmulpd {rn-sae}, %zmm4, %zmm4, %zmm7 > + vrsqrt14pd %zmm11, %zmm12 > + vcmppd $17, {sae}, %zmm10, %zmm11, %k1 > + vcmppd $21, {sae}, %zmm8, %zmm4, %k2 > + vcmppd $17, {sae}, %zmm4, %zmm9, %k0 > + vmovups poly_coeff_5+__svml_dasin_data_internal(%rip), %zmm10 > + > +/* polynomial */ > + vmovups poly_coeff_1+__svml_dasin_data_internal(%rip), %zmm8 > + vmovups poly_coeff_3+__svml_dasin_data_internal(%rip), %zmm9 > + vminpd {sae}, %zmm11, %zmm7, %zmm3 > + vxorpd %zmm12, %zmm12, %zmm12{%k1} > + vaddpd {rn-sae}, %zmm11, %zmm11, %zmm0 > + vxorpd %zmm6, %zmm4, %zmm5 > + vmulpd {rn-sae}, %zmm12, %zmm12, %zmm13 > + vmulpd {rn-sae}, %zmm12, %zmm0, %zmm7 > + vmovups poly_coeff_7+__svml_dasin_data_internal(%rip), %zmm11 > + vmovups poly_coeff_4+__svml_dasin_data_internal(%rip), %zmm12 > + vfmsub213pd {rn-sae}, %zmm14, %zmm13, %zmm0 > + vmovups sqrt_coeff_4+__svml_dasin_data_internal(%rip), %zmm13 > + vfmadd231pd {rn-sae}, %zmm3, %zmm9, %zmm12 > + vmovups poly_coeff_11+__svml_dasin_data_internal(%rip), %zmm9 > + vfmadd231pd {rn-sae}, %zmm0, %zmm15, %zmm2 > + vmovups poly_coeff_9+__svml_dasin_data_internal(%rip), %zmm15 > + vmulpd {rn-sae}, %zmm0, %zmm7, %zmm14 > + vfmadd213pd {rn-sae}, %zmm1, %zmm0, %zmm2 > + vmovups poly_coeff_2+__svml_dasin_data_internal(%rip), %zmm1 > + kmovw %k0, %edx > + vfmadd213pd {rn-sae}, %zmm13, %zmm0, %zmm2 > + vfmadd231pd {rn-sae}, %zmm3, %zmm8, %zmm1 > + vmovups poly_coeff_10+__svml_dasin_data_internal(%rip), %zmm8 > + vmulpd {rn-sae}, %zmm3, %zmm3, %zmm0 > + vfmsub213pd {rn-sae}, %zmm7, %zmm14, %zmm2 > + vmovups poly_coeff_6+__svml_dasin_data_internal(%rip), %zmm7 > + vfmadd231pd {rn-sae}, %zmm3, %zmm15, %zmm8 > + vfmadd213pd {rn-sae}, %zmm12, %zmm0, %zmm1 > + vblendmpd %zmm2, %zmm4, %zmm2{%k2} > + vfmadd231pd {rn-sae}, %zmm3, %zmm10, %zmm7 > + vmovups poly_coeff_8+__svml_dasin_data_internal(%rip), %zmm10 > + vmovups Pi2H+__svml_dasin_data_internal(%rip), %zmm4 > + vfmadd231pd {rn-sae}, %zmm3, %zmm11, %zmm10 > + vmovups poly_coeff_12+__svml_dasin_data_internal(%rip), %zmm11 > + vfmadd213pd {rn-sae}, %zmm10, %zmm0, %zmm7 > + vfmadd231pd {rn-sae}, %zmm3, %zmm9, %zmm11 > + vmulpd {rn-sae}, %zmm0, %zmm0, %zmm10 > + vfmadd213pd {rn-sae}, %zmm7, %zmm10, %zmm1 > + vfmadd213pd {rn-sae}, %zmm8, %zmm0, %zmm1 > + vfmadd213pd {rn-sae}, %zmm11, %zmm0, %zmm1 > + vmulpd {rn-sae}, %zmm3, %zmm1, %zmm3 > + vfmadd213pd {rn-sae}, %zmm2, %zmm2, %zmm3 > + vaddpd {rn-sae}, %zmm4, %zmm3, %zmm3{%k2} > + vxorpd %zmm5, %zmm3, %zmm0 > + testl %edx, %edx > + > +/* Go to special inputs processing branch */ > + jne L(SPECIAL_VALUES_BRANCH) > + # LOE rbx r12 r13 r14 r15 edx zmm0 zmm6 > + > +/* Restore registers > + * and exit the function > + */ > + > +L(EXIT): > + movq %rbp, %rsp > + popq %rbp > + cfi_def_cfa(7, 8) > + cfi_restore(6) > + ret > + cfi_def_cfa(6, 16) > + cfi_offset(6, -16) > + > +/* Branch to process > + * special inputs > + */ > + > +L(SPECIAL_VALUES_BRANCH): > + vmovups %zmm6, 64(%rsp) > + vmovups %zmm0, 128(%rsp) > + # LOE rbx r12 r13 r14 r15 edx zmm0 > + > + xorl %eax, %eax > + # LOE rbx r12 r13 r14 r15 eax edx > + > + vzeroupper > + movq %r12, 16(%rsp) > + /* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -176; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x50, 0xff, 0xff, 0xff, 0x22 > + movl %eax, %r12d > + movq %r13, 8(%rsp) > + /* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -184; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22 > + movl %edx, %r13d > + movq %r14, (%rsp) > + /* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -192; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 0xff, 0xff, 0xff, 0x22 > + # LOE rbx r15 r12d r13d > + > +/* Range mask > + * bits check > + */ > + > +L(RANGEMASK_CHECK): > + btl %r12d, %r13d > + > +/* Call scalar math function */ > + jc L(SCALAR_MATH_CALL) > + # LOE rbx r15 r12d r13d > + > +/* Special inputs > + * processing loop > + */ > + > +L(SPECIAL_VALUES_LOOP): > + incl %r12d > + cmpl $8, %r12d > + > +/* Check bits in range mask */ > + jl L(RANGEMASK_CHECK) > + # LOE rbx r15 r12d r13d > + > + movq 16(%rsp), %r12 > + cfi_restore(12) > + movq 8(%rsp), %r13 > + cfi_restore(13) > + movq (%rsp), %r14 > + cfi_restore(14) > + vmovups 128(%rsp), %zmm0 > + > +/* Go to exit */ > + jmp L(EXIT) > + /* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -176; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x50, 0xff, 0xff, 0xff, 0x22 > + /* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -184; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22 > + /* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -192; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 0xff, 0xff, 0xff, 0x22 > + # LOE rbx r12 r13 r14 r15 zmm0 > + > +/* Scalar math fucntion call > + * to process special input > + */ > + > +L(SCALAR_MATH_CALL): > + movl %r12d, %r14d > + movsd 64(%rsp,%r14,8), %xmm0 > + call asin@PLT > + # LOE rbx r14 r15 r12d r13d xmm0 > + > + movsd %xmm0, 128(%rsp,%r14,8) > + > +/* Process special inputs in loop */ > + jmp L(SPECIAL_VALUES_LOOP) > + # LOE rbx r15 r12d r13d > +END(_ZGVeN8v_asin_skx) > + > + .section .rodata, "a" > + .align 64 > + > +#ifdef __svml_dasin_data_internal_typedef > +typedef unsigned int VUINT32; > +typedef struct { > + __declspec(align(64)) VUINT32 AbsMask[8][2]; > + __declspec(align(64)) VUINT32 OneHalf[8][2]; > + __declspec(align(64)) VUINT32 SmallNorm[8][2]; > + __declspec(align(64)) VUINT32 One[8][2]; > + __declspec(align(64)) VUINT32 Two[8][2]; > + __declspec(align(64)) VUINT32 sqrt_coeff[4][8][2]; > + __declspec(align(64)) VUINT32 poly_coeff[12][8][2]; > + __declspec(align(64)) VUINT32 Pi2H[8][2]; > +} __svml_dasin_data_internal; > +#endif > +__svml_dasin_data_internal: > + /*== AbsMask ==*/ > + .quad 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff, 0x7fffffffffffffff > + /*== OneHalf ==*/ > + .align 64 > + .quad 0x3fe0000000000000, 0x3fe0000000000000, 0x3fe0000000000000, 0x3fe0000000000000, 0x3fe0000000000000, 0x3fe0000000000000, 0x3fe0000000000000, 0x3fe0000000000000 > + /*== SmallNorm ==*/ > + .align 64 > + .quad 0x3000000000000000, 0x3000000000000000, 0x3000000000000000, 0x3000000000000000, 