From patchwork Mon Dec 27 15:04:48 2021 Content-Type: text/plain; charset="utf-8" MIME-Version: 1.0 Content-Transfer-Encoding: 7bit X-Patchwork-Submitter: Sunil Pandey X-Patchwork-Id: 49270 Return-Path: X-Original-To: patchwork@sourceware.org Delivered-To: patchwork@sourceware.org Received: from server2.sourceware.org (localhost [IPv6:::1]) by sourceware.org (Postfix) with ESMTP id 30CE5385843A for ; Mon, 27 Dec 2021 15:13:25 +0000 (GMT) DKIM-Filter: OpenDKIM Filter v2.11.0 sourceware.org 30CE5385843A DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/relaxed; d=sourceware.org; s=default; t=1640618005; bh=MmzzQMl6TkhAUAiKXYQ6YM4j+wludqcQgDDalozPkaY=; h=To:Subject:Date:In-Reply-To:References:List-Id:List-Unsubscribe: List-Archive:List-Post:List-Help:List-Subscribe:From:Reply-To:Cc: From; b=DDrUvQZiafdSEFDjqFM7no3BmPPXZJOnRH7RETziXMyOBPfFJBDMa7E+NjYU4cQ6f MyARWR85pl9SvvP4thMTUIshEPRywLZyaCZq3+NeyG7ZTKQHfXPathrE9Yxq8sqJ63 OY3cKTsbpQ9jo/6rIaiza3EGkIcUfjmtLa/56hAA= X-Original-To: libc-alpha@sourceware.org Delivered-To: libc-alpha@sourceware.org Received: from mga03.intel.com (mga03.intel.com [134.134.136.65]) by sourceware.org (Postfix) with ESMTPS id 708D7385841C for ; Mon, 27 Dec 2021 15:05:01 +0000 (GMT) DMARC-Filter: OpenDMARC Filter v1.4.1 sourceware.org 708D7385841C X-IronPort-AV: E=McAfee;i="6200,9189,10209"; a="241172164" X-IronPort-AV: E=Sophos;i="5.88,239,1635231600"; d="scan'208";a="241172164" Received: from fmsmga001.fm.intel.com ([10.253.24.23]) by orsmga103.jf.intel.com with ESMTP/TLS/ECDHE-RSA-AES256-GCM-SHA384; 27 Dec 2021 07:04:59 -0800 X-ExtLoop1: 1 X-IronPort-AV: E=Sophos;i="5.88,239,1635231600"; d="scan'208";a="665770337" Received: from scymds01.sc.intel.com ([10.148.94.138]) by fmsmga001.fm.intel.com with ESMTP; 27 Dec 2021 07:04:59 -0800 Received: from gskx-1.sc.intel.com (gskx-1.sc.intel.com [172.25.149.211]) by scymds01.sc.intel.com with ESMTP id 1BRF4v2h002146; Mon, 27 Dec 2021 07:04:58 -0800 To: libc-alpha@sourceware.org Subject: [PATCH v3 09/18] x86-64: Add vector cbrt/cbrtf implementation to libmvec Date: Mon, 27 Dec 2021 07:04:48 -0800 Message-Id: <20211227150457.1680245-10-skpgkp2@gmail.com> X-Mailer: git-send-email 2.31.1 In-Reply-To: <20211227150457.1680245-1-skpgkp2@gmail.com> References: <20211227150457.1680245-1-skpgkp2@gmail.com> MIME-Version: 1.0 X-Spam-Status: No, score=-7.1 required=5.0 tests=BAYES_00, DKIM_ADSP_CUSTOM_MED, FORGED_GMAIL_RCVD, FREEMAIL_ENVFROM_END_DIGIT, FREEMAIL_FROM, GIT_PATCH_0, HK_RANDOM_ENVFROM, HK_RANDOM_FROM, KAM_DMARC_NONE, KAM_DMARC_STATUS, KAM_SHORT, KAM_STOCKGEN, NML_ADSP_CUSTOM_MED, SPF_HELO_NONE, SPF_SOFTFAIL, 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-Patchwork-Original-From: Sunil K Pandey via Libc-alpha From: Sunil Pandey Reply-To: Sunil K Pandey Cc: andrey.kolesov@intel.com, marius.cornea@intel.com Errors-To: libc-alpha-bounces+patchwork=sourceware.org@sourceware.org Sender: "Libc-alpha" Implement vectorized cbrt/cbrtf containing SSE, AVX, AVX2 and AVX512 versions for libmvec as per vector ABI. It also contains accuracy and ABI tests for vector cbrt/cbrtf 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_cbrt2_core-sse2.S | 20 + .../x86_64/fpu/multiarch/svml_d_cbrt2_core.c | 27 + .../fpu/multiarch/svml_d_cbrt2_core_sse4.S | 467 ++++++++++++++++ .../fpu/multiarch/svml_d_cbrt4_core-sse.S | 20 + .../x86_64/fpu/multiarch/svml_d_cbrt4_core.c | 27 + .../fpu/multiarch/svml_d_cbrt4_core_avx2.S | 505 +++++++++++++++++ .../fpu/multiarch/svml_d_cbrt8_core-avx2.S | 20 + .../x86_64/fpu/multiarch/svml_d_cbrt8_core.c | 27 + .../fpu/multiarch/svml_d_cbrt8_core_avx512.S | 253 +++++++++ .../fpu/multiarch/svml_s_cbrtf16_core-avx2.S | 20 + .../fpu/multiarch/svml_s_cbrtf16_core.c | 28 + .../multiarch/svml_s_cbrtf16_core_avx512.S | 235 ++++++++ .../fpu/multiarch/svml_s_cbrtf4_core-sse2.S | 20 + .../x86_64/fpu/multiarch/svml_s_cbrtf4_core.c | 28 + .../fpu/multiarch/svml_s_cbrtf4_core_sse4.S | 490 +++++++++++++++++ .../fpu/multiarch/svml_s_cbrtf8_core-sse.S | 20 + .../x86_64/fpu/multiarch/svml_s_cbrtf8_core.c | 28 + .../fpu/multiarch/svml_s_cbrtf8_core_avx2.S | 509 ++++++++++++++++++ sysdeps/x86_64/fpu/svml_d_cbrt2_core.S | 29 + sysdeps/x86_64/fpu/svml_d_cbrt4_core.S | 29 + sysdeps/x86_64/fpu/svml_d_cbrt4_core_avx.S | 25 + sysdeps/x86_64/fpu/svml_d_cbrt8_core.S | 25 + sysdeps/x86_64/fpu/svml_s_cbrtf16_core.S | 25 + sysdeps/x86_64/fpu/svml_s_cbrtf4_core.S | 29 + sysdeps/x86_64/fpu/svml_s_cbrtf8_core.S | 29 + sysdeps/x86_64/fpu/svml_s_cbrtf8_core_avx.S | 25 + .../x86_64/fpu/test-double-libmvec-cbrt-avx.c | 1 + .../fpu/test-double-libmvec-cbrt-avx2.c | 1 + .../fpu/test-double-libmvec-cbrt-avx512f.c | 1 + sysdeps/x86_64/fpu/test-double-libmvec-cbrt.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-cbrtf-avx.c | 1 + .../fpu/test-float-libmvec-cbrtf-avx2.c | 1 + .../fpu/test-float-libmvec-cbrtf-avx512f.c | 1 + sysdeps/x86_64/fpu/test-float-libmvec-cbrtf.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, 3031 insertions(+), 1 deletion(-) create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core-sse2.S create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core.c create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core_sse4.S create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core-sse.S create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core.c create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core_avx2.S create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core-avx2.S create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core.c create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core_avx512.S create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core-avx2.S create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core.c create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core_avx512.S create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core-sse2.S create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core.c create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core_sse4.S create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core-sse.S create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core.c create mode 100644 sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core_avx2.S create mode 100644 sysdeps/x86_64/fpu/svml_d_cbrt2_core.S create mode 100644 sysdeps/x86_64/fpu/svml_d_cbrt4_core.S create mode 100644 sysdeps/x86_64/fpu/svml_d_cbrt4_core_avx.S create mode 100644 sysdeps/x86_64/fpu/svml_d_cbrt8_core.S create mode 100644 sysdeps/x86_64/fpu/svml_s_cbrtf16_core.S create mode 100644 sysdeps/x86_64/fpu/svml_s_cbrtf4_core.S create mode 100644 sysdeps/x86_64/fpu/svml_s_cbrtf8_core.S create mode 100644 sysdeps/x86_64/fpu/svml_s_cbrtf8_core_avx.S create mode 100644 sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx.c create mode 100644 sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx2.c create mode 100644 sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx512f.c create mode 100644 sysdeps/x86_64/fpu/test-double-libmvec-cbrt.c create mode 100644 sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx.c create mode 100644 sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx2.c create mode 100644 sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx512f.c create mode 100644 sysdeps/x86_64/fpu/test-float-libmvec-cbrtf.c diff --git a/bits/libm-simd-decl-stubs.h b/bits/libm-simd-decl-stubs.h index 6347320521..7f1304ed1d 100644 --- a/bits/libm-simd-decl-stubs.h +++ b/bits/libm-simd-decl-stubs.h @@ -197,4 +197,15 @@ #define __DECL_SIMD_sinhf32x #define __DECL_SIMD_sinhf64x #define __DECL_SIMD_sinhf128x + +#define __DECL_SIMD_cbrt +#define __DECL_SIMD_cbrtf +#define __DECL_SIMD_cbrtl +#define __DECL_SIMD_cbrtf16 +#define __DECL_SIMD_cbrtf32 +#define __DECL_SIMD_cbrtf64 +#define __DECL_SIMD_cbrtf128 +#define __DECL_SIMD_cbrtf32x +#define __DECL_SIMD_cbrtf64x +#define __DECL_SIMD_cbrtf128x #endif diff --git a/math/bits/mathcalls.h b/math/bits/mathcalls.h index 673b3a93ba..26d18f0135 100644 --- a/math/bits/mathcalls.h +++ b/math/bits/mathcalls.h @@ -149,7 +149,7 @@ __MATHCALL_VEC (hypot,, (_Mdouble_ __x, _Mdouble_ __y)); #if defined __USE_XOPEN_EXTENDED || defined __USE_ISOC99 /* Return the cube root of X. */ -__MATHCALL (cbrt,, (_Mdouble_ __x)); +__MATHCALL_VEC (cbrt,, (_Mdouble_ __x)); #endif diff --git a/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist b/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist index f9d7b085ab..a6558d9810 100644 --- a/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist +++ b/sysdeps/unix/sysv/linux/x86_64/libmvec.abilist @@ -49,6 +49,7 @@ 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 _ZGVbN2v_cbrt F GLIBC_2.35 _ZGVbN2v_cosh F GLIBC_2.35 _ZGVbN2v_exp10 F GLIBC_2.35 _ZGVbN2v_exp2 F @@ -58,6 +59,7 @@ GLIBC_2.35 _ZGVbN2vv_hypot F GLIBC_2.35 _ZGVbN4v_acosf F GLIBC_2.35 _ZGVbN4v_asinf F GLIBC_2.35 _ZGVbN4v_atanf F +GLIBC_2.35 _ZGVbN4v_cbrtf F GLIBC_2.35 _ZGVbN4v_coshf F GLIBC_2.35 _ZGVbN4v_exp10f F GLIBC_2.35 _ZGVbN4v_exp2f F @@ -67,6 +69,7 @@ GLIBC_2.35 _ZGVbN4vv_hypotf F GLIBC_2.35 _ZGVcN4v_acos F GLIBC_2.35 _ZGVcN4v_asin F GLIBC_2.35 _ZGVcN4v_atan F +GLIBC_2.35 _ZGVcN4v_cbrt F GLIBC_2.35 _ZGVcN4v_cosh F GLIBC_2.35 _ZGVcN4v_exp10 F GLIBC_2.35 _ZGVcN4v_exp2 F @@ -76,6 +79,7 @@ GLIBC_2.35 _ZGVcN4vv_hypot F GLIBC_2.35 _ZGVcN8v_acosf F GLIBC_2.35 _ZGVcN8v_asinf F GLIBC_2.35 _ZGVcN8v_atanf F +GLIBC_2.35 _ZGVcN8v_cbrtf F GLIBC_2.35 _ZGVcN8v_coshf F GLIBC_2.35 _ZGVcN8v_exp10f F GLIBC_2.35 _ZGVcN8v_exp2f F @@ -85,6 +89,7 @@ GLIBC_2.35 _ZGVcN8vv_hypotf F GLIBC_2.35 _ZGVdN4v_acos F GLIBC_2.35 _ZGVdN4v_asin F GLIBC_2.35 _ZGVdN4v_atan F +GLIBC_2.35 _ZGVdN4v_cbrt F GLIBC_2.35 _ZGVdN4v_cosh F GLIBC_2.35 _ZGVdN4v_exp10 F GLIBC_2.35 _ZGVdN4v_exp2 F @@ -94,6 +99,7 @@ GLIBC_2.35 _ZGVdN4vv_hypot F GLIBC_2.35 _ZGVdN8v_acosf F GLIBC_2.35 _ZGVdN8v_asinf F GLIBC_2.35 _ZGVdN8v_atanf F +GLIBC_2.35 _ZGVdN8v_cbrtf F GLIBC_2.35 _ZGVdN8v_coshf F GLIBC_2.35 _ZGVdN8v_exp10f F GLIBC_2.35 _ZGVdN8v_exp2f F @@ -103,6 +109,7 @@ GLIBC_2.35 _ZGVdN8vv_hypotf F GLIBC_2.35 _ZGVeN16v_acosf F GLIBC_2.35 _ZGVeN16v_asinf F GLIBC_2.35 _ZGVeN16v_atanf F +GLIBC_2.35 _ZGVeN16v_cbrtf F GLIBC_2.35 _ZGVeN16v_coshf F GLIBC_2.35 _ZGVeN16v_exp10f F GLIBC_2.35 _ZGVeN16v_exp2f F @@ -112,6 +119,7 @@ GLIBC_2.35 _ZGVeN16vv_hypotf F GLIBC_2.35 _ZGVeN8v_acos F GLIBC_2.35 _ZGVeN8v_asin F GLIBC_2.35 _ZGVeN8v_atan F +GLIBC_2.35 _ZGVeN8v_cbrt F GLIBC_2.35 _ZGVeN8v_cosh F GLIBC_2.35 _ZGVeN8v_exp10 F GLIBC_2.35 _ZGVeN8v_exp2 F diff --git a/sysdeps/x86/fpu/bits/math-vector.h b/sysdeps/x86/fpu/bits/math-vector.h index 51a41cfebc..dcd45934ab 100644 --- a/sysdeps/x86/fpu/bits/math-vector.h +++ b/sysdeps/x86/fpu/bits/math-vector.h @@ -94,6 +94,10 @@ # define __DECL_SIMD_sinh __DECL_SIMD_x86_64 # undef __DECL_SIMD_sinhf # define __DECL_SIMD_sinhf __DECL_SIMD_x86_64 +# undef __DECL_SIMD_cbrt +# define __DECL_SIMD_cbrt __DECL_SIMD_x86_64 +# undef __DECL_SIMD_cbrtf +# define __DECL_SIMD_cbrtf __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 91e9b4fc83..dfb5f13ea3 100644 --- a/sysdeps/x86/fpu/finclude/math-vector-fortran.h +++ b/sysdeps/x86/fpu/finclude/math-vector-fortran.h @@ -46,6 +46,8 @@ !GCC$ builtin (expm1f) attributes simd (notinbranch) if('x86_64') !GCC$ builtin (sinh) attributes simd (notinbranch) if('x86_64') !GCC$ builtin (sinhf) attributes simd (notinbranch) if('x86_64') +!GCC$ builtin (cbrt) attributes simd (notinbranch) if('x86_64') +!GCC$ builtin (cbrtf) attributes simd (notinbranch) if('x86_64') !GCC$ builtin (cos) attributes simd (notinbranch) if('x32') !GCC$ builtin (cosf) attributes simd (notinbranch) if('x32') @@ -77,3 +79,5 @@ !GCC$ builtin (expm1f) attributes simd (notinbranch) if('x32') !GCC$ builtin (sinh) attributes simd (notinbranch) if('x32') !GCC$ builtin (sinhf) attributes simd (notinbranch) if('x32') +!GCC$ builtin (cbrt) attributes simd (notinbranch) if('x32') +!GCC$ builtin (cbrtf) attributes simd (notinbranch) if('x32') diff --git a/sysdeps/x86_64/fpu/Makeconfig b/sysdeps/x86_64/fpu/Makeconfig index 81e9fc95b2..dde737c0d6 100644 --- a/sysdeps/x86_64/fpu/Makeconfig +++ b/sysdeps/x86_64/fpu/Makeconfig @@ -25,6 +25,7 @@ libmvec-funcs = \ acos \ asin \ atan \ + cbrt \ cos \ cosh \ exp \ diff --git a/sysdeps/x86_64/fpu/Versions b/sysdeps/x86_64/fpu/Versions index 2710446d12..b70aeb3e2f 100644 --- a/sysdeps/x86_64/fpu/Versions +++ b/sysdeps/x86_64/fpu/Versions @@ -17,6 +17,7 @@ libmvec { _ZGVbN2v_acos; _ZGVcN4v_acos; _ZGVdN4v_acos; _ZGVeN8v_acos; _ZGVbN2v_asin; _ZGVcN4v_asin; _ZGVdN4v_asin; _ZGVeN8v_asin; _ZGVbN2v_atan; _ZGVcN4v_atan; _ZGVdN4v_atan; _ZGVeN8v_atan; + _ZGVbN2v_cbrt; _ZGVcN4v_cbrt; _ZGVdN4v_cbrt; _ZGVeN8v_cbrt; _ZGVbN2v_cosh; _ZGVcN4v_cosh; _ZGVdN4v_cosh; _ZGVeN8v_cosh; _ZGVbN2v_exp10; _ZGVcN4v_exp10; _ZGVdN4v_exp10; _ZGVeN8v_exp10; _ZGVbN2v_exp2; _ZGVcN4v_exp2; _ZGVdN4v_exp2; _ZGVeN8v_exp2; @@ -26,6 +27,7 @@ libmvec { _ZGVbN4v_acosf; _ZGVcN8v_acosf; _ZGVdN8v_acosf; _ZGVeN16v_acosf; _ZGVbN4v_asinf; _ZGVcN8v_asinf; _ZGVdN8v_asinf; _ZGVeN16v_asinf; _ZGVbN4v_atanf; _ZGVcN8v_atanf; _ZGVdN8v_atanf; _ZGVeN16v_atanf; + _ZGVbN4v_cbrtf; _ZGVcN8v_cbrtf; _ZGVdN8v_cbrtf; _ZGVeN16v_cbrtf; _ZGVbN4v_coshf; _ZGVcN8v_coshf; _ZGVdN8v_coshf; _ZGVeN16v_coshf; _ZGVbN4v_exp10f; _ZGVcN8v_exp10f; _ZGVdN8v_exp10f; _ZGVeN16v_exp10f; _ZGVbN4v_exp2f; _ZGVcN8v_exp2f; _ZGVdN8v_exp2f; _ZGVeN16v_exp2f; diff --git a/sysdeps/x86_64/fpu/libm-test-ulps b/sysdeps/x86_64/fpu/libm-test-ulps index f4b313119d..e039a993df 100644 --- a/sysdeps/x86_64/fpu/libm-test-ulps +++ b/sysdeps/x86_64/fpu/libm-test-ulps @@ -583,6 +583,26 @@ float: 1 float128: 1 ldouble: 1 +Function: "cbrt_vlen16": +float: 1 + +Function: "cbrt_vlen2": +double: 1 + +Function: "cbrt_vlen4": +double: 1 +float: 2 + +Function: "cbrt_vlen4_avx2": +double: 1 + +Function: "cbrt_vlen8": +double: 1 +float: 2 + +Function: "cbrt_vlen8_avx2": +float: 2 + Function: Real part of "ccos": double: 1 float: 1 diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core-sse2.S b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core-sse2.S new file mode 100644 index 0000000000..60f4c46a11 --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core-sse2.S @@ -0,0 +1,20 @@ +/* SSE2 version of vectorized cbrt, 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_cbrt _ZGVbN2v_cbrt_sse2 +#include "../svml_d_cbrt2_core.S" diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core.c b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core.c new file mode 100644 index 0000000000..07390b7150 --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core.c @@ -0,0 +1,27 @@ +/* Multiple versions of vectorized cbrt, 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_cbrt +#include "ifunc-mathvec-sse4_1.h" + +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); + +#ifdef SHARED +__hidden_ver1 (_ZGVbN2v_cbrt, __GI__ZGVbN2v_cbrt, __redirect__ZGVbN2v_cbrt) + __attribute__ ((visibility ("hidden"))); +#endif diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core_sse4.S b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core_sse4.S new file mode 100644 index 0000000000..72ecb25e05 --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt2_core_sse4.S @@ -0,0 +1,467 @@ +/* Function cbrt 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: + * + * x=2^{3*k+j} * 1.b1 b2 ... b5 b6 ... b52 + * Let r=(x*2^{-3k-j} - 1.b1 b2 ... b5 1)* rcp[b1 b2 ..b5], + * where rcp[b1 b2 .. b5]=1/(1.b1 b2 b3 b4 b5 1) in double precision + * cbrt(2^j * 1. b1 b2 .. b5 1) is approximated as T[j][b1..b5]+D[j][b1..b5] + * (T stores the high 53 bits, D stores the low order bits) + * Result=2^k*T+(2^k*T*r)*P+2^k*D + * where P=p1+p2*r+..+p8*r^7 + * + */ + +/* Offsets for data table __svml_dcbrt_data_internal + */ +#define _dRcp 0 +#define _dCbrtHiLo 256 +#define _dA7 1024 +#define _dA6 1040 +#define _dA5 1056 +#define _dA4 1072 +#define _dA3 1088 +#define _dA2 1104 +#define _dA1 1120 +#define _dNeg65Div64 1136 +#define _dSgnf6Mask 1152 +#define _dNegOne 1168 +#define _dMantissaMask 1184 +#define _lExpHiMask 1200 +#define _lExpLoMask 1216 +#define _l1556 1232 +#define _iRcpIndexMask 1248 +#define _iAbsMask 1264 +#define _iSignMask 1280 +#define _iBias 1296 +#define _iSub 1312 +#define _iCmp 1328 + +#include + + .text + .section .text.sse4,"ax",@progbits +ENTRY(_ZGVbN2v_cbrt_sse4) + subq $72, %rsp + cfi_def_cfa_offset(80) + +/* Calculate CbrtIndex */ + movaps %xmm0, %xmm10 + psrlq $52, %xmm10 + +/* Load 1/(1+iRcpIndex/32+1/64) reciprocal table value */ + lea __svml_dcbrt_data_internal(%rip), %r8 + pand _lExpLoMask+__svml_dcbrt_data_internal(%rip), %xmm10 + movdqu _l1556+__svml_dcbrt_data_internal(%rip), %xmm9 + pmuludq %xmm10, %xmm9 + +/* If the exponent field is zero - go to callout to process denormals */ + movq _iAbsMask+__svml_dcbrt_data_internal(%rip), %xmm7 + +/* Calculate Rcp table index */ + movq _iRcpIndexMask+__svml_dcbrt_data_internal(%rip), %xmm13 + +/* Get iX - high part of argument */ + pshufd $221, %xmm0, %xmm4 + +/* + * Declarations + * Load constants + */ + movq _iSignMask+__svml_dcbrt_data_internal(%rip), %xmm1 + pand %xmm4, %xmm7 + pand %xmm4, %xmm13 + +/* Compute 2^k */ + psrld $20, %xmm4 + movq _iBias+__svml_dcbrt_data_internal(%rip), %xmm2 + pand %xmm1, %xmm4 + pshufd $136, %xmm9, %xmm15 + por %xmm2, %xmm4 + psrld $14, %xmm15 + psrld $12, %xmm13 + paddd %xmm15, %xmm4 + pxor %xmm2, %xmm2 + pslld $20, %xmm4 + movdqa %xmm15, %xmm11 + movd %xmm13, %edx + paddd %xmm15, %xmm11 + pshufd $1, %xmm13, %xmm8 + punpckldq %xmm4, %xmm2 + +/* + * VAND( L, l2k, = l2k, lExpHiMask ); + * Argument reduction Z + */ + movups _dMantissaMask+__svml_dcbrt_data_internal(%rip), %xmm1 + movups _dSgnf6Mask+__svml_dcbrt_data_internal(%rip), %xmm4 + andps %xmm0, %xmm1 + movd %xmm8, %ecx + andps %xmm0, %xmm4 + orps _dNegOne+__svml_dcbrt_data_internal(%rip), %xmm1 + orps _dNeg65Div64+__svml_dcbrt_data_internal(%rip), %xmm4 + movslq %edx, %rdx + subpd %xmm4, %xmm1 + movslq %ecx, %rcx + movsd (%r8,%rdx), %xmm3 + movq _iSub+__svml_dcbrt_data_internal(%rip), %xmm5 + psubd %xmm5, %xmm7 + movhpd (%r8,%rcx), %xmm3 + mulpd %xmm1, %xmm3 + +/* Polynomial */ + movups _dA7+__svml_dcbrt_data_internal(%rip), %xmm5 + mulpd %xmm3, %xmm5 + addpd _dA6+__svml_dcbrt_data_internal(%rip), %xmm5 + mulpd %xmm3, %xmm5 + addpd _dA5+__svml_dcbrt_data_internal(%rip), %xmm5 + mulpd %xmm3, %xmm5 + addpd _dA4+__svml_dcbrt_data_internal(%rip), %xmm5 + mulpd %xmm3, %xmm5 + addpd _dA3+__svml_dcbrt_data_internal(%rip), %xmm5 + pshufd $136, %xmm10, %xmm12 + psubd %xmm15, %xmm12 + psubd %xmm11, %xmm12 + mulpd %xmm3, %xmm5 + pslld $8, %xmm12 + paddd %xmm12, %xmm13 + +/* Load cbrt(2^j*(1+iRcpIndex/32+1/64)) Hi & Lo values */ + movd %xmm13, %esi + pshufd $1, %xmm13, %xmm14 + movq _iCmp+__svml_dcbrt_data_internal(%rip), %xmm6 + movd %xmm14, %edi + pcmpgtd %xmm6, %xmm7 + movmskps %xmm7, %eax + addpd _dA2+__svml_dcbrt_data_internal(%rip), %xmm5 + movslq %esi, %rsi + movslq %edi, %rdi + mulpd %xmm3, %xmm5 + movsd 256(%r8,%rsi), %xmm6 + movhpd 256(%r8,%rdi), %xmm6 + +/* THi*2^k, TLo*2^k */ + mulpd %xmm2, %xmm6 + addpd _dA1+__svml_dcbrt_data_internal(%rip), %xmm5 + +/* THi*2^k*Z */ + mulpd %xmm6, %xmm3 + +/* Final reconstruction */ + mulpd %xmm3, %xmm5 + addpd %xmm5, %xmm6 + andl $3, %eax + +/* Go to special inputs processing branch */ + jne L(SPECIAL_VALUES_BRANCH) + # LOE rbx rbp r12 r13 r14 r15 eax xmm0 xmm6 + +/* Restore registers + * and exit the function + */ + +L(EXIT): + movaps %xmm6, %xmm0 + addq $72, %rsp + cfi_def_cfa_offset(8) + ret + cfi_def_cfa_offset(80) + +/* Branch to process + * special inputs + */ + +L(SPECIAL_VALUES_BRANCH): + movups %xmm0, 32(%rsp) + movups %xmm6, 48(%rsp) + # LOE rbx rbp r12 r13 r14 r15 eax xmm6 + + xorl %edx, %edx + movq %r12, 16(%rsp) + cfi_offset(12, -64) + movl %edx, %r12d + movq %r13, 8(%rsp) + cfi_offset(13, -72) + movl %eax, %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), %xmm6 + +/* 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 xmm6 + +/* Scalar math fucntion call + * to process special input + */ + +L(SCALAR_MATH_CALL): + movl %r12d, %r14d + movsd 32(%rsp,%r14,8), %xmm0 + call cbrt@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_cbrt_sse4) + + .section .rodata, "a" + .align 16 + +#ifdef __svml_dcbrt_data_internal_typedef +typedef unsigned int VUINT32; +typedef struct +{ + __declspec(align(16)) VUINT32 _dRcp[32][2]; + __declspec(align(16)) VUINT32 _dCbrtHiLo[96][2]; + __declspec(align(16)) VUINT32 _dA7[2][2]; + __declspec(align(16)) VUINT32 _dA6[2][2]; + __declspec(align(16)) VUINT32 _dA5[2][2]; + __declspec(align(16)) VUINT32 _dA4[2][2]; + __declspec(align(16)) VUINT32 _dA3[2][2]; + __declspec(align(16)) VUINT32 _dA2[2][2]; + __declspec(align(16)) VUINT32 _dA1[2][2]; + __declspec(align(16)) VUINT32 _dNeg65Div64[2][2]; + __declspec(align(16)) VUINT32 _dSgnf6Mask[2][2]; + __declspec(align(16)) VUINT32 _dNegOne[2][2]; + __declspec(align(16)) VUINT32 _dMantissaMask[2][2]; + __declspec(align(16)) VUINT32 _lExpHiMask[2][2]; + __declspec(align(16)) VUINT32 _lExpLoMask[2][2]; + __declspec(align(16)) VUINT32 _l1556[2][2]; + __declspec(align(16)) VUINT32 _iRcpIndexMask[4][1]; + __declspec(align(16)) VUINT32 _iAbsMask[4][1]; + __declspec(align(16)) VUINT32 _iSignMask[4][1]; + __declspec(align(16)) VUINT32 _iBias[4][1]; + __declspec(align(16)) VUINT32 _iSub[4][1]; + __declspec(align(16)) VUINT32 _iCmp[4][1]; +} __svml_dcbrt_data_internal; +#endif +__svml_dcbrt_data_internal: + /*== _dRcp ==*/ + .quad 0xBFEF81F81F81F820 /* (1/(1+0/32+1/64)) = -.984615 */ + .quad 0xBFEE9131ABF0B767 /* (1/(1+1/32+1/64)) = -.955224 */ + .quad 0xBFEDAE6076B981DB /* (1/(1+2/32+1/64)) = -.927536 */ + .quad 0xBFECD85689039B0B /* (1/(1+3/32+1/64)) = -.901408 */ + .quad 0xBFEC0E070381C0E0 /* (1/(1+4/32+1/64)) = -.876712 */ + .quad 0xBFEB4E81B4E81B4F /* (1/(1+5/32+1/64)) = -.853333 */ + .quad 0xBFEA98EF606A63BE /* (1/(1+6/32+1/64)) = -.831169 */ + .quad 0xBFE9EC8E951033D9 /* (1/(1+7/32+1/64)) = -.810127 */ + .quad 0xBFE948B0FCD6E9E0 /* (1/(1+8/32+1/64)) = -.790123 */ + .quad 0xBFE8ACB90F6BF3AA /* (1/(1+9/32+1/64)) = -.771084 */ + .quad 0xBFE8181818181818 /* (1/(1+10/32+1/64)) = -.752941 */ + .quad 0xBFE78A4C8178A4C8 /* (1/(1+11/32+1/64)) = -.735632 */ + .quad 0xBFE702E05C0B8170 /* (1/(1+12/32+1/64)) = -.719101 */ + .quad 0xBFE6816816816817 /* (1/(1+13/32+1/64)) = -.703297 */ + .quad 0xBFE6058160581606 /* (1/(1+14/32+1/64)) = -.688172 */ + .quad 0xBFE58ED2308158ED /* (1/(1+15/32+1/64)) = -.673684 */ + .quad 0xBFE51D07EAE2F815 /* (1/(1+16/32+1/64)) = -.659794 */ + .quad 0xBFE4AFD6A052BF5B /* (1/(1+17/32+1/64)) = -.646465 */ + .quad 0xBFE446F86562D9FB /* (1/(1+18/32+1/64)) = -.633663 */ + .quad 0xBFE3E22CBCE4A902 /* (1/(1+19/32+1/64)) = -.621359 */ + .quad 0xBFE3813813813814 /* (1/(1+20/32+1/64)) = -.609524 */ + .quad 0xBFE323E34A2B10BF /* (1/(1+21/32+1/64)) = -.598131 */ + .quad 0xBFE2C9FB4D812CA0 /* (1/(1+22/32+1/64)) = -.587156 */ + .quad 0xBFE27350B8812735 /* (1/(1+23/32+1/64)) = -.576577 */ + .quad 0xBFE21FB78121FB78 /* (1/(1+24/32+1/64)) = -.566372 */ + .quad 0xBFE1CF06ADA2811D /* (1/(1+25/32+1/64)) = -.556522 */ + .quad 0xBFE1811811811812 /* (1/(1+26/32+1/64)) = -.547009 */ + .quad 0xBFE135C81135C811 /* (1/(1+27/32+1/64)) = -.537815 */ + .quad 0xBFE0ECF56BE69C90 /* (1/(1+28/32+1/64)) = -.528926 */ + .quad 0xBFE0A6810A6810A7 /* (1/(1+29/32+1/64)) = -.520325 */ + .quad 0xBFE0624DD2F1A9FC /* (1/(1+30/32+1/64)) = -.512 */ + .quad 0xBFE0204081020408 /* (1/(1+31/32+1/64)) = -.503937 */ + /*== _dCbrtHiLo ==*/ + .align 16 + .quad 0x3FF01539221D4C97 /* HI((2^0*(1+0/32+1/64))^(1/3)) = 1.005181 */ + .quad 0x3FF03F06771A2E33 /* HI((2^0*(1+1/32+1/64))^(1/3)) = 1.015387 */ + .quad 0x3FF06800E629D671 /* HI((2^0*(1+2/32+1/64))^(1/3)) = 1.025391 */ + .quad 0x3FF090328731DEB2 /* HI((2^0*(1+3/32+1/64))^(1/3)) = 1.035204 */ + .quad 0x3FF0B7A4B1BD64AC /* HI((2^0*(1+4/32+1/64))^(1/3)) = 1.044835 */ + .quad 0x3FF0DE601024FB87 /* HI((2^0*(1+5/32+1/64))^(1/3)) = 1.054291 */ + .quad 0x3FF1046CB0597000 /* HI((2^0*(1+6/32+1/64))^(1/3)) = 1.06358 */ + .quad 0x3FF129D212A9BA9B /* HI((2^0*(1+7/32+1/64))^(1/3)) = 1.07271 */ + .quad 0x3FF14E9736CDAF38 /* HI((2^0*(1+8/32+1/64))^(1/3)) = 1.081687 */ + .quad 0x3FF172C2A772F507 /* HI((2^0*(1+9/32+1/64))^(1/3)) = 1.090518 */ + .quad 0x3FF1965A848001D3 /* HI((2^0*(1+10/32+1/64))^(1/3)) = 1.099207 */ + .quad 0x3FF1B9648C38C55D /* HI((2^0*(1+11/32+1/64))^(1/3)) = 1.107762 */ + .quad 0x3FF1DBE6236A0C45 /* HI((2^0*(1+12/32+1/64))^(1/3)) = 1.116186 */ + .quad 0x3FF1FDE45CBB1F9F /* HI((2^0*(1+13/32+1/64))^(1/3)) = 1.124485 */ + .quad 0x3FF21F63FF409042 /* HI((2^0*(1+14/32+1/64))^(1/3)) = 1.132664 */ + .quad 0x3FF240698C6746E5 /* HI((2^0*(1+15/32+1/64))^(1/3)) = 1.140726 */ + .quad 0x3FF260F9454BB99B /* HI((2^0*(1+16/32+1/64))^(1/3)) = 1.148675 */ + .quad 0x3FF281172F8E7073 /* HI((2^0*(1+17/32+1/64))^(1/3)) = 1.156516 */ + .quad 0x3FF2A0C719B4B6D0 /* HI((2^0*(1+18/32+1/64))^(1/3)) = 1.164252 */ + .quad 0x3FF2C00C9F2263EC /* HI((2^0*(1+19/32+1/64))^(1/3)) = 1.171887 */ + .quad 0x3FF2DEEB2BB7FB78 /* HI((2^0*(1+20/32+1/64))^(1/3)) = 1.179423 */ + .quad 0x3FF2FD65FF1EFBBC /* HI((2^0*(1+21/32+1/64))^(1/3)) = 1.186865 */ + .quad 0x3FF31B802FCCF6A2 /* HI((2^0*(1+22/32+1/64))^(1/3)) = 1.194214 */ + .quad 0x3FF3393CADC50708 /* HI((2^0*(1+23/32+1/64))^(1/3)) = 1.201474 */ + .quad 0x3FF3569E451E4C2A /* HI((2^0*(1+24/32+1/64))^(1/3)) = 1.208647 */ + .quad 0x3FF373A7A0554CDE /* HI((2^0*(1+25/32+1/64))^(1/3)) = 1.215736 */ + .quad 0x3FF3905B4A6D76CE /* HI((2^0*(1+26/32+1/64))^(1/3)) = 1.222743 */ + .quad 0x3FF3ACBBB0E756B6 /* HI((2^0*(1+27/32+1/64))^(1/3)) = 1.229671 */ + .quad 0x3FF3C8CB258FA340 /* HI((2^0*(1+28/32+1/64))^(1/3)) = 1.236522 */ + .quad 0x3FF3E48BE02AC0CE /* HI((2^0*(1+29/32+1/64))^(1/3)) = 1.243297 */ + .quad 0x3FF4000000000000 /* HI((2^0*(1+30/32+1/64))^(1/3)) = 1.25 */ + .quad 0x3FF41B298D47800E /* HI((2^0*(1+31/32+1/64))^(1/3)) = 1.256631 */ + .quad 0x3FF443604B34D9B2 /* HI((2^1*(1+0/32+1/64))^(1/3)) = 1.266449 */ + .quad 0x3FF4780B20906571 /* HI((2^1*(1+1/32+1/64))^(1/3)) = 1.279307 */ + .quad 0x3FF4ABAC3EE06706 /* HI((2^1*(1+2/32+1/64))^(1/3)) = 1.291912 */ + .quad 0x3FF4DE505DA66B8D /* HI((2^1*(1+3/32+1/64))^(1/3)) = 1.304276 */ + .quad 0x3FF51003420A5C07 /* HI((2^1*(1+4/32+1/64))^(1/3)) = 1.316409 */ + .quad 0x3FF540CFD6FD11C1 /* HI((2^1*(1+5/32+1/64))^(1/3)) = 1.328323 */ + .quad 0x3FF570C04260716B /* HI((2^1*(1+6/32+1/64))^(1/3)) = 1.340027 */ + .quad 0x3FF59FDDF7A45F38 /* HI((2^1*(1+7/32+1/64))^(1/3)) = 1.35153 */ + .quad 0x3FF5CE31C83539DF /* HI((2^1*(1+8/32+1/64))^(1/3)) = 1.36284 */ + .quad 0x3FF5FBC3F20966A4 /* HI((2^1*(1+9/32+1/64))^(1/3)) = 1.373966 */ + .quad 0x3FF6289C2C8F1B70 /* HI((2^1*(1+10/32+1/64))^(1/3)) = 1.384915 */ + .quad 0x3FF654C1B4316DCF /* HI((2^1*(1+11/32+1/64))^(1/3)) = 1.395693 */ + .quad 0x3FF6803B54A34E44 /* HI((2^1*(1+12/32+1/64))^(1/3)) = 1.406307 */ + .quad 0x3FF6AB0F72182659 /* HI((2^1*(1+13/32+1/64))^(1/3)) = 1.416763 */ + .quad 0x3FF6D544118C08BC /* HI((2^1*(1+14/32+1/64))^(1/3)) = 1.427067 */ + .quad 0x3FF6FEDEE0388D4A /* HI((2^1*(1+15/32+1/64))^(1/3)) = 1.437224 */ + .quad 0x3FF727E53A4F645E /* HI((2^1*(1+16/32+1/64))^(1/3)) = 1.44724 */ + .quad 0x3FF7505C31104114 /* HI((2^1*(1+17/32+1/64))^(1/3)) = 1.457119 */ + .quad 0x3FF77848904CD549 /* HI((2^1*(1+18/32+1/64))^(1/3)) = 1.466866 */ + .quad 0x3FF79FAEE36B2534 /* HI((2^1*(1+19/32+1/64))^(1/3)) = 1.476485 */ + .quad 0x3FF7C69379F4605B /* HI((2^1*(1+20/32+1/64))^(1/3)) = 1.48598 */ + .quad 0x3FF7ECFA6BBCA391 /* HI((2^1*(1+21/32+1/64))^(1/3)) = 1.495356 */ + .quad 0x3FF812E79CAE7EB9 /* HI((2^1*(1+22/32+1/64))^(1/3)) = 1.504615 */ + .quad 0x3FF8385EC043C71D /* HI((2^1*(1+23/32+1/64))^(1/3)) = 1.513762 */ + .quad 0x3FF85D635CB41B9D /* HI((2^1*(1+24/32+1/64))^(1/3)) = 1.5228 */ + .quad 0x3FF881F8CDE083DB /* HI((2^1*(1+25/32+1/64))^(1/3)) = 1.531731 */ + .quad 0x3FF8A6224802B8A8 /* HI((2^1*(1+26/32+1/64))^(1/3)) = 1.54056 */ + .quad 0x3FF8C9E2DA25E5E4 /* HI((2^1*(1+27/32+1/64))^(1/3)) = 1.549289 */ + .quad 0x3FF8ED3D706E1010 /* HI((2^1*(1+28/32+1/64))^(1/3)) = 1.55792 */ + .quad 0x3FF91034D632B6DF /* HI((2^1*(1+29/32+1/64))^(1/3)) = 1.566457 */ + .quad 0x3FF932CBB7F0CF2D /* HI((2^1*(1+30/32+1/64))^(1/3)) = 1.574901 */ + .quad 0x3FF95504A517BF3A /* HI((2^1*(1+31/32+1/64))^(1/3)) = 1.583256 */ + .quad 0x3FF987AF34F8BB19 /* HI((2^2*(1+0/32+1/64))^(1/3)) = 1.595626 */ + .quad 0x3FF9CA0A8337B317 /* HI((2^2*(1+1/32+1/64))^(1/3)) = 1.611826 */ + .quad 0x3FFA0B1709CC13D5 /* HI((2^2*(1+2/32+1/64))^(1/3)) = 1.627708 */ + .quad 0x3FFA4AE4CE6419ED /* HI((2^2*(1+3/32+1/64))^(1/3)) = 1.643285 */ + .quad 0x3FFA8982A5567031 /* HI((2^2*(1+4/32+1/64))^(1/3)) = 1.658572 */ + .quad 0x3FFAC6FE500AB570 /* HI((2^2*(1+5/32+1/64))^(1/3)) = 1.673582 */ + .quad 0x3FFB036497A15A17 /* HI((2^2*(1+6/32+1/64))^(1/3)) = 1.688328 */ + .quad 0x3FFB3EC164671755 /* HI((2^2*(1+7/32+1/64))^(1/3)) = 1.702821 */ + .quad 0x3FFB791FD288C46F /* HI((2^2*(1+8/32+1/64))^(1/3)) = 1.717071 */ + .quad 0x3FFBB28A44693BE4 /* HI((2^2*(1+9/32+1/64))^(1/3)) = 1.731089 */ + .quad 0x3FFBEB0A72EB6E31 /* HI((2^2*(1+10/32+1/64))^(1/3)) = 1.744883 */ + .quad 0x3FFC22A97BF5F697 /* HI((2^2*(1+11/32+1/64))^(1/3)) = 1.758462 */ + .quad 0x3FFC596FEF6AF983 /* HI((2^2*(1+12/32+1/64))^(1/3)) = 1.771835 */ + .quad 0x3FFC8F65DAC655A3 /* HI((2^2*(1+13/32+1/64))^(1/3)) = 1.785009 */ + .quad 0x3FFCC492D38CE8D9 /* HI((2^2*(1+14/32+1/64))^(1/3)) = 1.797992 */ + .quad 0x3FFCF8FE00B19367 /* HI((2^2*(1+15/32+1/64))^(1/3)) = 1.810789 */ + .quad 0x3FFD2CAE230F8709 /* HI((2^2*(1+16/32+1/64))^(1/3)) = 1.823408 */ + .quad 0x3FFD5FA99D15208F /* HI((2^2*(1+17/32+1/64))^(1/3)) = 1.835855 */ + .quad 0x3FFD91F679B6E505 /* HI((2^2*(1+18/32+1/64))^(1/3)) = 1.848135 */ + .quad 0x3FFDC39A72BF2302 /* HI((2^2*(1+19/32+1/64))^(1/3)) = 1.860255 */ + .quad 0x3FFDF49AF68C1570 /* HI((2^2*(1+20/32+1/64))^(1/3)) = 1.872218 */ + .quad 0x3FFE24FD2D4C23B8 /* HI((2^2*(1+21/32+1/64))^(1/3)) = 1.884031 */ + .quad 0x3FFE54C5FDC5EC73 /* HI((2^2*(1+22/32+1/64))^(1/3)) = 1.895697 */ + .quad 0x3FFE83FA11B81DBB /* HI((2^2*(1+23/32+1/64))^(1/3)) = 1.907221 */ + .quad 0x3FFEB29DD9DBAF25 /* HI((2^2*(1+24/32+1/64))^(1/3)) = 1.918608 */ + .quad 0x3FFEE0B59191D374 /* HI((2^2*(1+25/32+1/64))^(1/3)) = 1.929861 */ + .quad 0x3FFF0E454245E4BF /* HI((2^2*(1+26/32+1/64))^(1/3)) = 1.940984 */ + .quad 0x3FFF3B50C68A9DD3 /* HI((2^2*(1+27/32+1/64))^(1/3)) = 1.951981 */ + .quad 0x3FFF67DBCCF922DC /* HI((2^2*(1+28/32+1/64))^(1/3)) = 1.962856 */ + .quad 0x3FFF93E9DAD7A4A6 /* HI((2^2*(1+29/32+1/64))^(1/3)) = 1.973612 */ + .quad 0x3FFFBF7E4E8CC9CB /* HI((2^2*(1+30/32+1/64))^(1/3)) = 1.984251 */ + .quad 0x3FFFEA9C61E47CD3 /* HI((2^2*(1+31/32+1/64))^(1/3)) = 1.994778 */ + .align 16 + .quad 0x3F93750AD588F115, 0x3F93750AD588F115 /* _dA7 */ + .align 16 + .quad 0xBF98090D6221A247, 0xBF98090D6221A247 /* _dA6 */ + .align 16 + .quad 0x3F9EE7113506AC12, 0x3F9EE7113506AC12 /* _dA5 */ + .align 16 + .quad 0xBFA511E8D2B3183B, 0xBFA511E8D2B3183B /* _dA4 */ + .align 16 + .quad 0x3FAF9ADD3C0CA458, 0x3FAF9ADD3C0CA458 /* _dA3 */ + .align 16 + .quad 0xBFBC71C71C71C71C, 0xBFBC71C71C71C71C /* _dA2 */ + .align 16 + .quad 0x3FD5555555555555, 0x3FD5555555555555 /* _dA1 */ + .align 16 + .quad 0xBFF0400000000000, 0xBFF0400000000000 /* _dNeg65Div64 */ + .align 16 + .quad 0x000FC00000000000, 0x000FC00000000000 /* _dSgnf6Mask */ + .align 16 + .quad 0xBFF0000000000000, 0xBFF0000000000000 /* _dNegOne */ + .align 16 + .quad 0x000FFFFFFFFFFFFF, 0x000FFFFFFFFFFFFF /* _dMantissaMask */ + .align 16 + .quad 0xFFF0000000000000, 0xFFF0000000000000 /* _lExpHiMask */ + .align 16 + .quad 0x00000000000007FF, 0x00000000000007FF /* _lExpLoMask */ + .align 16 + .quad 0x0000000000001556, 0x0000000000001556 /* _l1556 */ + .align 16 + .long 0x000F8000, 0x000F8000, 0x000F8000, 0x000F8000 /* _iRcpIndexMask */ + .align 16 + .long 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF /* _iAbsMask */ + .align 16 + .long 0x00000800, 0x00000800, 0x00000800, 0x00000800 /* _iSignMask */ + .align 16 + .long 0x000002AA, 0x000002AA, 0x000002AA, 0x000002AA /* _iBias */ + .align 16 + .long 0x80100000, 0x80100000, 0x80100000, 0x80100000 /* _iSub */ + .align 16 + .long 0xffdfffff, 0xffdfffff, 0xffdfffff, 0xffdfffff /* _iCmp */ + .align 16 + .type __svml_dcbrt_data_internal,@object + .size __svml_dcbrt_data_internal,.-__svml_dcbrt_data_internal diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core-sse.S b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core-sse.S new file mode 100644 index 0000000000..3b54f31fbc --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core-sse.S @@ -0,0 +1,20 @@ +/* SSE version of vectorized cbrt, 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_cbrt _ZGVdN4v_cbrt_sse_wrapper +#include "../svml_d_cbrt4_core.S" diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core.c b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core.c new file mode 100644 index 0000000000..0b135877aa --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core.c @@ -0,0 +1,27 @@ +/* Multiple versions of vectorized cbrt, 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_cbrt +#include "ifunc-mathvec-avx2.h" + +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); + +#ifdef SHARED +__hidden_ver1 (_ZGVdN4v_cbrt, __GI__ZGVdN4v_cbrt, __redirect__ZGVdN4v_cbrt) + __attribute__ ((visibility ("hidden"))); +#endif diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core_avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core_avx2.S new file mode 100644 index 0000000000..2223c5309f --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt4_core_avx2.S @@ -0,0 +1,505 @@ +/* Function cbrt 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: + * + * x=2^{3*k+j} * 1.b1 b2 ... b5 b6 ... b52 + * Let r=(x*2^{-3k-j} - 1.b1 b2 ... b5 1)* rcp[b1 b2 ..b5], + * where rcp[b1 b2 .. b5]=1/(1.b1 b2 b3 b4 b5 1) in double precision + * cbrt(2^j * 1. b1 b2 .. b5 1) is approximated as T[j][b1..b5]+D[j][b1..b5] + * (T stores the high 53 bits, D stores the low order bits) + * Result=2^k*T+(2^k*T*r)*P+2^k*D + * where P=p1+p2*r+..