0x3000000000000000, 0x3000000000000000, 0x3000000000000000, 0x3000000000000000 > + /*== One ==*/ > + .align 64 > + .quad 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000 > + /*== Two ==*/ > + .align 64 > + .quad 0x4000000000000000, 0x4000000000000000, 0x4000000000000000, 0x4000000000000000, 0x4000000000000000, 0x4000000000000000, 0x4000000000000000, 0x4000000000000000 > + /*== sqrt_coeff[4] ==*/ > + .align 64 > + .quad 0xbf918000993B24C3, 0xbf918000993B24C3, 0xbf918000993B24C3, 0xbf918000993B24C3, 0xbf918000993B24C3, 0xbf918000993B24C3, 0xbf918000993B24C3, 0xbf918000993B24C3 /* sqrt_coeff4 */ > + .quad 0x3fa400006F70D42D, 0x3fa400006F70D42D, 0x3fa400006F70D42D, 0x3fa400006F70D42D, 0x3fa400006F70D42D, 0x3fa400006F70D42D, 0x3fa400006F70D42D, 0x3fa400006F70D42D /* sqrt_coeff3 */ > + .quad 0xbfb7FFFFFFFFFE97, 0xbfb7FFFFFFFFFE97, 0xbfb7FFFFFFFFFE97, 0xbfb7FFFFFFFFFE97, 0xbfb7FFFFFFFFFE97, 0xbfb7FFFFFFFFFE97, 0xbfb7FFFFFFFFFE97, 0xbfb7FFFFFFFFFE97 /* sqrt_coeff2 */ > + .quad 0x3fcFFFFFFFFFFF9D, 0x3fcFFFFFFFFFFF9D, 0x3fcFFFFFFFFFFF9D, 0x3fcFFFFFFFFFFF9D, 0x3fcFFFFFFFFFFF9D, 0x3fcFFFFFFFFFFF9D, 0x3fcFFFFFFFFFFF9D, 0x3fcFFFFFFFFFFF9D /* sqrt_coeff1 */ > + /*== poly_coeff[12] ==*/ > + .align 64 > + .quad 0x3fa07520C70EB909, 0x3fa07520C70EB909, 0x3fa07520C70EB909, 0x3fa07520C70EB909, 0x3fa07520C70EB909, 0x3fa07520C70EB909, 0x3fa07520C70EB909, 0x3fa07520C70EB909 /* poly_coeff12 */ > + .quad 0xbf90FB17F7DBB0ED, 0xbf90FB17F7DBB0ED, 0xbf90FB17F7DBB0ED, 0xbf90FB17F7DBB0ED, 0xbf90FB17F7DBB0ED, 0xbf90FB17F7DBB0ED, 0xbf90FB17F7DBB0ED, 0xbf90FB17F7DBB0ED /* poly_coeff11 */ > + .quad 0x3f943F44BFBC3BAE, 0x3f943F44BFBC3BAE, 0x3f943F44BFBC3BAE, 0x3f943F44BFBC3BAE, 0x3f943F44BFBC3BAE, 0x3f943F44BFBC3BAE, 0x3f943F44BFBC3BAE, 0x3f943F44BFBC3BAE /* poly_coeff10 */ > + .quad 0x3f7A583395D45ED5, 0x3f7A583395D45ED5, 0x3f7A583395D45ED5, 0x3f7A583395D45ED5, 0x3f7A583395D45ED5, 0x3f7A583395D45ED5, 0x3f7A583395D45ED5, 0x3f7A583395D45ED5 /* poly_coeff9 */ > + .quad 0x3f88F8DC2AFCCAD6, 0x3f88F8DC2AFCCAD6, 0x3f88F8DC2AFCCAD6, 0x3f88F8DC2AFCCAD6, 0x3f88F8DC2AFCCAD6, 0x3f88F8DC2AFCCAD6, 0x3f88F8DC2AFCCAD6, 0x3f88F8DC2AFCCAD6 /* poly_coeff8 */ > + .quad 0x3f8C6DBBCB88BD57, 0x3f8C6DBBCB88BD57, 0x3f8C6DBBCB88BD57, 0x3f8C6DBBCB88BD57, 0x3f8C6DBBCB88BD57, 0x3f8C6DBBCB88BD57, 0x3f8C6DBBCB88BD57, 0x3f8C6DBBCB88BD57 /* poly_coeff7 */ > + .quad 0x3f91C6DCF538AD2E, 0x3f91C6DCF538AD2E, 0x3f91C6DCF538AD2E, 0x3f91C6DCF538AD2E, 0x3f91C6DCF538AD2E, 0x3f91C6DCF538AD2E, 0x3f91C6DCF538AD2E, 0x3f91C6DCF538AD2E /* poly_coeff6 */ > + .quad 0x3f96E89CEBDEFadd, 0x3f96E89CEBDEFadd, 0x3f96E89CEBDEFadd, 0x3f96E89CEBDEFadd, 0x3f96E89CEBDEFadd, 0x3f96E89CEBDEFadd, 0x3f96E89CEBDEFadd, 0x3f96E89CEBDEFadd /* poly_coeff5 */ > + .quad 0x3f9F1C72E13AD8BE, 0x3f9F1C72E13AD8BE, 0x3f9F1C72E13AD8BE, 0x3f9F1C72E13AD8BE, 0x3f9F1C72E13AD8BE, 0x3f9F1C72E13AD8BE, 0x3f9F1C72E13AD8BE, 0x3f9F1C72E13AD8BE /* poly_coeff4 */ > + .quad 0x3fa6DB6DB3B445F8, 0x3fa6DB6DB3B445F8, 0x3fa6DB6DB3B445F8, 0x3fa6DB6DB3B445F8, 0x3fa6DB6DB3B445F8, 0x3fa6DB6DB3B445F8, 0x3fa6DB6DB3B445F8, 0x3fa6DB6DB3B445F8 /* poly_coeff3 */ > + .quad 0x3fb333333337E0DE, 0x3fb333333337E0DE, 0x3fb333333337E0DE, 0x3fb333333337E0DE, 0x3fb333333337E0DE, 0x3fb333333337E0DE, 0x3fb333333337E0DE, 0x3fb333333337E0DE /* poly_coeff2 */ > + .quad 0x3fc555555555529C, 0x3fc555555555529C, 0x3fc555555555529C, 0x3fc555555555529C, 0x3fc555555555529C, 0x3fc555555555529C, 0x3fc555555555529C, 0x3fc555555555529C /* poly_coeff1 */ > + /*== Pi2H ==*/ > + .align 64 > + .quad 0x3ff921fb54442d18, 0x3ff921fb54442d18, 0x3ff921fb54442d18, 0x3ff921fb54442d18, 0x3ff921fb54442d18, 0x3ff921fb54442d18, 0x3ff921fb54442d18, 0x3ff921fb54442d18 > + .align 64 > + .type __svml_dasin_data_internal,@object > + .size __svml_dasin_data_internal,.-__svml_dasin_data_internal > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core-avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core-avx2.S > new file mode 100644 > index 0000000000..e0582f27d4 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core-avx2.S > @@ -0,0 +1,20 @@ > +/* AVX2 version of vectorized asinf. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define _ZGVeN16v_asinf _ZGVeN16v_asinf_avx2_wrapper > +#include "../svml_s_asinf16_core.S" > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core.c b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core.c > new file mode 100644 > index 0000000000..4435055566 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core.c > @@ -0,0 +1,28 @@ > +/* Multiple versions of vectorized asinf, vector length is 16. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define SYMBOL_NAME _ZGVeN16v_asinf > +#include "ifunc-mathvec-avx512-skx.h" > + > +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); > + > +#ifdef SHARED > +__hidden_ver1 (_ZGVeN16v_asinf, __GI__ZGVeN16v_asinf, > + __redirect__ZGVeN16v_asinf) > + __attribute__ ((visibility ("hidden"))); > +#endif > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core_avx512.S b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core_avx512.S > new file mode 100644 > index 0000000000..7afdfd1317 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf16_core_avx512.S > @@ -0,0 +1,260 @@ > +/* Function asinf vectorized with AVX-512. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + https://www.gnu.org/licenses/. */ > + > +/* > + * ALGORITHM DESCRIPTION: > + * > + * SelMask = (|x| >= 0.5) ? 1 : 0; > + * R = SelMask ? sqrt(0.5 - 0.5*|x|) : |x| > + * asin(x) = (SelMask ? (Pi/2 - 2*Poly(R)) : Poly(R))*(-1)^sign(x) > + * > + * > + */ > + > +/* Offsets for data table __svml_sasin_data_internal > + */ > +#define AbsMask 0 > +#define OneHalf 64 > +#define SmallNorm 128 > +#define One 192 > +#define Two 256 > +#define sqrt_coeff_1 320 > +#define sqrt_coeff_2 384 > +#define poly_coeff_1 448 > +#define poly_coeff_2 512 > +#define poly_coeff_3 576 > +#define poly_coeff_4 640 > +#define poly_coeff_5 704 > +#define Pi2H 768 > + > +#include > + > + .text > + .section .text.exex512,"ax",@progbits > +ENTRY(_ZGVeN16v_asinf_skx) > + pushq %rbp > + cfi_def_cfa_offset(16) > + movq %rsp, %rbp > + cfi_def_cfa(6, 16) > + cfi_offset(6, -16) > + andq $-64, %rsp > + subq $192, %rsp > + vmovups __svml_sasin_data_internal(%rip), %zmm4 > + vmovups OneHalf+__svml_sasin_data_internal(%rip), %zmm6 > + > +/* SQ ~ -2*sqrt(Y) */ > + vmovups SmallNorm+__svml_sasin_data_internal(%rip), %zmm8 > + vmovups Two+__svml_sasin_data_internal(%rip), %zmm12 > + vmovups sqrt_coeff_1+__svml_sasin_data_internal(%rip), %zmm13 > + vmovups One+__svml_sasin_data_internal(%rip), %zmm7 > + vmovaps %zmm0, %zmm3 > + > +/* x = |arg| */ > + vandps %zmm3, %zmm4, %zmm2 > + vandnps %zmm3, %zmm4, %zmm1 > + > +/* x^2 */ > + vmulps {rn-sae}, %zmm2, %zmm2, %zmm5 > + vcmpps $17, {sae}, %zmm2, %zmm7, %k0 > + vcmpps $21, {sae}, %zmm6, %zmm2, %k2 > + vmovups poly_coeff_2+__svml_sasin_data_internal(%rip), %zmm7 > + kmovw %k0, %edx > + > +/* Y = 0.5 - 0.5*x */ > + vmovaps %zmm6, %zmm9 > + vfnmadd231ps {rn-sae}, %zmm2, %zmm6, %zmm9 > + vmovups poly_coeff_5+__svml_sasin_data_internal(%rip), %zmm6 > + vrsqrt14ps %zmm9, %zmm10 > + vcmpps $17, {sae}, %zmm8, %zmm9, %k1 > + vminps {sae}, %zmm9, %zmm5, %zmm0 > + vmovups sqrt_coeff_2+__svml_sasin_data_internal(%rip), %zmm8 > + vmovups poly_coeff_4+__svml_sasin_data_internal(%rip), %zmm5 > + vxorps %zmm10, %zmm10, %zmm10{%k1} > + vaddps {rn-sae}, %zmm9, %zmm9, %zmm14 > + vmulps {rn-sae}, %zmm10, %zmm10, %zmm11 > + vmulps {rn-sae}, %zmm10, %zmm14, %zmm4 > + vfmsub213ps {rn-sae}, %zmm12, %zmm11, %zmm14 > + vmulps {rn-sae}, %zmm14, %zmm4, %zmm15 > + vfmadd231ps {rn-sae}, %zmm14, %zmm13, %zmm8 > + vmovups poly_coeff_3+__svml_sasin_data_internal(%rip), %zmm14 > + > +/* polynomial */ > + vmovups poly_coeff_1+__svml_sasin_data_internal(%rip), %zmm13 > + vfmsub213ps {rn-sae}, %zmm4, %zmm15, %zmm8 > + vfmadd231ps {rn-sae}, %zmm0, %zmm14, %zmm5 > + vfmadd231ps {rn-sae}, %zmm0, %zmm13, %zmm7 > + vmulps {rn-sae}, %zmm0, %zmm0, %zmm15 > + vblendmps %zmm8, %zmm2, %zmm2{%k2} > + vfmadd213ps {rn-sae}, %zmm5, %zmm15, %zmm7 > + vfmadd213ps {rn-sae}, %zmm6, %zmm0, %zmm7 > + vmulps {rn-sae}, %zmm0, %zmm7, %zmm9 > + vmovups Pi2H+__svml_sasin_data_internal(%rip), %zmm0 > + vfmadd213ps {rn-sae}, %zmm2, %zmm2, %zmm9 > + vaddps {rn-sae}, %zmm0, %zmm9, %zmm9{%k2} > + vxorps %zmm1, %zmm9, %zmm0 > + testl %edx, %edx > + > +/* Go to special inputs processing branch */ > + jne L(SPECIAL_VALUES_BRANCH) > + # LOE rbx r12 r13 r14 r15 edx zmm0 zmm3 > + > +/* Restore registers > + * and exit the function > + */ > + > +L(EXIT): > + movq %rbp, %rsp > + popq %rbp > + cfi_def_cfa(7, 8) > + cfi_restore(6) > + ret > + cfi_def_cfa(6, 16) > + cfi_offset(6, -16) > + > +/* Branch to process > + * special inputs > + */ > + > +L(SPECIAL_VALUES_BRANCH): > + vmovups %zmm3, 64(%rsp) > + vmovups %zmm0, 128(%rsp) > + # LOE rbx r12 r13 r14 r15 edx zmm0 > + > + xorl %eax, %eax > + # LOE rbx r12 r13 r14 r15 eax edx > + > + vzeroupper > + movq %r12, 16(%rsp) > + /* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -176; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x50, 0xff, 0xff, 0xff, 0x22 > + movl %eax, %r12d > + movq %r13, 8(%rsp) > + /* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -184; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22 > + movl %edx, %r13d > + movq %r14, (%rsp) > + /* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -192; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 0xff, 0xff, 0xff, 0x22 > + # LOE rbx r15 r12d r13d > + > +/* Range mask > + * bits check > + */ > + > +L(RANGEMASK_CHECK): > + btl %r12d, %r13d > + > +/* Call scalar math function */ > + jc L(SCALAR_MATH_CALL) > + # LOE rbx r15 r12d r13d > + > +/* Special inputs > + * processing loop > + */ > + > +L(SPECIAL_VALUES_LOOP): > + incl %r12d > + cmpl $16, %r12d > + > +/* Check bits in range mask */ > + jl L(RANGEMASK_CHECK) > + # LOE rbx r15 r12d r13d > + > + movq 16(%rsp), %r12 > + cfi_restore(12) > + movq 8(%rsp), %r13 > + cfi_restore(13) > + movq (%rsp), %r14 > + cfi_restore(14) > + vmovups 128(%rsp), %zmm0 > + > +/* Go to exit */ > + jmp L(EXIT) > + /* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -176; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x50, 0xff, 0xff, 0xff, 0x22 > + /* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -184; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x48, 0xff, 0xff, 0xff, 0x22 > + /* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -64; DW_OP_and; DW_OP_const4s: -192; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xc0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0x40, 0xff, 0xff, 0xff, 0x22 > + # LOE rbx r12 r13 r14 r15 zmm0 > + > +/* Scalar math fucntion call > + * to process special input > + */ > + > +L(SCALAR_MATH_CALL): > + movl %r12d, %r14d > + movss 64(%rsp,%r14,4), %xmm0 > + call asinf@PLT > + # LOE rbx r14 r15 r12d r13d xmm0 > + > + movss %xmm0, 128(%rsp,%r14,4) > + > +/* Process special inputs in loop */ > + jmp L(SPECIAL_VALUES_LOOP) > + # LOE rbx r15 r12d r13d > +END(_ZGVeN16v_asinf_skx) > + > + .section .rodata, "a" > + .align 64 > + > +#ifdef __svml_sasin_data_internal_typedef > +typedef unsigned int VUINT32; > +typedef struct { > + __declspec(align(64)) VUINT32 AbsMask[16][1]; > + __declspec(align(64)) VUINT32 OneHalf[16][1]; > + __declspec(align(64)) VUINT32 SmallNorm[16][1]; > + __declspec(align(64)) VUINT32 One[16][1]; > + __declspec(align(64)) VUINT32 Two[16][1]; > + __declspec(align(64)) VUINT32 sqrt_coeff[2][16][1]; > + __declspec(align(64)) VUINT32 poly_coeff[5][16][1]; > + __declspec(align(64)) VUINT32 Pi2H[16][1]; > +} __svml_sasin_data_internal; > +#endif > +__svml_sasin_data_internal: > + /*== AbsMask ==*/ > + .long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff > + /*== OneHalf ==*/ > + .align 64 > + .