+p8*r^7 + * + */ + +/* Offsets for data table __svml_dcbrt_data_internal + */ +#define _dRcp 0 +#define _dCbrtHiLo 256 +#define _dA7 1024 +#define _dA6 1056 +#define _dA5 1088 +#define _dA4 1120 +#define _dA3 1152 +#define _dA2 1184 +#define _dA1 1216 +#define _dNeg65Div64 1248 +#define _dSgnf6Mask 1280 +#define _dNegOne 1312 +#define _dMantissaMask 1344 +#define _lExpHiMask 1376 +#define _lExpLoMask 1408 +#define _l1556 1440 +#define _iRcpIndexMask 1472 +#define _iAbsMask 1504 +#define _iSignMask 1536 +#define _iBias 1568 +#define _iSub 1600 +#define _iCmp 1632 + +#include + + .text + .section .text.avx2,"ax",@progbits +ENTRY(_ZGVdN4v_cbrt_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 + +/* Load 1/(1+iRcpIndex/32+1/64) reciprocal table value */ + lea __svml_dcbrt_data_internal(%rip), %rax + vmovapd %ymm0, %ymm5 + +/* + * Declarations + * Load constants + * Get iX - high part of argument + */ + vextractf128 $1, %ymm5, %xmm6 + +/* Calculate CbrtIndex */ + vpsrlq $52, %ymm5, %ymm15 + vshufps $221, %xmm6, %xmm5, %xmm4 + +/* Calculate Rcp table index */ + vandps _iRcpIndexMask+__svml_dcbrt_data_internal(%rip), %xmm4, %xmm10 + vpsrld $12, %xmm10, %xmm3 + vmovd %xmm3, %ecx + +/* If the exponent field is zero - go to callout to process denormals */ + vandps _iAbsMask+__svml_dcbrt_data_internal(%rip), %xmm4, %xmm7 + +/* Compute 2^k */ + vpsrld $20, %xmm4, %xmm4 + vpsubd _iSub+__svml_dcbrt_data_internal(%rip), %xmm7, %xmm8 + vandps _lExpLoMask+__svml_dcbrt_data_internal(%rip), %ymm15, %ymm0 + vpmuludq _l1556+__svml_dcbrt_data_internal(%rip), %ymm0, %ymm6 + vpextrd $2, %xmm3, %edi + movslq %ecx, %rcx + vpextrd $1, %xmm3, %esi + movslq %edi, %rdi + vpextrd $3, %xmm3, %r8d + movslq %esi, %rsi + movslq %r8d, %r8 + vpcmpgtd _iCmp+__svml_dcbrt_data_internal(%rip), %xmm8, %xmm9 + vmovsd (%rax,%rcx), %xmm11 + vmovmskps %xmm9, %edx + vmovsd (%rax,%rdi), %xmm13 + vmovhpd (%rax,%rsi), %xmm11, %xmm12 + vmovhpd (%rax,%r8), %xmm13, %xmm14 + vextractf128 $1, %ymm6, %xmm7 + vshufps $136, %xmm7, %xmm6, %xmm8 + vmovups __VUNPACK_ODD_ind1.613.0.1(%rip), %ymm7 + vextractf128 $1, %ymm0, %xmm1 + vshufps $136, %xmm1, %xmm0, %xmm9 + vpsrld $14, %xmm8, %xmm1 + vpsubd %xmm1, %xmm9, %xmm10 + vpaddd %xmm1, %xmm1, %xmm11 + +/* + * VAND( L, l2k, = l2k, lExpHiMask ); + * Argument reduction Z + */ + vandpd _dMantissaMask+__svml_dcbrt_data_internal(%rip), %ymm5, %ymm9 + vinsertf128 $1, %xmm14, %ymm12, %ymm2 + vpsubd %xmm11, %xmm10, %xmm12 + vpslld $8, %xmm12, %xmm13 + vpaddd %xmm13, %xmm3, %xmm15 + +/* Load cbrt(2^j*(1+iRcpIndex/32+1/64)) Hi & Lo values */ + vmovd %xmm15, %r9d + vpextrd $2, %xmm15, %r11d + movslq %r9d, %r9 + vpextrd $1, %xmm15, %r10d + movslq %r11d, %r11 + vpextrd $3, %xmm15, %ecx + movslq %r10d, %r10 + movslq %ecx, %rcx + vmovsd 256(%rax,%r9), %xmm3 + vmovsd 256(%rax,%r11), %xmm0 + vandpd _dSgnf6Mask+__svml_dcbrt_data_internal(%rip), %ymm5, %ymm10 + vmovhpd 256(%rax,%r10), %xmm3, %xmm14 + vmovhpd 256(%rax,%rcx), %xmm0, %xmm3 + vorpd _dNegOne+__svml_dcbrt_data_internal(%rip), %ymm9, %ymm11 + vorpd _dNeg65Div64+__svml_dcbrt_data_internal(%rip), %ymm10, %ymm12 + vsubpd %ymm12, %ymm11, %ymm13 + vmulpd %ymm13, %ymm2, %ymm2 + vinsertf128 $1, %xmm3, %ymm14, %ymm0 + vpand _iSignMask+__svml_dcbrt_data_internal(%rip), %xmm4, %xmm3 + vpor _iBias+__svml_dcbrt_data_internal(%rip), %xmm3, %xmm4 + vpaddd %xmm1, %xmm4, %xmm1 + vpslld $20, %xmm1, %xmm6 + +/* Polynomial */ + vmovupd _dA7+__svml_dcbrt_data_internal(%rip), %ymm1 + vfmadd213pd _dA6+__svml_dcbrt_data_internal(%rip), %ymm2, %ymm1 + vfmadd213pd _dA5+__svml_dcbrt_data_internal(%rip), %ymm2, %ymm1 + vfmadd213pd _dA4+__svml_dcbrt_data_internal(%rip), %ymm2, %ymm1 + vfmadd213pd _dA3+__svml_dcbrt_data_internal(%rip), %ymm2, %ymm1 + vfmadd213pd _dA2+__svml_dcbrt_data_internal(%rip), %ymm2, %ymm1 + vfmadd213pd _dA1+__svml_dcbrt_data_internal(%rip), %ymm2, %ymm1 + vpermps %ymm6, %ymm7, %ymm8 + vandps __VUNPACK_ODD_mask.613.0.1(%rip), %ymm8, %ymm14 + +/* THi*2^k, TLo*2^k */ + vmulpd %ymm14, %ymm0, %ymm0 + +/* THi*2^k*Z */ + vmulpd %ymm0, %ymm2, %ymm2 + +/* Final reconstruction */ + vmulpd %ymm2, %ymm1, %ymm3 + vaddpd %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 cbrt@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_cbrt_avx2) + .section .rodata, "a" + .align 32 + +__VUNPACK_ODD_ind1.613.0.1: + .rept 3 + .long 0 + .endr + .long 1 + .long 0 + .long 2 + .long 0 + .long 3 + .align 32 + +__VUNPACK_ODD_mask.613.0.1: + .long 0 + .long -1 + .long 0 + .long -1 + .long 0 + .long -1 + .long 0 + .long -1 + + .section .rodata, "a" + .align 32 + +#ifdef __svml_dcbrt_data_internal_typedef +typedef unsigned int VUINT32; +typedef struct +{ + __declspec(align(32)) VUINT32 _dRcp[32][2]; + __declspec(align(32)) VUINT32 _dCbrtHiLo[96][2]; + __declspec(align(32)) VUINT32 _dA7[4][2]; + __declspec(align(32)) VUINT32 _dA6[4][2]; + __declspec(align(32)) VUINT32 _dA5[4][2]; + __declspec(align(32)) VUINT32 _dA4[4][2]; + __declspec(align(32)) VUINT32 _dA3[4][2]; + __declspec(align(32)) VUINT32 _dA2[4][2]; + __declspec(align(32)) VUINT32 _dA1[4][2]; + __declspec(align(32)) VUINT32 _dNeg65Div64[4][2]; + __declspec(align(32)) VUINT32 _dSgnf6Mask[4][2]; + __declspec(align(32)) VUINT32 _dNegOne[4][2]; + __declspec(align(32)) VUINT32 _dMantissaMask[4][2]; + __declspec(align(32)) VUINT32 _lExpHiMask[4][2]; + __declspec(align(32)) VUINT32 _lExpLoMask[4][2]; + __declspec(align(32)) VUINT32 _l1556[4][2]; + __declspec(align(32)) VUINT32 _iRcpIndexMask[8][1]; + __declspec(align(32)) VUINT32 _iAbsMask[8][1]; + __declspec(align(32)) VUINT32 _iSignMask[8][1]; + __declspec(align(32)) VUINT32 _iBias[8][1]; + __declspec(align(32)) VUINT32 _iSub[8][1]; + __declspec(align(32)) VUINT32 _iCmp[8][1]; +} __svml_dcbrt_data_internal; +#endif +__svml_dcbrt_data_internal: + /*== _dRcp ==*/ + .quad 0xBFEF81F81F81F820 /* (1/(1+0/32+1/64)) = -.984615 */ + .quad 0xBFEE9131ABF0B767 /* (1/(1+1/32+1/64)) = -.955224 */ + .quad 0xBFEDAE6076B981DB /* (1/(1+2/32+1/64)) = -.927536 */ + .quad 0xBFECD85689039B0B /* (1/(1+3/32+1/64)) = -.901408 */ + .quad 0xBFEC0E070381C0E0 /* (1/(1+4/32+1/64)) = -.876712 */ + .quad 0xBFEB4E81B4E81B4F /* (1/(1+5/32+1/64)) = -.853333 */ + .quad 0xBFEA98EF606A63BE /* (1/(1+6/32+1/64)) = -.831169 */ + .quad 0xBFE9EC8E951033D9 /* (1/(1+7/32+1/64)) = -.810127 */ + .quad 0xBFE948B0FCD6E9E0 /* (1/(1+8/32+1/64)) = -.790123 */ + .quad 0xBFE8ACB90F6BF3AA /* (1/(1+9/32+1/64)) = -.771084 */ + .quad 0xBFE8181818181818 /* (1/(1+10/32+1/64)) = -.752941 */ + .quad 0xBFE78A4C8178A4C8 /* (1/(1+11/32+1/64)) = -.735632 */ + .quad 0xBFE702E05C0B8170 /* (1/(1+12/32+1/64)) = -.719101 */ + .quad 0xBFE6816816816817 /* (1/(1+13/32+1/64)) = -.703297 */ + .quad 0xBFE6058160581606 /* (1/(1+14/32+1/64)) = -.688172 */ + .quad 0xBFE58ED2308158ED /* (1/(1+15/32+1/64)) = -.673684 */ + .quad 0xBFE51D07EAE2F815 /* (1/(1+16/32+1/64)) = -.659794 */ + .quad 0xBFE4AFD6A052BF5B /* (1/(1+17/32+1/64)) = -.646465 */ + .quad 0xBFE446F86562D9FB /* (1/(1+18/32+1/64)) = -.633663 */ + .quad 0xBFE3E22CBCE4A902 /* (1/(1+19/32+1/64)) = -.621359 */ + .quad 0xBFE3813813813814 /* (1/(1+20/32+1/64)) = -.609524 */ + .quad 0xBFE323E34A2B10BF /* (1/(1+21/32+1/64)) = -.598131 */ + .quad 0xBFE2C9FB4D812CA0 /* (1/(1+22/32+1/64)) = -.587156 */ + .quad 0xBFE27350B8812735 /* (1/(1+23/32+1/64)) = -.576577 */ + .quad 0xBFE21FB78121FB78 /* (1/(1+24/32+1/64)) = -.566372 */ + .quad 0xBFE1CF06ADA2811D /* (1/(1+25/32+1/64)) = -.556522 */ + .quad 0xBFE1811811811812 /* (1/(1+26/32+1/64)) = -.547009 */ + .quad 0xBFE135C81135C811 /* (1/(1+27/32+1/64)) = -.537815 */ + .quad 0xBFE0ECF56BE69C90 /* (1/(1+28/32+1/64)) = -.528926 */ + .quad 0xBFE0A6810A6810A7 /* (1/(1+29/32+1/64)) = -.520325 */ + .quad 0xBFE0624DD2F1A9FC /* (1/(1+30/32+1/64)) = -.512 */ + .quad 0xBFE0204081020408 /* (1/(1+31/32+1/64)) = -.503937 */ + /*== _dCbrtHiLo ==*/ + .align 32 + .quad 0x3FF01539221D4C97 /* HI((2^0*(1+0/32+1/64))^(1/3)) = 1.005181 */ + .quad 0x3FF03F06771A2E33 /* HI((2^0*(1+1/32+1/64))^(1/3)) = 1.015387 */ + .quad 0x3FF06800E629D671 /* HI((2^0*(1+2/32+1/64))^(1/3)) = 1.025391 */ + .quad 0x3FF090328731DEB2 /* HI((2^0*(1+3/32+1/64))^(1/3)) = 1.035204 */ + .quad 0x3FF0B7A4B1BD64AC /* HI((2^0*(1+4/32+1/64))^(1/3)) = 1.044835 */ + .quad 0x3FF0DE601024FB87 /* HI((2^0*(1+5/32+1/64))^(1/3)) = 1.054291 */ + .quad 0x3FF1046CB0597000 /* HI((2^0*(1+6/32+1/64))^(1/3)) = 1.06358 */ + .quad 0x3FF129D212A9BA9B /* HI((2^0*(1+7/32+1/64))^(1/3)) = 1.07271 */ + .quad 0x3FF14E9736CDAF38 /* HI((2^0*(1+8/32+1/64))^(1/3)) = 1.081687 */ + .quad 0x3FF172C2A772F507 /* HI((2^0*(1+9/32+1/64))^(1/3)) = 1.090518 */ + .quad 0x3FF1965A848001D3 /* HI((2^0*(1+10/32+1/64))^(1/3)) = 1.099207 */ + .quad 0x3FF1B9648C38C55D /* HI((2^0*(1+11/32+1/64))^(1/3)) = 1.107762 */ + .quad 0x3FF1DBE6236A0C45 /* HI((2^0*(1+12/32+1/64))^(1/3)) = 1.116186 */ + .quad 0x3FF1FDE45CBB1F9F /* HI((2^0*(1+13/32+1/64))^(1/3)) = 1.124485 */ + .quad 0x3FF21F63FF409042 /* HI((2^0*(1+14/32+1/64))^(1/3)) = 1.132664 */ + .quad 0x3FF240698C6746E5 /* HI((2^0*(1+15/32+1/64))^(1/3)) = 1.140726 */ + .quad 0x3FF260F9454BB99B /* HI((2^0*(1+16/32+1/64))^(1/3)) = 1.148675 */ + .quad 0x3FF281172F8E7073 /* HI((2^0*(1+17/32+1/64))^(1/3)) = 1.156516 */ + .quad 0x3FF2A0C719B4B6D0 /* HI((2^0*(1+18/32+1/64))^(1/3)) = 1.164252 */ + .quad 0x3FF2C00C9F2263EC /* HI((2^0*(1+19/32+1/64))^(1/3)) = 1.171887 */ + .quad 0x3FF2DEEB2BB7FB78 /* HI((2^0*(1+20/32+1/64))^(1/3)) = 1.179423 */ + .quad 0x3FF2FD65FF1EFBBC /* HI((2^0*(1+21/32+1/64))^(1/3)) = 1.186865 */ + .quad 0x3FF31B802FCCF6A2 /* HI((2^0*(1+22/32+1/64))^(1/3)) = 1.194214 */ + .quad 0x3FF3393CADC50708 /* HI((2^0*(1+23/32+1/64))^(1/3)) = 1.201474 */ + .quad 0x3FF3569E451E4C2A /* HI((2^0*(1+24/32+1/64))^(1/3)) = 1.208647 */ + .quad 0x3FF373A7A0554CDE /* HI((2^0*(1+25/32+1/64))^(1/3)) = 1.215736 */ + .quad 0x3FF3905B4A6D76CE /* HI((2^0*(1+26/32+1/64))^(1/3)) = 1.222743 */ + .quad 0x3FF3ACBBB0E756B6 /* HI((2^0*(1+27/32+1/64))^(1/3)) = 1.229671 */ + .quad 0x3FF3C8CB258FA340 /* HI((2^0*(1+28/32+1/64))^(1/3)) = 1.236522 */ + .quad 0x3FF3E48BE02AC0CE /* HI((2^0*(1+29/32+1/64))^(1/3)) = 1.243297 */ + .quad 0x3FF4000000000000 /* HI((2^0*(1+30/32+1/64))^(1/3)) = 1.25 */ + .quad 0x3FF41B298D47800E /* HI((2^0*(1+31/32+1/64))^(1/3)) = 1.256631 */ + .quad 0x3FF443604B34D9B2 /* HI((2^1*(1+0/32+1/64))^(1/3)) = 1.266449 */ + .quad 0x3FF4780B20906571 /* HI((2^1*(1+1/32+1/64))^(1/3)) = 1.279307 */ + .quad 0x3FF4ABAC3EE06706 /* HI((2^1*(1+2/32+1/64))^(1/3)) = 1.291912 */ + .quad 0x3FF4DE505DA66B8D /* HI((2^1*(1+3/32+1/64))^(1/3)) = 1.304276 */ + .quad 0x3FF51003420A5C07 /* HI((2^1*(1+4/32+1/64))^(1/3)) = 1.316409 */ + .quad 0x3FF540CFD6FD11C1 /* HI((2^1*(1+5/32+1/64))^(1/3)) = 1.328323 */ + .quad 0x3FF570C04260716B /* HI((2^1*(1+6/32+1/64))^(1/3)) = 1.340027 */ + .quad 0x3FF59FDDF7A45F38 /* HI((2^1*(1+7/32+1/64))^(1/3)) = 1.35153 */ + .quad 0x3FF5CE31C83539DF /* HI((2^1*(1+8/32+1/64))^(1/3)) = 1.36284 */ + .quad 0x3FF5FBC3F20966A4 /* HI((2^1*(1+9/32+1/64))^(1/3)) = 1.373966 */ + .quad 0x3FF6289C2C8F1B70 /* HI((2^1*(1+10/32+1/64))^(1/3)) = 1.384915 */ + .quad 0x3FF654C1B4316DCF /* HI((2^1*(1+11/32+1/64))^(1/3)) = 1.395693 */ + .quad 0x3FF6803B54A34E44 /* HI((2^1*(1+12/32+1/64))^(1/3)) = 1.406307 */ + .quad 0x3FF6AB0F72182659 /* HI((2^1*(1+13/32+1/64))^(1/3)) = 1.416763 */ + .quad 0x3FF6D544118C08BC /* HI((2^1*(1+14/32+1/64))^(1/3)) = 1.427067 */ + .quad 0x3FF6FEDEE0388D4A /* HI((2^1*(1+15/32+1/64))^(1/3)) = 1.437224 */ + .quad 0x3FF727E53A4F645E /* HI((2^1*(1+16/32+1/64))^(1/3)) = 1.44724 */ + .quad 0x3FF7505C31104114 /* HI((2^1*(1+17/32+1/64))^(1/3)) = 1.457119 */ + .quad 0x3FF77848904CD549 /* HI((2^1*(1+18/32+1/64))^(1/3)) = 1.466866 */ + .quad 0x3FF79FAEE36B2534 /* HI((2^1*(1+19/32+1/64))^(1/3)) = 1.476485 */ + .quad 0x3FF7C69379F4605B /* HI((2^1*(1+20/32+1/64))^(1/3)) = 1.48598 */ + .quad 0x3FF7ECFA6BBCA391 /* HI((2^1*(1+21/32+1/64))^(1/3)) = 1.495356 */ + .quad 0x3FF812E79CAE7EB9 /* HI((2^1*(1+22/32+1/64))^(1/3)) = 1.504615 */ + .quad 0x3FF8385EC043C71D /* HI((2^1*(1+23/32+1/64))^(1/3)) = 1.513762 */ + .quad 0x3FF85D635CB41B9D /* HI((2^1*(1+24/32+1/64))^(1/3)) = 1.5228 */ + .quad 0x3FF881F8CDE083DB /* HI((2^1*(1+25/32+1/64))^(1/3)) = 1.531731 */ + .quad 0x3FF8A6224802B8A8 /* HI((2^1*(1+26/32+1/64))^(1/3)) = 1.54056 */ + .quad 0x3FF8C9E2DA25E5E4 /* HI((2^1*(1+27/32+1/64))^(1/3)) = 1.549289 */ + .quad 0x3FF8ED3D706E1010 /* HI((2^1*(1+28/32+1/64))^(1/3)) = 1.55792 */ + .quad 0x3FF91034D632B6DF /* HI((2^1*(1+29/32+1/64))^(1/3)) = 1.566457 */ + .quad 0x3FF932CBB7F0CF2D /* HI((2^1*(1+30/32+1/64))^(1/3)) = 1.574901 */ + .quad 0x3FF95504A517BF3A /* HI((2^1*(1+31/32+1/64))^(1/3)) = 1.583256 */ + .quad 0x3FF987AF34F8BB19 /* HI((2^2*(1+0/32+1/64))^(1/3)) = 1.595626 */ + .quad 0x3FF9CA0A8337B317 /* HI((2^2*(1+1/32+1/64))^(1/3)) = 1.611826 */ + .quad 0x3FFA0B1709CC13D5 /* HI((2^2*(1+2/32+1/64))^(1/3)) = 1.627708 */ + .quad 0x3FFA4AE4CE6419ED /* HI((2^2*(1+3/32+1/64))^(1/3)) = 1.643285 */ + .quad 0x3FFA8982A5567031 /* HI((2^2*(1+4/32+1/64))^(1/3)) = 1.658572 */ + .quad 0x3FFAC6FE500AB570 /* HI((2^2*(1+5/32+1/64))^(1/3)) = 1.673582 */ + .quad 0x3FFB036497A15A17 /* HI((2^2*(1+6/32+1/64))^(1/3)) = 1.688328 */ + .quad 0x3FFB3EC164671755 /* HI((2^2*(1+7/32+1/64))^(1/3)) = 1.702821 */ + .quad 0x3FFB791FD288C46F /* HI((2^2*(1+8/32+1/64))^(1/3)) = 1.717071 */ + .quad 0x3FFBB28A44693BE4 /* HI((2^2*(1+9/32+1/64))^(1/3)) = 1.731089 */ + .quad 0x3FFBEB0A72EB6E31 /* HI((2^2*(1+10/32+1/64))^(1/3)) = 1.744883 */ + .quad 0x3FFC22A97BF5F697 /* HI((2^2*(1+11/32+1/64))^(1/3)) = 1.758462 */ + .quad 0x3FFC596FEF6AF983 /* HI((2^2*(1+12/32+1/64))^(1/3)) = 1.771835 */ + .quad 0x3FFC8F65DAC655A3 /* HI((2^2*(1+13/32+1/64))^(1/3)) = 1.785009 */ + .quad 0x3FFCC492D38CE8D9 /* HI((2^2*(1+14/32+1/64))^(1/3)) = 1.797992 */ + .quad 0x3FFCF8FE00B19367 /* HI((2^2*(1+15/32+1/64))^(1/3)) = 1.810789 */ + .quad 0x3FFD2CAE230F8709 /* HI((2^2*(1+16/32+1/64))^(1/3)) = 1.823408 */ + .quad 0x3FFD5FA99D15208F /* HI((2^2*(1+17/32+1/64))^(1/3)) = 1.835855 */ + .quad 0x3FFD91F679B6E505 /* HI((2^2*(1+18/32+1/64))^(1/3)) = 1.848135 */ + .quad 0x3FFDC39A72BF2302 /* HI((2^2*(1+19/32+1/64))^(1/3)) = 1.860255 */ + .quad 0x3FFDF49AF68C1570 /* HI((2^2*(1+20/32+1/64))^(1/3)) = 1.872218 */ + .quad 0x3FFE24FD2D4C23B8 /* HI((2^2*(1+21/32+1/64))^(1/3)) = 1.884031 */ + .quad 0x3FFE54C5FDC5EC73 /* HI((2^2*(1+22/32+1/64))^(1/3)) = 1.895697 */ + .quad 0x3FFE83FA11B81DBB /* HI((2^2*(1+23/32+1/64))^(1/3)) = 1.907221 */ + .quad 0x3FFEB29DD9DBAF25 /* HI((2^2*(1+24/32+1/64))^(1/3)) = 1.918608 */ + .quad 0x3FFEE0B59191D374 /* HI((2^2*(1+25/32+1/64))^(1/3)) = 1.929861 */ + .quad 0x3FFF0E454245E4BF /* HI((2^2*(1+26/32+1/64))^(1/3)) = 1.940984 */ + .quad 0x3FFF3B50C68A9DD3 /* HI((2^2*(1+27/32+1/64))^(1/3)) = 1.951981 */ + .quad 0x3FFF67DBCCF922DC /* HI((2^2*(1+28/32+1/64))^(1/3)) = 1.962856 */ + .quad 0x3FFF93E9DAD7A4A6 /* HI((2^2*(1+29/32+1/64))^(1/3)) = 1.973612 */ + .quad 0x3FFFBF7E4E8CC9CB /* HI((2^2*(1+30/32+1/64))^(1/3)) = 1.984251 */ + .quad 0x3FFFEA9C61E47CD3 /* HI((2^2*(1+31/32+1/64))^(1/3)) = 1.994778 */ + .align 32 + .quad 0x3F93750AD588F115, 0x3F93750AD588F115, 0x3F93750AD588F115, 0x3F93750AD588F115 /* _dA7 */ + .align 32 + .quad 0xBF98090D6221A247, 0xBF98090D6221A247, 0xBF98090D6221A247, 0xBF98090D6221A247 /* _dA6 */ + .align 32 + .quad 0x3F9EE7113506AC12, 0x3F9EE7113506AC12, 0x3F9EE7113506AC12, 0x3F9EE7113506AC12 /* _dA5 */ + .align 32 + .quad 0xBFA511E8D2B3183B, 0xBFA511E8D2B3183B, 0xBFA511E8D2B3183B, 0xBFA511E8D2B3183B /* _dA4 */ + .align 32 + .quad 0x3FAF9ADD3C0CA458, 0x3FAF9ADD3C0CA458, 0x3FAF9ADD3C0CA458, 0x3FAF9ADD3C0CA458 /* _dA3 */ + .align 32 + .quad 0xBFBC71C71C71C71C, 0xBFBC71C71C71C71C, 0xBFBC71C71C71C71C, 0xBFBC71C71C71C71C /* _dA2 */ + .align 32 + .quad 0x3FD5555555555555, 0x3FD5555555555555, 0x3FD5555555555555, 0x3FD5555555555555 /* _dA1 */ + .align 32 + .quad 0xBFF0400000000000, 0xBFF0400000000000, 0xBFF0400000000000, 0xBFF0400000000000 /* _dNeg65Div64 */ + .align 32 + .quad 0x000FC00000000000, 0x000FC00000000000, 0x000FC00000000000, 0x000FC00000000000 /* _dSgnf6Mask */ + .align 32 + .quad 0xBFF0000000000000, 0xBFF0000000000000, 0xBFF0000000000000, 0xBFF0000000000000 /* _dNegOne */ + .align 32 + .quad 0x000FFFFFFFFFFFFF, 0x000FFFFFFFFFFFFF, 0x000FFFFFFFFFFFFF, 0x000FFFFFFFFFFFFF /* _dMantissaMask */ + .align 32 + .quad 0xFFF0000000000000, 0xFFF0000000000000, 0xFFF0000000000000, 0xFFF0000000000000 /* _lExpHiMask */ + .align 32 + .quad 0x00000000000007FF, 0x00000000000007FF, 0x00000000000007FF, 0x00000000000007FF /* _lExpLoMask */ + .align 32 + .quad 0x0000000000001556, 0x0000000000001556, 0x0000000000001556, 0x0000000000001556 /* _l1556 */ + .align 32 + .long 0x000F8000, 0x000F8000, 0x000F8000, 0x000F8000, 0x000F8000, 0x000F8000, 0x000F8000, 0x000F8000 /* _iRcpIndexMask */ + .align 32 + .long 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF, 0x7FFFFFFF /* _iAbsMask */ + .align 32 + .long 0x00000800, 0x00000800, 0x00000800, 0x00000800, 0x00000800, 0x00000800, 0x00000800, 0x00000800 /* _iSignMask */ + .align 32 + .long 0x000002AA, 0x000002AA, 0x000002AA, 0x000002AA, 0x000002AA, 0x000002AA, 0x000002AA, 0x000002AA /* _iBias */ + .align 32 + .long 0x80100000, 0x80100000, 0x80100000, 0x80100000, 0x80100000, 0x80100000, 0x80100000, 0x80100000 /* _iSub */ + .align 32 + .long 0xffdfffff, 0xffdfffff, 0xffdfffff, 0xffdfffff, 0xffdfffff, 0xffdfffff, 0xffdfffff, 0xffdfffff /* _iCmp */ + .align 32 + .type __svml_dcbrt_data_internal,@object + .size __svml_dcbrt_data_internal,.-__svml_dcbrt_data_internal diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core-avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core-avx2.S new file mode 100644 index 0000000000..3831e582ce --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core-avx2.S @@ -0,0 +1,20 @@ +/* AVX2 version of vectorized cbrt, 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_cbrt _ZGVeN8v_cbrt_avx2_wrapper +#include "../svml_d_cbrt8_core.S" diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core.c b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core.c new file mode 100644 index 0000000000..28c147216f --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core.c @@ -0,0 +1,27 @@ +/* Multiple versions of vectorized cbrt, 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_cbrt +#include "ifunc-mathvec-avx512-skx.h" + +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); + +#ifdef SHARED +__hidden_ver1 (_ZGVeN8v_cbrt, __GI__ZGVeN8v_cbrt, __redirect__ZGVeN8v_cbrt) + __attribute__ ((visibility ("hidden"))); +#endif diff --git a/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core_avx512.S b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core_avx512.S new file mode 100644 index 0000000000..b9c071b54c --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_d_cbrt8_core_avx512.S @@ -0,0 +1,253 @@ +/* Function cbrt 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: + * + * x=2^{3*k+j} * 1.b1 b2 ... b5 b6 ... b52 + * Let r=(x*2^{-3k-j} - 1.b1 b2 ... b5 1)* rcp[b1 b2 ..b5], + * where rcp[b1 b2 .. b5]=1/(1.b1 b2 b3 b4 b5 1) in double precision + * cbrt(2^j * 1. b1 b2 .. b5 1) is approximated as T[j][b1..b5]+D[j][b1..b5] + * (T stores the high 53 bits, D stores the low order bits) + * Result=2^k*T+(2^k*T*r)*P+2^k*D + * where P=p1+p2*r+..+p8*r^7 + * + */ + +/* Offsets for data table __svml_dcbrt_data_internal_avx512 + */ +#define etbl_H 0 +#define etbl_L 64 +#define cbrt_tbl_H 128 +#define BiasL 256 +#define SZero 320 +#define OneThird 384 +#define Bias3 448 +#define Three 512 +#define One 576 +#define poly_coeff10 640 +#define poly_coeff9 704 +#define poly_coeff8 768 +#define poly_coeff7 832 +#define poly_coeff6 896 +#define poly_coeff5 960 +#define poly_coeff4 1024 +#define poly_coeff3 1088 +#define poly_coeff2 1152 +#define poly_coeff1 1216 + +#include + + .text + .section .text.evex512,"ax",@progbits +ENTRY(_ZGVeN8v_cbrt_skx) + vgetmantpd $0, {sae}, %zmm0, %zmm14 + +/* GetExp(x) */ + vgetexppd {sae}, %zmm0, %zmm7 + vmovups BiasL+__svml_dcbrt_data_internal_avx512(%rip), %zmm8 + +/* exponent/3 */ + vmovups OneThird+__svml_dcbrt_data_internal_avx512(%rip), %zmm9 + vmovups Bias3+__svml_dcbrt_data_internal_avx512(%rip), %zmm10 + +/* Reduced argument: R = DblRcp*Mantissa - 1 */ + vmovups One+__svml_dcbrt_data_internal_avx512(%rip), %zmm2 + +/* exponent%3 (to be used as index) */ + vmovups Three+__svml_dcbrt_data_internal_avx512(%rip), %zmm11 + +/* DblRcp ~ 1/Mantissa */ + vrcp14pd %zmm14, %zmm13 + vaddpd {rn-sae}, %zmm8, %zmm7, %zmm12 + vandpd SZero+__svml_dcbrt_data_internal_avx512(%rip), %zmm0, %zmm6 + +/* round DblRcp to 3 fractional bits (RN mode, no Precision exception) */ + vrndscalepd $72, {sae}, %zmm13, %zmm15 + vfmsub231pd {rn-sae}, %zmm12, %zmm9, %zmm10 + +/* polynomial */ + vmovups poly_coeff10+__svml_dcbrt_data_internal_avx512(%rip), %zmm0 + vmovups poly_coeff8+__svml_dcbrt_data_internal_avx512(%rip), %zmm7 + vmovups poly_coeff7+__svml_dcbrt_data_internal_avx512(%rip), %zmm9 + vfmsub231pd {rn-sae}, %zmm15, %zmm14, %zmm2 + vrndscalepd $9, {sae}, %zmm10, %zmm5 + +/* Table lookup */ + vmovups cbrt_tbl_H+__svml_dcbrt_data_internal_avx512(%rip), %zmm10 + vmovups poly_coeff6+__svml_dcbrt_data_internal_avx512(%rip), %zmm8 + vmovups poly_coeff3+__svml_dcbrt_data_internal_avx512(%rip), %zmm13 + vfmadd231pd {rn-sae}, %zmm2, %zmm7, %zmm9 + vfnmadd231pd {rn-sae}, %zmm5, %zmm11, %zmm12 + vmovups poly_coeff5+__svml_dcbrt_data_internal_avx512(%rip), %zmm11 + vmovups poly_coeff1+__svml_dcbrt_data_internal_avx512(%rip), %zmm14 + +/* Prepare table index */ + vpsrlq $49, %zmm15, %zmm1 + +/* Table lookup: 2^(exponent%3) */ + vpermpd __svml_dcbrt_data_internal_avx512(%rip), %zmm12, %zmm4 + vpermpd etbl_L+__svml_dcbrt_data_internal_avx512(%rip), %zmm12, %zmm3 + vpermt2pd cbrt_tbl_H+64+__svml_dcbrt_data_internal_avx512(%rip), %zmm1, %zmm10 + vmovups poly_coeff9+__svml_dcbrt_data_internal_avx512(%rip), %zmm1 + vfmadd231pd {rn-sae}, %zmm2, %zmm8, %zmm11 + vmovups poly_coeff2+__svml_dcbrt_data_internal_avx512(%rip), %zmm12 + vscalefpd {rn-sae}, %zmm5, %zmm10, %zmm15 + vfmadd231pd {rn-sae}, %zmm2, %zmm0, %zmm1 + vmovups poly_coeff4+__svml_dcbrt_data_internal_avx512(%rip), %zmm5 + vfmadd231pd {rn-sae}, %zmm2, %zmm12, %zmm14 + vmulpd {rn-sae}, %zmm2, %zmm2, %zmm0 + vfmadd231pd {rn-sae}, %zmm2, %zmm5, %zmm13 + +/* Sh*R */ + vmulpd {rn-sae}, %zmm2, %zmm4, %zmm2 + vfmadd213pd {rn-sae}, %zmm9, %zmm0, %zmm1 + vfmadd213pd {rn-sae}, %zmm11, %zmm0, %zmm1 + vfmadd213pd {rn-sae}, %zmm13, %zmm0, %zmm1 + vfmadd213pd {rn-sae}, %zmm14, %zmm0, %zmm1 + +/* Sl + (Sh*R)*Poly */ + vfmadd213pd {rn-sae}, %zmm3, %zmm1, %zmm2 + +/* + * branch-free + * scaled_Th*(Sh+Sl+Sh*R*Poly) + */ + vaddpd {rn-sae}, %zmm4, %zmm2, %zmm3 + vmulpd {rn-sae}, %zmm15, %zmm3, %zmm4 + vorpd %zmm6, %zmm4, %zmm0 + ret + +END(_ZGVeN8v_cbrt_skx) + + .section .rodata, "a" + .align 64 + +#ifdef __svml_dcbrt_data_internal_avx512_typedef +typedef unsigned int VUINT32; +typedef struct { + __declspec(align(64)) VUINT32 etbl_H[8][2]; + __declspec(align(64)) VUINT32 etbl_L[8][2]; + __declspec(align(64)) VUINT32 cbrt_tbl_H[16][2]; + __declspec(align(64)) VUINT32 BiasL[8][2]; + __declspec(align(64)) VUINT32 SZero[8][2]; + __declspec(align(64)) VUINT32 OneThird[8][2]; + __declspec(align(64)) VUINT32 Bias3[8][2]; + __declspec(align(64)) VUINT32 Three[8][2]; + __declspec(align(64)) VUINT32 One[8][2]; + __declspec(align(64)) VUINT32 poly_coeff10[8][2]; + __declspec(align(64)) VUINT32 poly_coeff9[8][2]; + __declspec(align(64)) VUINT32 poly_coeff8[8][2]; + __declspec(align(64)) VUINT32 poly_coeff7[8][2]; + __declspec(align(64)) VUINT32 poly_coeff6[8][2]; + __declspec(align(64)) VUINT32 poly_coeff5[8][2]; + __declspec(align(64)) VUINT32 poly_coeff4[8][2]; + __declspec(align(64)) VUINT32 poly_coeff3[8][2]; + __declspec(align(64)) VUINT32 poly_coeff2[8][2]; + __declspec(align(64)) VUINT32 poly_coeff1[8][2]; + } __svml_dcbrt_data_internal_avx512; +#endif +__svml_dcbrt_data_internal_avx512: + /*== etbl_H ==*/ + .quad 0x3ff0000000000000 + .quad 0x3ff428a2f98d728b + .quad 0x3ff965fea53d6e3d + .quad 0x0000000000000000 + .quad 0xbff0000000000000 + .quad 0xbff428a2f98d728b + .quad 0xbff965fea53d6e3d + .quad 0x0000000000000000 + /*== etbl_L ==*/ + .align 64 + .quad 0x0000000000000000 + .quad 0xbc7ddc22548ea41e + .quad 0xbc9f53e999952f09 + .quad 0x0000000000000000 + .quad 0x0000000000000000 + .quad 0x3c7ddc22548ea41e + .quad 0x3c9f53e999952f09 + .quad 0x0000000000000000 + /*== cbrt_tbl_H ==*/ + .align 64 + .quad 0x3ff428a2f98d728b + .quad 0x3ff361f35ca116ff + .quad 0x3ff2b6b5edf6b54a + .quad 0x3ff220e6dd675180 + .quad 0x3ff19c3b38e975a8 + .quad 0x3ff12589c21fb842 + .quad 0x3ff0ba6ee5f9aad4 + .quad 0x3ff059123d3a9848 + .quad 0x3ff0000000000000 + .quad 0x0000000000000000 + .quad 0x0000000000000000 + .quad 0x0000000000000000 + .quad 0x0000000000000000 + .quad 0x0000000000000000 + .quad 0x0000000000000000 + .quad 0x0000000000000000 + /*== BiasL ==*/ + .align 64 + .quad 0x4338000000000000, 0x4338000000000000, 0x4338000000000000, 0x4338000000000000, 0x4338000000000000, 0x4338000000000000, 0x4338000000000000, 0x4338000000000000 + /*== Zero ==*/ + .align 64 + .quad 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000, 0x8000000000000000 + /*== OneThird ==*/ + .align 64 + .quad 0x3fd5555555555556, 0x3fd5555555555556, 0x3fd5555555555556, 0x3fd5555555555556, 0x3fd5555555555556, 0x3fd5555555555556, 0x3fd5555555555556, 0x3fd5555555555556 + /*== Bias3 ==*/ + .align 64 + .quad 0x4320000000000000, 0x4320000000000000, 0x4320000000000000, 0x4320000000000000, 0x4320000000000000, 0x4320000000000000, 0x4320000000000000, 0x4320000000000000 + /*== Three ==*/ + .align 64 + .quad 0x4008000000000000, 0x4008000000000000, 0x4008000000000000, 0x4008000000000000, 0x4008000000000000, 0x4008000000000000, 0x4008000000000000, 0x4008000000000000 + /*==One ==*/ + .align 64 + .quad 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000, 0x3ff0000000000000 + /*== poly_coeff10 ==*/ + .align 64 + .quad 0xbf882e3b6adeca62, 0xbf882e3b6adeca62, 0xbf882e3b6adeca62, 0xbf882e3b6adeca62, 0xbf882e3b6adeca62, 0xbf882e3b6adeca62, 0xbf882e3b6adeca62, 0xbf882e3b6adeca62 + /*== poly_coeff9 ==*/ + .align 64 + .quad 0x3f8bda24bae48875, 0x3f8bda24bae48875, 0x3f8bda24bae48875, 0x3f8bda24bae48875, 0x3f8bda24bae48875, 0x3f8bda24bae48875, 0x3f8bda24bae48875, 0x3f8bda24bae48875 + /*== poly_coeff8 ==*/ + .align 64 + .quad 0xbf9036b87c71d55f, 0xbf9036b87c71d55f, 0xbf9036b87c71d55f, 0xbf9036b87c71d55f, 0xbf9036b87c71d55f, 0xbf9036b87c71d55f, 0xbf9036b87c71d55f, 0xbf9036b87c71d55f + /*== poly_coeff7 ==*/ + .align 64 + .quad 0x3f9374ed9398b914, 0x3f9374ed9398b914, 0x3f9374ed9398b914, 0x3f9374ed9398b914, 0x3f9374ed9398b914, 0x3f9374ed9398b914, 0x3f9374ed9398b914, 0x3f9374ed9398b914 + /*== poly_coeff6 ==*/ + .align 64 + .quad 0xbf98090d77f2468e, 0xbf98090d77f2468e, 0xbf98090d77f2468e, 0xbf98090d77f2468e, 0xbf98090d77f2468e, 0xbf98090d77f2468e, 0xbf98090d77f2468e, 0xbf98090d77f2468e + /*== poly_coeff5 ==*/ + .align 64 + .quad 0x3f9ee71141dcf569, 0x3f9ee71141dcf569, 0x3f9ee71141dcf569, 0x3f9ee71141dcf569, 0x3f9ee71141dcf569, 0x3f9ee71141dcf569, 0x3f9ee71141dcf569, 0x3f9ee71141dcf569 + /*== poly_coeff4 ==*/ + .align 64 + .quad 0xbfa511e8d2b0363e, 0xbfa511e8d2b0363e, 0xbfa511e8d2b0363e, 0xbfa511e8d2b0363e, 0xbfa511e8d2b0363e, 0xbfa511e8d2b0363e, 0xbfa511e8d2b0363e, 0xbfa511e8d2b0363e + /*== poly_coeff3 ==*/ + .align 64 + .quad 0x3faf9add3c0b7e31, 0x3faf9add3c0b7e31, 0x3faf9add3c0b7e31, 0x3faf9add3c0b7e31, 0x3faf9add3c0b7e31, 0x3faf9add3c0b7e31, 0x3faf9add3c0b7e31, 0x3faf9add3c0b7e31 + /*== poly_coeff2 ==*/ + .align 64 + .quad 0xbfbc71c71c71c741, 0xbfbc71c71c71c741, 0xbfbc71c71c71c741, 0xbfbc71c71c71c741, 0xbfbc71c71c71c741, 0xbfbc71c71c71c741, 0xbfbc71c71c71c741, 0xbfbc71c71c71c741 + /*== poly_coeff1 ==*/ + .align 64 + .quad 0x3fd5555555555557, 0x3fd5555555555557, 0x3fd5555555555557, 0x3fd5555555555557, 0x3fd5555555555557, 0x3fd5555555555557, 0x3fd5555555555557, 0x3fd5555555555557 + .align 64 + .type __svml_dcbrt_data_internal_avx512,@object + .size __svml_dcbrt_data_internal_avx512,.-__svml_dcbrt_data_internal_avx512 diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core-avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core-avx2.S new file mode 100644 index 0000000000..faa847fba6 --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core-avx2.S @@ -0,0 +1,20 @@ +/* AVX2 version of vectorized cbrtf. + 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_cbrtf _ZGVeN16v_cbrtf_avx2_wrapper +#include "../svml_s_cbrtf16_core.S" diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core.c b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core.c new file mode 100644 index 0000000000..785a68cc0d --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core.c @@ -0,0 +1,28 @@ +/* Multiple versions of vectorized cbrtf, 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_cbrtf +#include "ifunc-mathvec-avx512-skx.h" + +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); + +#ifdef SHARED +__hidden_ver1 (_ZGVeN16v_cbrtf, __GI__ZGVeN16v_cbrtf, + __redirect__ZGVeN16v_cbrtf) + __attribute__ ((visibility ("hidden"))); +#endif diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core_avx512.S b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core_avx512.S new file mode 100644 index 0000000000..55b017682b --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf16_core_avx512.S @@ -0,0 +1,235 @@ +/* Function cbrtf 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: + * + * x=2^{3*k+j} * 1.b1 b2 ... b5 b6 ... b52 + * Let r=(x*2^{-3k-j} - 1.b1 b2 ... b5 1)* rcp[b1 b2 ..b5], + * where rcp[b1 b2 .. b5]=1/(1.b1 b2 b3 b4 b5 1) in single precision + * cbrtf(2^j * 1. b1 b2 .. b5 1) is approximated as T[j][b1..b5]+D[j][b1..b5] + * (T stores the high 24 bits, D stores the low order bits) + * Result=2^k*T+(2^k*T*r)*P+2^k*D + * where P=p1+p2*r+.. + * + */ + +/* Offsets for data table __svml_scbrt_data_internal_avx512 + */ +#define etbl_H 0 +#define etbl_L 64 +#define cbrt_tbl_H 128 +#define BiasL 256 +#define SZero 320 +#define OneThird 384 +#define Bias3 448 +#define Three 512 +#define One 576 +#define poly_coeff3 640 +#define poly_coeff2 704 +#define poly_coeff1 768 + +#include + + .text + .section .text.exex512,"ax",@progbits +ENTRY(_ZGVeN16v_cbrtf_skx) + vgetmantps $0, {sae}, %zmm0, %zmm8 + +/* GetExp(x) */ + vgetexpps {sae}, %zmm0, %zmm1 + vmovups BiasL+__svml_scbrt_data_internal_avx512(%rip), %zmm2 + +/* exponent/3 */ + vmovups OneThird+__svml_scbrt_data_internal_avx512(%rip), %zmm3 + vmovups Bias3+__svml_scbrt_data_internal_avx512(%rip), %zmm4 + vmovups One+__svml_scbrt_data_internal_avx512(%rip), %zmm15 + +/* exponent%3 (to be used as index) */ + vmovups Three+__svml_scbrt_data_internal_avx512(%rip), %zmm5 + +/* polynomial */ + vmovups poly_coeff3+__svml_scbrt_data_internal_avx512(%rip), %zmm11 + vmovups poly_coeff1+__svml_scbrt_data_internal_avx512(%rip), %zmm14 + +/* Table lookup */ + vmovups cbrt_tbl_H+__svml_scbrt_data_internal_avx512(%rip), %zmm12 + +/* DblRcp ~ 1/Mantissa */ + vrcp14ps %zmm8, %zmm7 + vaddps {rn-sae}, %zmm2, %zmm1, %zmm6 + vandps SZero+__svml_scbrt_data_internal_avx512(%rip), %zmm0, %zmm0 + +/* round DblRcp to 3 fractional bits (RN mode, no Precision exception) */ + vrndscaleps $88, {sae}, %zmm7, %zmm9 + vfmsub231ps {rn-sae}, %zmm6, %zmm3, %zmm4 + vmovups poly_coeff2+__svml_scbrt_data_internal_avx512(%rip), %zmm7 + +/* Reduced argument: R = DblRcp*Mantissa - 1 */ + vfmsub231ps {rn-sae}, %zmm9, %zmm8, %zmm15 + vrndscaleps $9, {sae}, %zmm4, %zmm13 + +/* Prepare table index */ + vpsrld $19, %zmm9, %zmm10 + vfmadd231ps {rn-sae}, %zmm15, %zmm11, %zmm7 + vfnmadd231ps {rn-sae}, %zmm13, %zmm5, %zmm6 + vpermt2ps cbrt_tbl_H+64+__svml_scbrt_data_internal_avx512(%rip), %zmm10, %zmm12 + vfmadd213ps {rn-sae}, %zmm14, %zmm15, %zmm7 + vscalefps {rn-sae}, %zmm13, %zmm12, %zmm2 + +/* Table lookup: 2^(exponent%3) */ + vpermps __svml_scbrt_data_internal_avx512(%rip), %zmm6, %zmm1 + vpermps etbl_L+__svml_scbrt_data_internal_avx512(%rip), %zmm6, %zmm6 + +/* Sh*R */ + vmulps {rn-sae}, %zmm15, %zmm1, %zmm14 + +/* Sl + (Sh*R)*Poly */ + vfmadd213ps {rn-sae}, %zmm6, %zmm7, %zmm14 + +/* + * branch-free + * scaled_Th*(Sh+Sl+Sh*R*Poly) + */ + vaddps {rn-sae}, %zmm1, %zmm14, %zmm15 + vmulps {rn-sae}, %zmm2, %zmm15, %zmm3 + vorps %zmm0, %zmm3, %zmm0 + ret + +END(_ZGVeN16v_cbrtf_skx) + + .section .rodata, "a" + .align 64 + +#ifdef __svml_scbrt_data_internal_avx512_typedef +typedef unsigned int VUINT32; +typedef struct { + __declspec(align(64)) VUINT32 etbl_H[16][1]; + __declspec(align(64)) VUINT32 etbl_L[16][1]; + __declspec(align(64)) VUINT32 cbrt_tbl_H[32][1]; + __declspec(align(64)) VUINT32 BiasL[16][1]; + __declspec(align(64)) VUINT32 SZero[16][1]; + __declspec(align(64)) VUINT32 OneThird[16][1]; + __declspec(align(64)) VUINT32 Bias3[16][1]; + __declspec(align(64)) VUINT32 Three[16][1]; + __declspec(align(64)) VUINT32 One[16][1]; + __declspec(align(64)) VUINT32 poly_coeff3[16][1]; + __declspec(align(64)) VUINT32 poly_coeff2[16][1]; + __declspec(align(64)) VUINT32 poly_coeff1[16][1]; + } __svml_scbrt_data_internal_avx512; +#endif +__svml_scbrt_data_internal_avx512: + /*== etbl_H ==*/ + .long 0x3f800000 + .long 0x3fa14518 + .long 0x3fcb2ff5 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + /*== etbl_L ==*/ + .align 64 + .long 0x00000000 + .long 0xb2ce51af + .long 0x32a7adc8 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + /*== cbrt_tbl_H ==*/ + .align 64 + .long 0x3fa14518 + .long 0x3f9e0b2b + .long 0x3f9b0f9b + .long 0x3f984a9a + .long 0x3f95b5af + .long 0x3f934b6c + .long 0x3f910737 + .long 0x3f8ee526 + .long 0x3f8ce1da + .long 0x3f8afa6a + .long 0x3f892c4e + .long 0x3f87754e + .long 0x3f85d377 + .long 0x3f844510 + .long 0x3f82c892 + .long 0x3f815c9f + .long 0x3f800000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + .long 0x00000000 + /*== BiasL ==*/ + .align 64 + .long 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000, 0x4b400000 + /*== Zero ==*/ + .align 64 + .long 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000, 0x80000000 + /*== OneThird ==*/ + .align 64 + .long 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab, 0x3eaaaaab + /*== Bias3 ==*/ + .align 64 + .long 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000, 0x4a800000 + /*== Three ==*/ + .align 64 + .long 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000, 0x40400000 + /*==One ==*/ + .align 64 + .long 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000, 0x3f800000 + /*== poly_coeff3 ==*/ + .align 64 + .long 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c, 0x3d7d057c + /*== poly_coeff2 ==*/ + .align 64 + .long 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363, 0xbde3a363 + /*== poly_coeff1 ==*/ + .align 64 + .long 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa, 0x3eaaaaaa + .align 64 + .type __svml_scbrt_data_internal_avx512,@object + .size __svml_scbrt_data_internal_avx512,.