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 > + /*== SmallNorm ==*/ > + .align 64 > + .long 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000 > + /*== One ==*/ > + .align 64 > + .long 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 > + /*== Two ==*/ > + .align 64 > + .long 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000 > + /*== sqrt_coeff[2] ==*/ > + .align 64 > + .long 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004 /* sqrt_coeff2 */ > + .long 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001 /* sqrt_coeff1 */ > + /*== poly_coeff[5] ==*/ > + .align 64 > + .long 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07 /* poly_coeff5 */ > + .long 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B /* poly_coeff4 */ > + .long 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4 /* poly_coeff3 */ > + .long 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12 /* poly_coeff2 */ > + .long 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF /* poly_coeff1 */ > + /*== Pi2H ==*/ > + .align 64 > + .long 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB > + .align 64 > + .type __svml_sasin_data_internal,@object > + .size __svml_sasin_data_internal,.-__svml_sasin_data_internal > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core-sse2.S b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core-sse2.S > new file mode 100644 > index 0000000000..b958db7795 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core-sse2.S > @@ -0,0 +1,20 @@ > +/* SSE2 version of vectorized asinf, vector length is 4. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define _ZGVbN4v_asinf _ZGVbN4v_asinf_sse2 > +#include "../svml_s_asinf4_core.S" > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core.c b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core.c > new file mode 100644 > index 0000000000..5a7aa94264 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core.c > @@ -0,0 +1,28 @@ > +/* Multiple versions of vectorized asinf, vector length is 4. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define SYMBOL_NAME _ZGVbN4v_asinf > +#include "ifunc-mathvec-sse4_1.h" > + > +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); > + > +#ifdef SHARED > +__hidden_ver1 (_ZGVbN4v_asinf, __GI__ZGVbN4v_asinf, > + __redirect__ZGVbN4v_asinf) > + __attribute__ ((visibility ("hidden"))); > +#endif > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core_sse4.S b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core_sse4.S > new file mode 100644 > index 0000000000..ddcceeb7b9 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf4_core_sse4.S > @@ -0,0 +1,252 @@ > +/* Function asinf vectorized with SSE4. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + https://www.gnu.org/licenses/. */ > + > +/* > + * ALGORITHM DESCRIPTION: > + * > + * SelMask = (|x| >= 0.5) ? 1 : 0; > + * R = SelMask ? sqrt(0.5 - 0.5*|x|) : |x| > + * asin(x) = (SelMask ? (Pi/2 - 2*Poly(R)) : Poly(R))*(-1)^sign(x) > + * > + * > + */ > + > +/* Offsets for data table __svml_sasin_data_internal > + */ > +#define AbsMask 0 > +#define OneHalf 16 > +#define SmallNorm 32 > +#define One 48 > +#define Two 64 > +#define sqrt_coeff 80 > +#define poly_coeff 112 > +#define Pi2H 192 > + > +#include > + > + .text > + .section .text.sse4,"ax",@progbits > +ENTRY(_ZGVbN4v_asinf_sse4) > + subq $72, %rsp > + cfi_def_cfa_offset(80) > + movaps %xmm0, %xmm2 > + movups __svml_sasin_data_internal(%rip), %xmm1 > + movups OneHalf+__svml_sasin_data_internal(%rip), %xmm5 > + > +/* x = |arg| */ > + movaps %xmm1, %xmm0 > + andps %xmm2, %xmm0 > + > +/* Y = 0.5 - 0.5*x */ > + movaps %xmm5, %xmm3 > + mulps %xmm0, %xmm3 > + movaps %xmm5, %xmm8 > + > +/* x^2 */ > + movaps %xmm0, %xmm14 > + movaps %xmm0, %xmm15 > + mulps %xmm0, %xmm14 > + subps %xmm3, %xmm8 > + cmpnltps %xmm5, %xmm15 > + > +/* SQ ~ -2*sqrt(Y) */ > + rsqrtps %xmm8, %xmm6 > + minps %xmm8, %xmm14 > + movaps %xmm8, %xmm9 > + movaps %xmm14, %xmm10 > + cmpltps SmallNorm+__svml_sasin_data_internal(%rip), %xmm9 > + mulps %xmm14, %xmm10 > + addps %xmm8, %xmm8 > + andnps %xmm6, %xmm9 > + movaps %xmm15, %xmm3 > + movaps %xmm9, %xmm7 > + andnps %xmm0, %xmm3 > + mulps %xmm9, %xmm7 > + andnps %xmm2, %xmm1 > + mulps %xmm8, %xmm9 > + mulps %xmm7, %xmm8 > + > +/* polynomial */ > + movups poly_coeff+__svml_sasin_data_internal(%rip), %xmm11 > + mulps %xmm14, %xmm11 > + subps Two+__svml_sasin_data_internal(%rip), %xmm8 > + movups poly_coeff+32+__svml_sasin_data_internal(%rip), %xmm12 > + mulps %xmm14, %xmm12 > + addps poly_coeff+16+__svml_sasin_data_internal(%rip), %xmm11 > + mulps %xmm10, %xmm11 > + addps poly_coeff+48+__svml_sasin_data_internal(%rip), %xmm12 > + movups sqrt_coeff+__svml_sasin_data_internal(%rip), %xmm13 > + addps %xmm11, %xmm12 > + mulps %xmm8, %xmm13 > + mulps %xmm9, %xmm8 > + mulps %xmm14, %xmm12 > + addps sqrt_coeff+16+__svml_sasin_data_internal(%rip), %xmm13 > + addps poly_coeff+64+__svml_sasin_data_internal(%rip), %xmm12 > + mulps %xmm8, %xmm13 > + mulps %xmm12, %xmm14 > + subps %xmm9, %xmm13 > + andps %xmm15, %xmm13 > + orps %xmm13, %xmm3 > + mulps %xmm3, %xmm14 > + movups One+__svml_sasin_data_internal(%rip), %xmm4 > + addps %xmm14, %xmm3 > + cmpltps %xmm0, %xmm4 > + movups Pi2H+__svml_sasin_data_internal(%rip), %xmm0 > + andps %xmm15, %xmm0 > + movmskps %xmm4, %edx > + addps %xmm3, %xmm0 > + pxor %xmm1, %xmm0 > + testl %edx, %edx > + > +/* Go to special inputs processing branch */ > + jne L(SPECIAL_VALUES_BRANCH) > + # LOE rbx rbp r12 r13 r14 r15 edx xmm0 xmm2 > + > +/* Restore registers > + * and exit the function > + */ > + > +L(EXIT): > + addq $72, %rsp > + cfi_def_cfa_offset(8) > + ret > + cfi_def_cfa_offset(80) > + > +/* Branch to process > + * special inputs > + */ > + > +L(SPECIAL_VALUES_BRANCH): > + movups %xmm2, 32(%rsp) > + movups %xmm0, 48(%rsp) > + # LOE rbx rbp r12 r13 r14 r15 edx > + > + xorl %eax, %eax > + movq %r12, 16(%rsp) > + cfi_offset(12, -64) > + movl %eax, %r12d > + movq %r13, 8(%rsp) > + cfi_offset(13, -72) > + movl %edx, %r13d > + movq %r14, (%rsp) > + cfi_offset(14, -80) > + # LOE rbx rbp r15 r12d r13d > + > +/* Range mask > + * bits check > + */ > + > +L(RANGEMASK_CHECK): > + btl %r12d, %r13d > + > +/* Call scalar math function */ > + jc L(SCALAR_MATH_CALL) > + # LOE rbx rbp r15 r12d r13d > + > +/* Special inputs > + * processing loop > + */ > + > +L(SPECIAL_VALUES_LOOP): > + incl %r12d > + cmpl $4, %r12d > + > +/* Check bits in range mask */ > + jl L(RANGEMASK_CHECK) > + # LOE rbx rbp r15 r12d r13d > + > + movq 16(%rsp), %r12 > + cfi_restore(12) > + movq 8(%rsp), %r13 > + cfi_restore(13) > + movq (%rsp), %r14 > + cfi_restore(14) > + movups 48(%rsp), %xmm0 > + > +/* Go to exit */ > + jmp L(EXIT) > + cfi_offset(12, -64) > + cfi_offset(13, -72) > + cfi_offset(14, -80) > + # LOE rbx rbp r12 r13 r14 r15 xmm0 > + > +/* Scalar math fucntion call > + * to process special input > + */ > + > +L(SCALAR_MATH_CALL): > + movl %r12d, %r14d > + movss 32(%rsp,%r14,4), %xmm0 > + call asinf@PLT > + # LOE rbx rbp r14 r15 r12d r13d xmm0 > + > + movss %xmm0, 48(%rsp,%r14,4) > + > +/* Process special inputs in loop */ > + jmp L(SPECIAL_VALUES_LOOP) > + # LOE rbx rbp r15 r12d r13d > +END(_ZGVbN4v_asinf_sse4) > + > + .section .rodata, "a" > + .align 16 > + > +#ifdef __svml_sasin_data_internal_typedef > +typedef unsigned int VUINT32; > +typedef struct { > + __declspec(align(16)) VUINT32 AbsMask[4][1]; > + __declspec(align(16)) VUINT32 OneHalf[4][1]; > + __declspec(align(16)) VUINT32 SmallNorm[4][1]; > + __declspec(align(16)) VUINT32 One[4][1]; > + __declspec(align(16)) VUINT32 Two[4][1]; > + __declspec(align(16)) VUINT32 sqrt_coeff[2][4][1]; > + __declspec(align(16)) VUINT32 poly_coeff[5][4][1]; > + __declspec(align(16)) VUINT32 Pi2H[4][1]; > +} __svml_sasin_data_internal; > +#endif > +__svml_sasin_data_internal: > + /*== AbsMask ==*/ > + .long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff > + /*== OneHalf ==*/ > + .align 16 > + .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 > + /*== SmallNorm ==*/ > + .align 16 > + .long 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000 > + /*== One ==*/ > + .align 16 > + .long 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 > + /*== Two ==*/ > + .align 16 > + .long 0x40000000, 0x40000000, 0x40000000, 0x40000000 > + /*== sqrt_coeff[2] ==*/ > + .align 16 > + .long 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004 /* sqrt_coeff2 */ > + .long 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001 /* sqrt_coeff1 */ > + /*== poly_coeff[5] ==*/ > + .align 16 > + .long 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07 /* poly_coeff5 */ > + .long 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B /* poly_coeff4 */ > + .long 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4 /* poly_coeff3 */ > + .long 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12 /* poly_coeff2 */ > + .long 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF /* poly_coeff1 */ > + /*== Pi2H ==*/ > + .align 16 > + .long 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB > + .align 16 > + .type __svml_sasin_data_internal,@object > + .size __svml_sasin_data_internal,.-__svml_sasin_data_internal > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core-sse.S b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core-sse.S > new file mode 100644 > index 0000000000..6273c919d6 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core-sse.S > @@ -0,0 +1,20 @@ > +/* SSE version of vectorized asinf, vector length is 8. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define _ZGVdN8v_asinf _ZGVdN8v_asinf_sse_wrapper > +#include "../svml_s_asinf8_core.S" > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core.c b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core.c > new file mode 100644 > index 0000000000..946b25b43f > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core.c > @@ -0,0 +1,28 @@ > +/* Multiple versions of vectorized asinf, vector length is 8. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#define SYMBOL_NAME _ZGVdN8v_asinf > +#include "ifunc-mathvec-avx2.h" > + > +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); > + > +#ifdef SHARED > +__hidden_ver1 (_ZGVdN8v_asinf, __GI__ZGVdN8v_asinf, > + __redirect__ZGVdN8v_asinf) > + __attribute__ ((visibility ("hidden"))); > +#endif > diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core_avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core_avx2.S > new file mode 100644 > index 0000000000..89c156dbbb > --- /dev/null > +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_asinf8_core_avx2.S > @@ -0,0 +1,249 @@ > +/* Function asinf vectorized with AVX2. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + https://www.gnu.org/licenses/. */ > + > +/* > + * ALGORITHM DESCRIPTION: > + * > + * SelMask = (|x| >= 0.5) ? 1 : 0; > + * R = SelMask ? sqrt(0.5 - 0.5*|x|) : |x| > + * asin(x) = (SelMask ? (Pi/2 - 2*Poly(R)) : Poly(R))*(-1)^sign(x) > + * > + * > + */ > + > +/* Offsets for data table __svml_sasin_data_internal > + */ > +#define AbsMask 0 > +#define OneHalf 32 > +#define SmallNorm 64 > +#define One 96 > +#define Two 128 > +#define sqrt_coeff 160 > +#define poly_coeff 224 > +#define Pi2H 384 > + > +#include > + > + .text > + .section .text.avx2,"ax",@progbits > +ENTRY(_ZGVdN8v_asinf_avx2) > + pushq %rbp > + cfi_def_cfa_offset(16) > + movq %rsp, %rbp > + cfi_def_cfa(6, 16) > + cfi_offset(6, -16) > + andq $-32, %rsp > + subq $96, %rsp > + vmovups __svml_sasin_data_internal(%rip), %ymm5 > + vmovups OneHalf+__svml_sasin_data_internal(%rip), %ymm9 > + vmovups One+__svml_sasin_data_internal(%rip), %ymm6 > + vmovaps %ymm0, %ymm4 > + > +/* x = |arg| */ > + vandps %ymm4, %ymm5, %ymm3 > + > +/* Y = 0.5 - 0.5*x */ > + vmovaps %ymm9, %ymm12 > + vfnmadd231ps %ymm3, %ymm9, %ymm12 > + > +/* x^2 */ > + vmulps %ymm3, %ymm3, %ymm7 > + vcmplt_oqps %ymm3, %ymm6, %ymm8 > + > +/* SQ ~ -2*sqrt(Y) */ > + vcmplt_oqps SmallNorm+__svml_sasin_data_internal(%rip), %ymm12, %ymm10 > + vminps %ymm12, %ymm7, %ymm1 > + vaddps %ymm12, %ymm12, %ymm15 > + vcmpnlt_uqps %ymm9, %ymm3, %ymm0 > + vrsqrtps %ymm12, %ymm11 > + vmovups poly_coeff+64+__svml_sasin_data_internal(%rip), %ymm7 > + vmulps %ymm1, %ymm1, %ymm6 > + vmovups sqrt_coeff+__svml_sasin_data_internal(%rip), %ymm9 > + vfmadd213ps poly_coeff+96+__svml_sasin_data_internal(%rip), %ymm1, %ymm7 > + vmovmskps %ymm8, %edx > + > +/* polynomial */ > + vmovups poly_coeff+__svml_sasin_data_internal(%rip), %ymm8 > + vandnps %ymm11, %ymm10, %ymm13 > + vmulps %ymm13, %ymm13, %ymm14 > + vfmadd213ps poly_coeff+32+__svml_sasin_data_internal(%rip), %ymm1, %ymm8 > + vandnps %ymm4, %ymm5, %ymm2 > + vmulps %ymm15, %ymm13, %ymm5 > + vfmsub213ps Two+__svml_sasin_data_internal(%rip), %ymm14, %ymm15 > + vfmadd213ps %ymm7, %ymm6, %ymm8 > + vfmadd213ps sqrt_coeff+32+__svml_sasin_data_internal(%rip), %ymm15, %ymm9 > + vmulps %ymm15, %ymm5, %ymm15 > + vfmadd213ps poly_coeff+128+__svml_sasin_data_internal(%rip), %ymm1, %ymm8 > + vfmsub213ps %ymm5, %ymm15, %ymm9 > + vmulps %ymm8, %ymm1, %ymm1 > + vblendvps %ymm0, %ymm9, %ymm3, %ymm3 > + vfmadd213ps %ymm3, %ymm3, %ymm1 > + vandps Pi2H+__svml_sasin_data_internal(%rip), %ymm0, %ymm0 > + vaddps %ymm1, %ymm0, %ymm10 > + vxorps %ymm2, %ymm10, %ymm0 > + testl %edx, %edx > + > +/* Go to special inputs processing branch */ > + jne L(SPECIAL_VALUES_BRANCH) > + # LOE rbx r12 r13 r14 r15 edx ymm0 ymm4 > + > +/* Restore registers > + * and exit the function > + */ > + > +L(EXIT): > + movq %rbp, %rsp > + popq %rbp > + cfi_def_cfa(7, 8) > + cfi_restore(6) > + ret > + cfi_def_cfa(6, 16) > + cfi_offset(6, -16) > + > +/* Branch to process > + * special inputs > + */ > + > +L(SPECIAL_VALUES_BRANCH): > + vmovups %ymm4, 32(%rsp) > + vmovups %ymm0, 64(%rsp) > + # LOE rbx r12 r13 r14 r15 edx ymm0 > + > + xorl %eax, %eax > + # LOE rbx r12 r13 r14 r15 eax edx > + > + vzeroupper > + movq %r12, 16(%rsp) > + /* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -80; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xb0, 0xff, 0xff, 0xff, 0x22 > + movl %eax, %r12d > + movq %r13, 8(%rsp) > + /* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -88; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa8, 0xff, 0xff, 0xff, 0x22 > + movl %edx, %r13d > + movq %r14, (%rsp) > + /* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -96; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa0, 0xff, 0xff, 0xff, 0x22 > + # LOE rbx r15 r12d r13d > + > +/* Range mask > + * bits check > + */ > + > +L(RANGEMASK_CHECK): > + btl %r12d, %r13d > + > +/* Call scalar math function */ > + jc L(SCALAR_MATH_CALL) > + # LOE rbx r15 r12d r13d > + > +/* Special inputs > + * processing loop > + */ > + > +L(SPECIAL_VALUES_LOOP): > + incl %r12d > + cmpl $8, %r12d > + > +/* Check bits in range mask */ > + jl L(RANGEMASK_CHECK) > + # LOE rbx r15 r12d r13d > + > + movq 16(%rsp), %r12 > + cfi_restore(12) > + movq 8(%rsp), %r13 > + cfi_restore(13) > + movq (%rsp), %r14 > + cfi_restore(14) > + vmovups 64(%rsp), %ymm0 > + > +/* Go to exit */ > + jmp L(EXIT) > + /* DW_CFA_expression: r12 (r12) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -80; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0c, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xb0, 0xff, 0xff, 0xff, 0x22 > + /* DW_CFA_expression: r13 (r13) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -88; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0d, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa8, 0xff, 0xff, 0xff, 0x22 > + /* DW_CFA_expression: r14 (r14) (DW_OP_lit8; DW_OP_minus; DW_OP_const4s: -32; DW_OP_and; DW_OP_const4s: -96; DW_OP_plus) */ > + .cfi_escape 0x10, 0x0e, 0x0e, 0x38, 0x1c, 0x0d, 0xe0, 0xff, 0xff, 0xff, 0x1a, 0x0d, 0xa0, 0xff, 0xff, 0xff, 0x22 > + # LOE rbx r12 r13 r14 r15 ymm0 > + > +/* Scalar math fucntion call > + * to process special input > + */ > + > +L(SCALAR_MATH_CALL): > + movl %r12d, %r14d > + movss 32(%rsp,%r14,4), %xmm0 > + call asinf@PLT > + # LOE rbx r14 r15 r12d r13d xmm0 > + > + movss %xmm0, 64(%rsp,%r14,4) > + > +/* Process special inputs in loop */ > + jmp L(SPECIAL_VALUES_LOOP) > + # LOE rbx r15 r12d r13d > +END(_ZGVdN8v_asinf_avx2) > + > + .section .rodata, "a" > + .align 32 > + > +#ifdef __svml_sasin_data_internal_typedef > +typedef unsigned int VUINT32; > +typedef struct { > + __declspec(align(32)) VUINT32 AbsMask[8][1]; > + __declspec(align(32)) VUINT32 OneHalf[8][1]; > + __declspec(align(32)) VUINT32 SmallNorm[8][1]; > + __declspec(align(32)) VUINT32 One[8][1]; > + __declspec(align(32)) VUINT32 Two[8][1]; > + __declspec(align(32)) VUINT32 sqrt_coeff[2][8][1]; > + __declspec(align(32)) VUINT32 poly_coeff[5][8][1]; > + __declspec(align(32)) VUINT32 Pi2H[8][1]; > +} __svml_sasin_data_internal; > +#endif > +__svml_sasin_data_internal: > + /*== AbsMask ==*/ > + .long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff > + /*== OneHalf ==*/ > + .align 32 > + .long 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000, 0x3f000000 > + /*== SmallNorm ==*/ > + .align 32 > + .long 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000, 0x2f800000 > + /*== One ==*/ > + .align 32 > + .long 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 > + /*== Two ==*/ > + .align 32 > + .long 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000, 0x40000000 > + /*== sqrt_coeff[2] ==*/ > + .align 32 > + .long 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004, 0xbdC00004 /* sqrt_coeff2 */ > + .long 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001, 0x3e800001 /* sqrt_coeff1 */ > + /*== poly_coeff[5] ==*/ > + .align 32 > + .long 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07, 0x3d2EDC07 /* poly_coeff5 */ > + .long 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B, 0x3CC32A6B /* poly_coeff4 */ > + .long 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4, 0x3d3A9AB4 /* poly_coeff3 */ > + .long 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12, 0x3d997C12 /* poly_coeff2 */ > + .long 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF, 0x3e2AAAFF /* poly_coeff1 */ > + /*== Pi2H ==*/ > + .align 32 > + .long 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB, 0x3fc90FDB > + .align 32 > + .type __svml_sasin_data_internal,@object > + .size __svml_sasin_data_internal,.-__svml_sasin_data_internal > diff --git a/sysdeps/x86_64/fpu/svml_d_asin2_core.S b/sysdeps/x86_64/fpu/svml_d_asin2_core.S > new file mode 100644 > index 0000000000..8ff8bc58df > --- /dev/null > +++ b/sysdeps/x86_64/fpu/svml_d_asin2_core.S > @@ -0,0 +1,29 @@ > +/* Function asin vectorized with SSE2. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#include > +#include "svml_d_wrapper_impl.h" > + > + .text > +ENTRY (_ZGVbN2v_asin) > +WRAPPER_IMPL_SSE2 asin > +END (_ZGVbN2v_asin) > + > +#ifndef USE_MULTIARCH > + libmvec_hidden_def (_ZGVbN2v_asin) > +#endif > diff --git a/sysdeps/x86_64/fpu/svml_d_asin4_core.S b/sysdeps/x86_64/fpu/svml_d_asin4_core.S > new file mode 100644 > index 0000000000..dbe33952bc > --- /dev/null > +++ b/sysdeps/x86_64/fpu/svml_d_asin4_core.S > @@ -0,0 +1,29 @@ > +/* Function asin vectorized with AVX2, wrapper version. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#include > +#include "svml_d_wrapper_impl.h" > + > + .text > +ENTRY (_ZGVdN4v_asin) > +WRAPPER_IMPL_AVX _ZGVbN2v_asin > +END (_ZGVdN4v_asin) > + > +#ifndef USE_MULTIARCH > + libmvec_hidden_def (_ZGVdN4v_asin) > +#endif > diff --git a/sysdeps/x86_64/fpu/svml_d_asin4_core_avx.S b/sysdeps/x86_64/fpu/svml_d_asin4_core_avx.S > new file mode 100644 > index 0000000000..513a31bde5 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/svml_d_asin4_core_avx.S > @@ -0,0 +1,25 @@ > +/* Function asin vectorized in AVX ISA as wrapper to SSE4 ISA version. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#include > +#include "svml_d_wrapper_impl.h" > + > + .text > +ENTRY (_ZGVcN4v_asin) > +WRAPPER_IMPL_AVX _ZGVbN2v_asin > +END (_ZGVcN4v_asin) > diff --git a/sysdeps/x86_64/fpu/svml_d_asin8_core.S b/sysdeps/x86_64/fpu/svml_d_asin8_core.S > new file mode 100644 > index 0000000000..06694298cf > --- /dev/null > +++ b/sysdeps/x86_64/fpu/svml_d_asin8_core.S > @@ -0,0 +1,25 @@ > +/* Function asin vectorized with AVX-512, wrapper to AVX2. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#include > +#include "svml_d_wrapper_impl.h" > + > + .text > +ENTRY (_ZGVeN8v_asin) > +WRAPPER_IMPL_AVX512 _ZGVdN4v_asin > +END (_ZGVeN8v_asin) > diff --git a/sysdeps/x86_64/fpu/svml_s_asinf16_core.S b/sysdeps/x86_64/fpu/svml_s_asinf16_core.S > new file mode 100644 > index 0000000000..015d583e3f > --- /dev/null > +++ b/sysdeps/x86_64/fpu/svml_s_asinf16_core.S > @@ -0,0 +1,25 @@ > +/* Function asinf vectorized with AVX-512. Wrapper to AVX2 version. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#include > +#include "svml_s_wrapper_impl.h" > + > + .text > +ENTRY (_ZGVeN16v_asinf) > +WRAPPER_IMPL_AVX512 _ZGVdN8v_asinf > +END (_ZGVeN16v_asinf) > diff --git a/sysdeps/x86_64/fpu/svml_s_asinf4_core.S b/sysdeps/x86_64/fpu/svml_s_asinf4_core.S > new file mode 100644 > index 0000000000..d80f06c16d > --- /dev/null > +++ b/sysdeps/x86_64/fpu/svml_s_asinf4_core.S > @@ -0,0 +1,29 @@ > +/* Function asinf vectorized with SSE2, wrapper version. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#include > +#include "svml_s_wrapper_impl.h" > + > + .text > +ENTRY (_ZGVbN4v_asinf) > +WRAPPER_IMPL_SSE2 asinf > +END (_ZGVbN4v_asinf) > + > +#ifndef USE_MULTIARCH > + libmvec_hidden_def (_ZGVbN4v_asinf) > +#endif > diff --git a/sysdeps/x86_64/fpu/svml_s_asinf8_core.S b/sysdeps/x86_64/fpu/svml_s_asinf8_core.S > new file mode 100644 > index 0000000000..304ad0a7f5 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/svml_s_asinf8_core.S > @@ -0,0 +1,29 @@ > +/* Function asinf vectorized with AVX2, wrapper version. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#include > +#include "svml_s_wrapper_impl.h" > + > + .text > +ENTRY (_ZGVdN8v_asinf) > +WRAPPER_IMPL_AVX _ZGVbN4v_asinf > +END (_ZGVdN8v_asinf) > + > +#ifndef USE_MULTIARCH > + libmvec_hidden_def (_ZGVdN8v_asinf) > +#endif > diff --git a/sysdeps/x86_64/fpu/svml_s_asinf8_core_avx.S b/sysdeps/x86_64/fpu/svml_s_asinf8_core_avx.S > new file mode 100644 > index 0000000000..a2f7dc112e > --- /dev/null > +++ b/sysdeps/x86_64/fpu/svml_s_asinf8_core_avx.S > @@ -0,0 +1,25 @@ > +/* Function asinf vectorized in AVX ISA as wrapper to SSE4 ISA version. > + Copyright (C) 2021 Free Software Foundation, Inc. > + This file is part of the GNU C Library. > + > + The GNU C Library is free software; you can redistribute it and/or > + modify it under the terms of the GNU Lesser General Public > + License as published by the Free Software Foundation; either > + version 2.1 of the License, or (at your option) any later version. > + > + The GNU C Library is distributed in the hope that it will be useful, > + but WITHOUT ANY WARRANTY; without even the implied warranty of > + MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU > + Lesser General Public License for more details. > + > + You should have received a copy of the GNU Lesser General Public > + License along with the GNU C Library; if not, see > + . */ > + > +#include > +#include "svml_s_wrapper_impl.h" > + > + .text > +ENTRY (_ZGVcN8v_asinf) > +WRAPPER_IMPL_AVX _ZGVbN4v_asinf > +END (_ZGVcN8v_asinf) > diff --git a/sysdeps/x86_64/fpu/test-double-libmvec-asin-avx.c b/sysdeps/x86_64/fpu/test-double-libmvec-asin-avx.c > new file mode 100644 > index 0000000000..e37cfdce58 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/test-double-libmvec-asin-avx.c > @@ -0,0 +1 @@ > +#include "test-double-libmvec-asin.c" > diff --git a/sysdeps/x86_64/fpu/test-double-libmvec-asin-avx2.c b/sysdeps/x86_64/fpu/test-double-libmvec-asin-avx2.c > new file mode 100644 > index 0000000000..e37cfdce58 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/test-double-libmvec-asin-avx2.c > @@ -0,0 +1 @@ > +#include "test-double-libmvec-asin.c" > diff --git a/sysdeps/x86_64/fpu/test-double-libmvec-asin-avx512f.c b/sysdeps/x86_64/fpu/test-double-libmvec-asin-avx512f.c > new file mode 100644 > index 0000000000..e37cfdce58 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/test-double-libmvec-asin-avx512f.c > @@ -0,0 +1 @@ > +#include "test-double-libmvec-asin.c" > diff --git a/sysdeps/x86_64/fpu/test-double-libmvec-asin.c b/sysdeps/x86_64/fpu/test-double-libmvec-asin.c > new file mode 100644 > index 0000000000..d2e16e67f4 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/test-double-libmvec-asin.c > @@ -0,0 +1,3 @@ > +#define LIBMVEC_TYPE double > +#define LIBMVEC_FUNC asin > +#include "test-vector-abi-arg1.h" > diff --git a/sysdeps/x86_64/fpu/test-double-vlen2-wrappers.c b/sysdeps/x86_64/fpu/test-double-vlen2-wrappers.c > index 467c913990..5746bb5be3 100644 > --- a/sysdeps/x86_64/fpu/test-double-vlen2-wrappers.c > +++ b/sysdeps/x86_64/fpu/test-double-vlen2-wrappers.c > @@ -29,6 +29,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp), _ZGVbN2v_exp) > VECTOR_WRAPPER_ff (WRAPPER_NAME (pow), _ZGVbN2vv_pow) > VECTOR_WRAPPER (WRAPPER_NAME (acos), _ZGVbN2v_acos) > VECTOR_WRAPPER (WRAPPER_NAME (atan), _ZGVbN2v_atan) > +VECTOR_WRAPPER (WRAPPER_NAME (asin), _ZGVbN2v_asin) > > #define VEC_INT_TYPE __m128i > > diff --git a/sysdeps/x86_64/fpu/test-double-vlen4-avx2-wrappers.c b/sysdeps/x86_64/fpu/test-double-vlen4-avx2-wrappers.c > index b72a7de84e..8d3d5493ed 100644 > --- a/sysdeps/x86_64/fpu/test-double-vlen4-avx2-wrappers.c > +++ b/sysdeps/x86_64/fpu/test-double-vlen4-avx2-wrappers.c > @@ -32,6 +32,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp), _ZGVdN4v_exp) > VECTOR_WRAPPER_ff (WRAPPER_NAME (pow), _ZGVdN4vv_pow) > VECTOR_WRAPPER (WRAPPER_NAME (acos), _ZGVdN4v_acos) > VECTOR_WRAPPER (WRAPPER_NAME (atan), _ZGVdN4v_atan) > +VECTOR_WRAPPER (WRAPPER_NAME (asin), _ZGVdN4v_asin) > > #ifndef __ILP32__ > # define VEC_INT_TYPE __m256i > diff --git a/sysdeps/x86_64/fpu/test-double-vlen4-wrappers.c b/sysdeps/x86_64/fpu/test-double-vlen4-wrappers.c > index d2434df21e..f43328f2ff 100644 > --- a/sysdeps/x86_64/fpu/test-double-vlen4-wrappers.c > +++ b/sysdeps/x86_64/fpu/test-double-vlen4-wrappers.c > @@ -29,6 +29,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp), _ZGVcN4v_exp) > VECTOR_WRAPPER_ff (WRAPPER_NAME (pow), _ZGVcN4vv_pow) > VECTOR_WRAPPER (WRAPPER_NAME (acos), _ZGVcN4v_acos) > VECTOR_WRAPPER (WRAPPER_NAME (atan), _ZGVcN4v_atan) > +VECTOR_WRAPPER (WRAPPER_NAME (asin), _ZGVcN4v_asin) > > #define VEC_INT_TYPE __m128i > > diff --git a/sysdeps/x86_64/fpu/test-double-vlen8-wrappers.c b/sysdeps/x86_64/fpu/test-double-vlen8-wrappers.c > index f7aaf8159e..8b566c199a 100644 > --- a/sysdeps/x86_64/fpu/test-double-vlen8-wrappers.c > +++ b/sysdeps/x86_64/fpu/test-double-vlen8-wrappers.c > @@ -29,6 +29,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp), _ZGVeN8v_exp) > VECTOR_WRAPPER_ff (WRAPPER_NAME (pow), _ZGVeN8vv_pow) > VECTOR_WRAPPER (WRAPPER_NAME (acos), _ZGVeN8v_acos) > VECTOR_WRAPPER (WRAPPER_NAME (atan), _ZGVeN8v_atan) > +VECTOR_WRAPPER (WRAPPER_NAME (asin), _ZGVeN8v_asin) > > #ifndef __ILP32__ > # define VEC_INT_TYPE __m512i > diff --git a/sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx.c b/sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx.c > new file mode 100644 > index 0000000000..6aa8f5f370 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx.c > @@ -0,0 +1 @@ > +#include "test-float-libmvec-asinf.c" > diff --git a/sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx2.c b/sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx2.c > new file mode 100644 > index 0000000000..6aa8f5f370 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx2.c > @@ -0,0 +1 @@ > +#include "test-float-libmvec-asinf.c" > diff --git a/sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx512f.c b/sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx512f.c > new file mode 100644 > index 0000000000..6aa8f5f370 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/test-float-libmvec-asinf-avx512f.c > @@ -0,0 +1 @@ > +#include "test-float-libmvec-asinf.c" > diff --git a/sysdeps/x86_64/fpu/test-float-libmvec-asinf.c b/sysdeps/x86_64/fpu/test-float-libmvec-asinf.c > new file mode 100644 > index 0000000000..2bbe2395a0 > --- /dev/null > +++ b/sysdeps/x86_64/fpu/test-float-libmvec-asinf.c > @@ -0,0 +1,3 @@ > +#define LIBMVEC_TYPE float > +#define LIBMVEC_FUNC asinf > +#include "test-vector-abi-arg1.h" > diff --git a/sysdeps/x86_64/fpu/test-float-vlen16-wrappers.c b/sysdeps/x86_64/fpu/test-float-vlen16-wrappers.c > index af769c56fa..3d3218a310 100644 > --- a/sysdeps/x86_64/fpu/test-float-vlen16-wrappers.c > +++ b/sysdeps/x86_64/fpu/test-float-vlen16-wrappers.c > @@ -29,6 +29,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (expf), _ZGVeN16v_expf) > VECTOR_WRAPPER_ff (WRAPPER_NAME (powf), _ZGVeN16vv_powf) > VECTOR_WRAPPER (WRAPPER_NAME (acosf), _ZGVeN16v_acosf) > VECTOR_WRAPPER (WRAPPER_NAME (atanf), _ZGVeN16v_atanf) > +VECTOR_WRAPPER (WRAPPER_NAME (asinf), _ZGVeN16v_asinf) > > #define VEC_INT_TYPE __m512i > > diff --git a/sysdeps/x86_64/fpu/test-float-vlen4-wrappers.c b/sysdeps/x86_64/fpu/test-float-vlen4-wrappers.c > index 76e61d2f1e..7d75b9f60f 100644 > --- a/sysdeps/x86_64/fpu/test-float-vlen4-wrappers.c > +++ b/sysdeps/x86_64/fpu/test-float-vlen4-wrappers.c > @@ -29,6 +29,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (expf), _ZGVbN4v_expf) > VECTOR_WRAPPER_ff (WRAPPER_NAME (powf), _ZGVbN4vv_powf) > VECTOR_WRAPPER (WRAPPER_NAME (acosf), _ZGVbN4v_acosf) > VECTOR_WRAPPER (WRAPPER_NAME (atanf), _ZGVbN4v_atanf) > +VECTOR_WRAPPER (WRAPPER_NAME (asinf), _ZGVbN4v_asinf) > > #define VEC_INT_TYPE __m128i > > diff --git a/sysdeps/x86_64/fpu/test-float-vlen8-avx2-wrappers.c b/sysdeps/x86_64/fpu/test-float-vlen8-avx2-wrappers.c > index 5e27eaaf29..405dde49bc 100644 > --- a/sysdeps/x86_64/fpu/test-float-vlen8-avx2-wrappers.c > +++ b/sysdeps/x86_64/fpu/test-float-vlen8-avx2-wrappers.c > @@ -32,6 +32,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (expf), _ZGVdN8v_expf) > VECTOR_WRAPPER_ff (WRAPPER_NAME (powf), _ZGVdN8vv_powf) > VECTOR_WRAPPER (WRAPPER_NAME (acosf), _ZGVdN8v_acosf) > VECTOR_WRAPPER (WRAPPER_NAME (atanf), _ZGVdN8v_atanf) > +VECTOR_WRAPPER (WRAPPER_NAME (asinf), _ZGVdN8v_asinf) > > /* Redefinition of wrapper to be compatible with _ZGVdN8vvv_sincosf. */ > #undef VECTOR_WRAPPER_fFF > diff --git a/sysdeps/x86_64/fpu/test-float-vlen8-wrappers.c b/sysdeps/x86_64/fpu/test-float-vlen8-wrappers.c > index 28daf79aa9..7558443f2e 100644 > --- a/sysdeps/x86_64/fpu/test-float-vlen8-wrappers.c > +++ b/sysdeps/x86_64/fpu/test-float-vlen8-wrappers.c > @@ -29,6 +29,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (expf), _ZGVcN8v_expf) > VECTOR_WRAPPER_ff (WRAPPER_NAME (powf), _ZGVcN8vv_powf) > VECTOR_WRAPPER (WRAPPER_NAME (acosf), _ZGVcN8v_acosf) > VECTOR_WRAPPER (WRAPPER_NAME (atanf), _ZGVcN8v_atanf) > +VECTOR_WRAPPER (WRAPPER_NAME (asinf), _ZGVcN8v_asinf) > > #define VEC_INT_TYPE __m128i > > -- > 2.31.1 > LGTM. Reviewed-by: H.J. Lu Thanks. H.J.