-__svml_scbrt_data_internal_avx512 diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core-sse2.S b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core-sse2.S new file mode 100644 index 0000000000..76fc254e7a --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core-sse2.S @@ -0,0 +1,20 @@ +/* SSE2 version of vectorized cbrtf, 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_cbrtf _ZGVbN4v_cbrtf_sse2 +#include "../svml_s_cbrtf4_core.S" diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core.c b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core.c new file mode 100644 index 0000000000..564a549b39 --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core.c @@ -0,0 +1,28 @@ +/* Multiple versions of vectorized cbrtf, 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_cbrtf +#include "ifunc-mathvec-sse4_1.h" + +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); + +#ifdef SHARED +__hidden_ver1 (_ZGVbN4v_cbrtf, __GI__ZGVbN4v_cbrtf, + __redirect__ZGVbN4v_cbrtf) + __attribute__ ((visibility ("hidden"))); +#endif diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core_sse4.S b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core_sse4.S new file mode 100644 index 0000000000..af42dd5164 --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf4_core_sse4.S @@ -0,0 +1,490 @@ +/* Function cbrtf 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: + * + * x=2^{3*k+j} * 1.b1 b2 ... b5 b6 ... b52 + * Let r=(x*2^{-3k-j} - 1.b1 b2 ... b5 1)* rcp[b1 b2 ..b5], + * where rcp[b1 b2 .. b5]=1/(1.b1 b2 b3 b4 b5 1) in single precision + * cbrtf(2^j * 1. b1 b2 .. b5 1) is approximated as T[j][b1..b5]+D[j][b1..b5] + * (T stores the high 24 bits, D stores the low order bits) + * Result=2^k*T+(2^k*T*r)*P+2^k*D + * where P=p1+p2*r+.. + * + */ + +/* Offsets for data table __svml_scbrt_data_internal + */ +#define _sRcp 0 +#define _sCbrtHL 128 +#define _sP2 512 +#define _sP1 528 +#define _sMantissaMask 544 +#define _sMantissaMask1 560 +#define _sExpMask 576 +#define _sExpMask1 592 +#define _iRcpIndexMask 608 +#define _iBExpMask 624 +#define _iSignMask 640 +#define _iBias 656 +#define _iOne 672 +#define _i555 688 +#define _iAbsMask 704 +#define _iSubConst 720 +#define _iCmpConst 736 + +#include + + .text + .section .text.sse4,"ax",@progbits +ENTRY(_ZGVbN4v_cbrtf_sse4) + subq $72, %rsp + cfi_def_cfa_offset(80) + +/* + * Load constants + * Reciprocal index calculation + */ + movaps %xmm0, %xmm2 + movdqu _iRcpIndexMask+__svml_scbrt_data_internal(%rip), %xmm3 + psrld $16, %xmm2 + pand %xmm2, %xmm3 + +/* Load reciprocal value */ + lea __svml_scbrt_data_internal(%rip), %rdx + pshufd $1, %xmm3, %xmm5 + +/* Get signed biased exponent */ + psrld $7, %xmm2 + movd %xmm3, %eax + movd %xmm5, %ecx + +/* Get absolute biased exponent */ + movdqu _iBExpMask+__svml_scbrt_data_internal(%rip), %xmm15 + +/* + * Calculate exponent/3 + * i555Exp=(2^{12}-1)/3*exponent + */ + movdqu _i555+__svml_scbrt_data_internal(%rip), %xmm14 + pand %xmm2, %xmm15 + movslq %eax, %rax + movdqa %xmm14, %xmm5 + movslq %ecx, %rcx + psrlq $32, %xmm14 + pmuludq %xmm15, %xmm5 + movd (%rdx,%rax), %xmm4 + movd (%rdx,%rcx), %xmm6 + punpckldq %xmm6, %xmm4 + movdqa %xmm15, %xmm6 + psrlq $32, %xmm15 + pmuludq %xmm14, %xmm15 + pshufd $2, %xmm3, %xmm7 + psllq $32, %xmm15 + pshufd $3, %xmm3, %xmm8 + movd %xmm7, %esi + movd %xmm8, %edi + +/* Argument reduction */ + movups _sMantissaMask+__svml_scbrt_data_internal(%rip), %xmm12 + movups _sMantissaMask1+__svml_scbrt_data_internal(%rip), %xmm11 + andps %xmm0, %xmm12 + pand .FLT_17(%rip), %xmm5 + andps %xmm0, %xmm11 + movslq %esi, %rsi + por %xmm15, %xmm5 + movslq %edi, %rdi + +/* Get K (exponent=3*k+j) */ + psrld $12, %xmm5 + orps _sExpMask+__svml_scbrt_data_internal(%rip), %xmm12 + orps _sExpMask1+__svml_scbrt_data_internal(%rip), %xmm11 + psubd _iOne+__svml_scbrt_data_internal(%rip), %xmm6 + +/* r=y-y` */ + subps %xmm11, %xmm12 + +/* Get J */ + psubd %xmm5, %xmm6 + movdqu _iAbsMask+__svml_scbrt_data_internal(%rip), %xmm1 + psubd %xmm5, %xmm6 + movd (%rdx,%rsi), %xmm10 + pand %xmm0, %xmm1 + movd (%rdx,%rdi), %xmm9 + psubd %xmm5, %xmm6 + punpckldq %xmm9, %xmm10 + +/* Get 128*J */ + pslld $7, %xmm6 + punpcklqdq %xmm10, %xmm4 + +/* + * iCbrtIndex=4*l+128*j + * Zero index if callout expected + */ + paddd %xmm6, %xmm3 + psubd _iSubConst+__svml_scbrt_data_internal(%rip), %xmm1 + pcmpgtd _iCmpConst+__svml_scbrt_data_internal(%rip), %xmm1 + +/* r=(y-y`)*rcp_table(y`) */ + mulps %xmm12, %xmm4 + movmskps %xmm1, %eax + +/* Biased exponent-1 */ + movdqu _iSignMask+__svml_scbrt_data_internal(%rip), %xmm13 + pandn %xmm3, %xmm1 + +/* + * Add 2/3*(bias-1)+1 to (k+1/3*(bias-1)) + * Attach sign to exponent + */ + movdqu _iBias+__svml_scbrt_data_internal(%rip), %xmm12 + pand %xmm13, %xmm2 + paddd %xmm5, %xmm12 + +/* Load Cbrt table Hi & Lo values */ + movd %xmm1, %r8d + por %xmm2, %xmm12 + pshufd $1, %xmm1, %xmm2 + pslld $23, %xmm12 + pshufd $2, %xmm1, %xmm7 + pshufd $3, %xmm1, %xmm1 + movd %xmm2, %r9d + movd %xmm7, %r10d + movd %xmm1, %r11d + +/* Polynomial: p1+r*(p2*r+r*(p3+r*p4)) */ + movups _sP2+__svml_scbrt_data_internal(%rip), %xmm11 + mulps %xmm4, %xmm11 + movslq %r8d, %r8 + addps _sP1+__svml_scbrt_data_internal(%rip), %xmm11 + movslq %r9d, %r9 + movslq %r10d, %r10 + movslq %r11d, %r11 + movd 128(%rdx,%r8), %xmm10 + movd 128(%rdx,%r9), %xmm3 + movd 128(%rdx,%r10), %xmm9 + movd 128(%rdx,%r11), %xmm8 + punpckldq %xmm3, %xmm10 + punpckldq %xmm8, %xmm9 + punpcklqdq %xmm9, %xmm10 + +/* sCbrtHi *= 2^k */ + mulps %xmm10, %xmm12 + +/* T`*r */ + mulps %xmm12, %xmm4 + +/* (T`*r)*P */ + mulps %xmm4, %xmm11 + +/* + * T`*r*P+D` + * result = T`+(T`*r*P+D`) + */ + addps %xmm11, %xmm12 + testl %eax, %eax + +/* Go to special inputs processing branch */ + jne L(SPECIAL_VALUES_BRANCH) + # LOE rbx rbp r12 r13 r14 r15 eax xmm0 xmm12 + +/* Restore registers + * and exit the function + */ + +L(EXIT): + movaps %xmm12, %xmm0 + addq $72, %rsp + cfi_def_cfa_offset(8) + ret + cfi_def_cfa_offset(80) + +/* Branch to process + * special inputs + */ + +L(SPECIAL_VALUES_BRANCH): + movups %xmm0, 32(%rsp) + movups %xmm12, 48(%rsp) + # LOE rbx rbp r12 r13 r14 r15 eax + + xorl %edx, %edx + movq %r12, 16(%rsp) + cfi_offset(12, -64) + movl %edx, %r12d + movq %r13, 8(%rsp) + cfi_offset(13, -72) + movl %eax, %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), %xmm12 + +/* 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 xmm12 + +/* Scalar math fucntion call + * to process special input + */ + +L(SCALAR_MATH_CALL): + movl %r12d, %r14d + movss 32(%rsp,%r14,4), %xmm0 + call cbrtf@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_cbrtf_sse4) + + .section .rodata, "a" + .align 16 + +#ifdef __svml_scbrt_data_internal_typedef +typedef unsigned int VUINT32; +typedef struct +{ + __declspec(align(16)) VUINT32 _sRcp[32][1]; + __declspec(align(16)) VUINT32 _sCbrtHL[96][1]; + __declspec(align(16)) VUINT32 _sP2[4][1]; + __declspec(align(16)) VUINT32 _sP1[4][1]; + __declspec(align(16)) VUINT32 _sMantissaMask[4][1]; + __declspec(align(16)) VUINT32 _sMantissaMask1[4][1]; + __declspec(align(16)) VUINT32 _sExpMask[4][1]; + __declspec(align(16)) VUINT32 _sExpMask1[4][1]; + __declspec(align(16)) VUINT32 _iRcpIndexMask[4][1]; + __declspec(align(16)) VUINT32 _iBExpMask[4][1]; + __declspec(align(16)) VUINT32 _iSignMask[4][1]; + __declspec(align(16)) VUINT32 _iBias[4][1]; + __declspec(align(16)) VUINT32 _iOne[4][1]; + __declspec(align(16)) VUINT32 _i555[4][1]; + __declspec(align(16)) VUINT32 _iAbsMask[4][1]; + __declspec(align(16)) VUINT32 _iSubConst[4][1]; + __declspec(align(16)) VUINT32 _iCmpConst[4][1]; +} __svml_scbrt_data_internal; +#endif +__svml_scbrt_data_internal: + /*== _sRcp ==*/ + .long 0xBF7C0FC1 /* (1/(1+0/32+1/64)) = -.984615 */ + .long 0xBF74898D /* (1/(1+1/32+1/64)) = -.955224 */ + .long 0xBF6D7304 /* (1/(1+2/32+1/64)) = -.927536 */ + .long 0xBF66C2B4 /* (1/(1+3/32+1/64)) = -.901408 */ + .long 0xBF607038 /* (1/(1+4/32+1/64)) = -.876712 */ + .long 0xBF5A740E /* (1/(1+5/32+1/64)) = -.853333 */ + .long 0xBF54C77B /* (1/(1+6/32+1/64)) = -.831169 */ + .long 0xBF4F6475 /* (1/(1+7/32+1/64)) = -.810127 */ + .long 0xBF4A4588 /* (1/(1+8/32+1/64)) = -.790123 */ + .long 0xBF4565C8 /* (1/(1+9/32+1/64)) = -.771084 */ + .long 0xBF40C0C1 /* (1/(1+10/32+1/64)) = -.752941 */ + .long 0xBF3C5264 /* (1/(1+11/32+1/64)) = -.735632 */ + .long 0xBF381703 /* (1/(1+12/32+1/64)) = -.719101 */ + .long 0xBF340B41 /* (1/(1+13/32+1/64)) = -.703297 */ + .long 0xBF302C0B /* (1/(1+14/32+1/64)) = -.688172 */ + .long 0xBF2C7692 /* (1/(1+15/32+1/64)) = -.673684 */ + .long 0xBF28E83F /* (1/(1+16/32+1/64)) = -.659794 */ + .long 0xBF257EB5 /* (1/(1+17/32+1/64)) = -.646465 */ + .long 0xBF2237C3 /* (1/(1+18/32+1/64)) = -.633663 */ + .long 0xBF1F1166 /* (1/(1+19/32+1/64)) = -.621359 */ + .long 0xBF1C09C1 /* (1/(1+20/32+1/64)) = -.609524 */ + .long 0xBF191F1A /* (1/(1+21/32+1/64)) = -.598131 */ + .long 0xBF164FDA /* (1/(1+22/32+1/64)) = -.587156 */ + .long 0xBF139A86 /* (1/(1+23/32+1/64)) = -.576577 */ + .long 0xBF10FDBC /* (1/(1+24/32+1/64)) = -.566372 */ + .long 0xBF0E7835 /* (1/(1+25/32+1/64)) = -.556522 */ + .long 0xBF0C08C1 /* (1/(1+26/32+1/64)) = -.547009 */ + .long 0xBF09AE41 /* (1/(1+27/32+1/64)) = -.537815 */ + .long 0xBF0767AB /* (1/(1+28/32+1/64)) = -.528926 */ + .long 0xBF053408 /* (1/(1+29/32+1/64)) = -.520325 */ + .long 0xBF03126F /* (1/(1+30/32+1/64)) = -.512 */ + .long 0xBF010204 /* (1/(1+31/32+1/64)) = -.503937 */ + /*== _sCbrtHL ==*/ + .align 16 + .long 0x3F80A9C9 /* HI((2^0*(1+0/32+1/64))^(1/3)) = 1.005181 */ + .long 0x3F81F833 /* HI((2^0*(1+1/32+1/64))^(1/3)) = 1.015387 */ + .long 0x3F834007 /* HI((2^0*(1+2/32+1/64))^(1/3)) = 1.025391 */ + .long 0x3F848194 /* HI((2^0*(1+3/32+1/64))^(1/3)) = 1.035204 */ + .long 0x3F85BD25 /* HI((2^0*(1+4/32+1/64))^(1/3)) = 1.044835 */ + .long 0x3F86F300 /* HI((2^0*(1+5/32+1/64))^(1/3)) = 1.054291 */ + .long 0x3F882365 /* HI((2^0*(1+6/32+1/64))^(1/3)) = 1.06358 */ + .long 0x3F894E90 /* HI((2^0*(1+7/32+1/64))^(1/3)) = 1.07271 */ + .long 0x3F8A74B9 /* HI((2^0*(1+8/32+1/64))^(1/3)) = 1.081687 */ + .long 0x3F8B9615 /* HI((2^0*(1+9/32+1/64))^(1/3)) = 1.090518 */ + .long 0x3F8CB2D4 /* HI((2^0*(1+10/32+1/64))^(1/3)) = 1.099207 */ + .long 0x3F8DCB24 /* HI((2^0*(1+11/32+1/64))^(1/3)) = 1.107762 */ + .long 0x3F8EDF31 /* HI((2^0*(1+12/32+1/64))^(1/3)) = 1.116186 */ + .long 0x3F8FEF22 /* HI((2^0*(1+13/32+1/64))^(1/3)) = 1.124485 */ + .long 0x3F90FB1F /* HI((2^0*(1+14/32+1/64))^(1/3)) = 1.132664 */ + .long 0x3F92034C /* HI((2^0*(1+15/32+1/64))^(1/3)) = 1.140726 */ + .long 0x3F9307CA /* HI((2^0*(1+16/32+1/64))^(1/3)) = 1.148675 */ + .long 0x3F9408B9 /* HI((2^0*(1+17/32+1/64))^(1/3)) = 1.156516 */ + .long 0x3F950638 /* HI((2^0*(1+18/32+1/64))^(1/3)) = 1.164252 */ + .long 0x3F960064 /* HI((2^0*(1+19/32+1/64))^(1/3)) = 1.171887 */ + .long 0x3F96F759 /* HI((2^0*(1+20/32+1/64))^(1/3)) = 1.179423 */ + .long 0x3F97EB2F /* HI((2^0*(1+21/32+1/64))^(1/3)) = 1.186865 */ + .long 0x3F98DC01 /* HI((2^0*(1+22/32+1/64))^(1/3)) = 1.194214 */ + .long 0x3F99C9E5 /* HI((2^0*(1+23/32+1/64))^(1/3)) = 1.201474 */ + .long 0x3F9AB4F2 /* HI((2^0*(1+24/32+1/64))^(1/3)) = 1.208647 */ + .long 0x3F9B9D3D /* HI((2^0*(1+25/32+1/64))^(1/3)) = 1.215736 */ + .long 0x3F9C82DA /* HI((2^0*(1+26/32+1/64))^(1/3)) = 1.222743 */ + .long 0x3F9D65DD /* HI((2^0*(1+27/32+1/64))^(1/3)) = 1.229671 */ + .long 0x3F9E4659 /* HI((2^0*(1+28/32+1/64))^(1/3)) = 1.236522 */ + .long 0x3F9F245F /* HI((2^0*(1+29/32+1/64))^(1/3)) = 1.243297 */ + .long 0x3FA00000 /* HI((2^0*(1+30/32+1/64))^(1/3)) = 1.25 */ + .long 0x3FA0D94C /* HI((2^0*(1+31/32+1/64))^(1/3)) = 1.256631 */ + .long 0x3FA21B02 /* HI((2^1*(1+0/32+1/64))^(1/3)) = 1.266449 */ + .long 0x3FA3C059 /* HI((2^1*(1+1/32+1/64))^(1/3)) = 1.279307 */ + .long 0x3FA55D61 /* HI((2^1*(1+2/32+1/64))^(1/3)) = 1.291912 */ + .long 0x3FA6F282 /* HI((2^1*(1+3/32+1/64))^(1/3)) = 1.304276 */ + .long 0x3FA8801A /* HI((2^1*(1+4/32+1/64))^(1/3)) = 1.316409 */ + .long 0x3FAA067E /* HI((2^1*(1+5/32+1/64))^(1/3)) = 1.328323 */ + .long 0x3FAB8602 /* HI((2^1*(1+6/32+1/64))^(1/3)) = 1.340027 */ + .long 0x3FACFEEF /* HI((2^1*(1+7/32+1/64))^(1/3)) = 1.35153 */ + .long 0x3FAE718E /* HI((2^1*(1+8/32+1/64))^(1/3)) = 1.36284 */ + .long 0x3FAFDE1F /* HI((2^1*(1+9/32+1/64))^(1/3)) = 1.373966 */ + .long 0x3FB144E1 /* HI((2^1*(1+10/32+1/64))^(1/3)) = 1.384915 */ + .long 0x3FB2A60D /* HI((2^1*(1+11/32+1/64))^(1/3)) = 1.395692 */ + .long 0x3FB401DA /* HI((2^1*(1+12/32+1/64))^(1/3)) = 1.406307 */ + .long 0x3FB5587B /* HI((2^1*(1+13/32+1/64))^(1/3)) = 1.416763 */ + .long 0x3FB6AA20 /* HI((2^1*(1+14/32+1/64))^(1/3)) = 1.427067 */ + .long 0x3FB7F6F7 /* HI((2^1*(1+15/32+1/64))^(1/3)) = 1.437224 */ + .long 0x3FB93F29 /* HI((2^1*(1+16/32+1/64))^(1/3)) = 1.44724 */ + .long 0x3FBA82E1 /* HI((2^1*(1+17/32+1/64))^(1/3)) = 1.457119 */ + .long 0x3FBBC244 /* HI((2^1*(1+18/32+1/64))^(1/3)) = 1.466866 */ + .long 0x3FBCFD77 /* HI((2^1*(1+19/32+1/64))^(1/3)) = 1.476485 */ + .long 0x3FBE349B /* HI((2^1*(1+20/32+1/64))^(1/3)) = 1.48598 */ + .long 0x3FBF67D3 /* HI((2^1*(1+21/32+1/64))^(1/3)) = 1.495356 */ + .long 0x3FC0973C /* HI((2^1*(1+22/32+1/64))^(1/3)) = 1.504615 */ + .long 0x3FC1C2F6 /* HI((2^1*(1+23/32+1/64))^(1/3)) = 1.513762 */ + .long 0x3FC2EB1A /* HI((2^1*(1+24/32+1/64))^(1/3)) = 1.5228 */ + .long 0x3FC40FC6 /* HI((2^1*(1+25/32+1/64))^(1/3)) = 1.531731 */ + .long 0x3FC53112 /* HI((2^1*(1+26/32+1/64))^(1/3)) = 1.54056 */ + .long 0x3FC64F16 /* HI((2^1*(1+27/32+1/64))^(1/3)) = 1.549289 */ + .long 0x3FC769EB /* HI((2^1*(1+28/32+1/64))^(1/3)) = 1.55792 */ + .long 0x3FC881A6 /* HI((2^1*(1+29/32+1/64))^(1/3)) = 1.566457 */ + .long 0x3FC9965D /* HI((2^1*(1+30/32+1/64))^(1/3)) = 1.574901 */ + .long 0x3FCAA825 /* HI((2^1*(1+31/32+1/64))^(1/3)) = 1.583256 */ + .long 0x3FCC3D79 /* HI((2^2*(1+0/32+1/64))^(1/3)) = 1.595626 */ + .long 0x3FCE5054 /* HI((2^2*(1+1/32+1/64))^(1/3)) = 1.611826 */ + .long 0x3FD058B8 /* HI((2^2*(1+2/32+1/64))^(1/3)) = 1.627707 */ + .long 0x3FD25726 /* HI((2^2*(1+3/32+1/64))^(1/3)) = 1.643285 */ + .long 0x3FD44C15 /* HI((2^2*(1+4/32+1/64))^(1/3)) = 1.658572 */ + .long 0x3FD637F2 /* HI((2^2*(1+5/32+1/64))^(1/3)) = 1.673582 */ + .long 0x3FD81B24 /* HI((2^2*(1+6/32+1/64))^(1/3)) = 1.688328 */ + .long 0x3FD9F60B /* HI((2^2*(1+7/32+1/64))^(1/3)) = 1.702821 */ + .long 0x3FDBC8FE /* HI((2^2*(1+8/32+1/64))^(1/3)) = 1.717071 */ + .long 0x3FDD9452 /* HI((2^2*(1+9/32+1/64))^(1/3)) = 1.731089 */ + .long 0x3FDF5853 /* HI((2^2*(1+10/32+1/64))^(1/3)) = 1.744883 */ + .long 0x3FE1154B /* HI((2^2*(1+11/32+1/64))^(1/3)) = 1.758462 */ + .long 0x3FE2CB7F /* HI((2^2*(1+12/32+1/64))^(1/3)) = 1.771835 */ + .long 0x3FE47B2E /* HI((2^2*(1+13/32+1/64))^(1/3)) = 1.785009 */ + .long 0x3FE62496 /* HI((2^2*(1+14/32+1/64))^(1/3)) = 1.797992 */ + .long 0x3FE7C7F0 /* HI((2^2*(1+15/32+1/64))^(1/3)) = 1.810789 */ + .long 0x3FE96571 /* HI((2^2*(1+16/32+1/64))^(1/3)) = 1.823408 */ + .long 0x3FEAFD4C /* HI((2^2*(1+17/32+1/64))^(1/3)) = 1.835855 */ + .long 0x3FEC8FB3 /* HI((2^2*(1+18/32+1/64))^(1/3)) = 1.848135 */ + .long 0x3FEE1CD3 /* HI((2^2*(1+19/32+1/64))^(1/3)) = 1.860255 */ + .long 0x3FEFA4D7 /* HI((2^2*(1+20/32+1/64))^(1/3)) = 1.872218 */ + .long 0x3FF127E9 /* HI((2^2*(1+21/32+1/64))^(1/3)) = 1.88403 */ + .long 0x3FF2A62F /* HI((2^2*(1+22/32+1/64))^(1/3)) = 1.895697 */ + .long 0x3FF41FD0 /* HI((2^2*(1+23/32+1/64))^(1/3)) = 1.907221 */ + .long 0x3FF594EE /* HI((2^2*(1+24/32+1/64))^(1/3)) = 1.918607 */ + .long 0x3FF705AC /* HI((2^2*(1+25/32+1/64))^(1/3)) = 1.929861 */ + .long 0x3FF8722A /* HI((2^2*(1+26/32+1/64))^(1/3)) = 1.940984 */ + .long 0x3FF9DA86 /* HI((2^2*(1+27/32+1/64))^(1/3)) = 1.951981 */ + .long 0x3FFB3EDE /* HI((2^2*(1+28/32+1/64))^(1/3)) = 1.962856 */ + .long 0x3FFC9F4E /* HI((2^2*(1+29/32+1/64))^(1/3)) = 1.973612 */ + .long 0x3FFDFBF2 /* HI((2^2*(1+30/32+1/64))^(1/3)) = 1.984251 */ + .long 0x3FFF54E3 /* HI((2^2*(1+31/32+1/64))^(1/3)) = 1.994778 */ + .align 16 + .long 0xBDE3A962, 0xBDE3A962, 0xBDE3A962, 0xBDE3A962 /* _sP2 */ + .align 16 + .long 0x3EAAAC91, 0x3EAAAC91, 0x3EAAAC91, 0x3EAAAC91 /* _sP1 */ + .align 16 + .long 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff /* _sMantissaMask (EXP_MSK3) */ + .align 16 + .long 0x007e0000, 0x007e0000, 0x007e0000, 0x007e0000 /* _sMantissaMask1 (SIG_MASK) */ + .align 16 + .long 0xBF800000, 0xBF800000, 0xBF800000, 0xBF800000 /* _sExpMask (EXP_MASK) */ + .align 16 + .long 0xBF820000, 0xBF820000, 0xBF820000, 0xBF820000 /* _sExpMask1 (EXP_MASK2) */ + .align 16 + .long 0x0000007c, 0x0000007c, 0x0000007c, 0x0000007c /* _iRcpIndexMask */ + .align 16 + .long 0x000000ff, 0x000000ff, 0x000000ff, 0x000000ff /* _iBExpMask */ + .align 16 + .long 0x00000100, 0x00000100, 0x00000100, 0x00000100 /* _iSignMask */ + .align 16 + .long 0x00000055, 0x00000055, 0x00000055, 0x00000055 /* _iBias */ + .align 16 + .long 0x00000001, 0x00000001, 0x00000001, 0x00000001 /* _iOne */ + .align 16 + .long 0x00000555, 0x00000555, 0x00000555, 0x00000555 /* _i555 */ + .align 16 + .long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff /* _iAbsMask */ + .align 16 + .long 0x80800000, 0x80800000, 0x80800000, 0x80800000 /* _iSubConst */ + .align 16 + .long 0xFEFFFFFF, 0xFEFFFFFF, 0xFEFFFFFF, 0xFEFFFFFF /* _iCmpConst */ + .align 16 + .type __svml_scbrt_data_internal,@object + .size __svml_scbrt_data_internal,.-__svml_scbrt_data_internal + .align 16 + +.FLT_17: + .long 0xffffffff,0x00000000,0xffffffff,0x00000000 + .type .FLT_17,@object + .size .FLT_17,16 diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core-sse.S b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core-sse.S new file mode 100644 index 0000000000..8eaa457fa6 --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core-sse.S @@ -0,0 +1,20 @@ +/* SSE version of vectorized cbrtf, 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_cbrtf _ZGVdN8v_cbrtf_sse_wrapper +#include "../svml_s_cbrtf8_core.S" diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core.c b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core.c new file mode 100644 index 0000000000..089d28461f --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core.c @@ -0,0 +1,28 @@ +/* Multiple versions of vectorized cbrtf, 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_cbrtf +#include "ifunc-mathvec-avx2.h" + +libc_ifunc_redirected (REDIRECT_NAME, SYMBOL_NAME, IFUNC_SELECTOR ()); + +#ifdef SHARED +__hidden_ver1 (_ZGVdN8v_cbrtf, __GI__ZGVdN8v_cbrtf, + __redirect__ZGVdN8v_cbrtf) + __attribute__ ((visibility ("hidden"))); +#endif diff --git a/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core_avx2.S b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core_avx2.S new file mode 100644 index 0000000000..acd20d9db8 --- /dev/null +++ b/sysdeps/x86_64/fpu/multiarch/svml_s_cbrtf8_core_avx2.S @@ -0,0 +1,509 @@ +/* Function cbrtf 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: + * + * x=2^{3*k+j} * 1.b1 b2 ... b5 b6 ... b52 + * Let r=(x*2^{-3k-j} - 1.b1 b2 ... b5 1)* rcp[b1 b2 ..b5], + * where rcp[b1 b2 .. b5]=1/(1.b1 b2 b3 b4 b5 1) in single precision + * cbrtf(2^j * 1. b1 b2 .. b5 1) is approximated as T[j][b1..b5]+D[j][b1..b5] + * (T stores the high 24 bits, D stores the low order bits) + * Result=2^k*T+(2^k*T*r)*P+2^k*D + * where P=p1+p2*r+.. + * + */ + +/* Offsets for data table __svml_scbrt_data_internal + */ +#define _sRcp 0 +#define _sCbrtHL 128 +#define _sP2 512 +#define _sP1 544 +#define _sMantissaMask 576 +#define _sMantissaMask1 608 +#define _sExpMask 640 +#define _sExpMask1 672 +#define _iRcpIndexMask 704 +#define _iBExpMask 736 +#define _iSignMask 768 +#define _iBias 800 +#define _iOne 832 +#define _i555 864 +#define _iAbsMask 896 +#define _iSubConst 928 +#define _iCmpConst 960 + +#include + + .text + .section .text.avx2,"ax",@progbits +ENTRY(_ZGVdN8v_cbrtf_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 + +/* Load reciprocal value */ + lea __svml_scbrt_data_internal(%rip), %rdx + vmovaps %ymm0, %ymm5 + +/* + * Load constants + * Reciprocal index calculation + */ + vpsrld $16, %ymm5, %ymm3 + vpand _iRcpIndexMask+__svml_scbrt_data_internal(%rip), %ymm3, %ymm4 + vextractf128 $1, %ymm4, %xmm15 + vmovd %xmm4, %eax + vmovd %xmm15, %r8d + vpextrd $1, %xmm15, %r9d + vpextrd $2, %xmm15, %r10d + vpextrd $3, %xmm15, %r11d + movslq %r8d, %r8 + movslq %r9d, %r9 + movslq %r10d, %r10 + movslq %r11d, %r11 + vpextrd $1, %xmm4, %ecx + vpextrd $2, %xmm4, %esi + vpextrd $3, %xmm4, %edi + movslq %eax, %rax + movslq %ecx, %rcx + movslq %esi, %rsi + movslq %edi, %rdi + vmovd (%rdx,%r8), %xmm13 + vmovd (%rdx,%r9), %xmm14 + vmovd (%rdx,%r10), %xmm1 + vmovd (%rdx,%r11), %xmm0 + vpunpckldq %xmm14, %xmm13, %xmm2 + vpunpckldq %xmm0, %xmm1, %xmm13 + +/* Get signed biased exponent */ + vpsrld $7, %ymm3, %ymm0 + vmovd (%rdx,%rax), %xmm6 + vmovd (%rdx,%rcx), %xmm7 + vmovd (%rdx,%rsi), %xmm8 + vmovd (%rdx,%rdi), %xmm9 + vpunpckldq %xmm7, %xmm6, %xmm10 + vpunpckldq %xmm9, %xmm8, %xmm11 + vpunpcklqdq %xmm11, %xmm10, %xmm12 + vpunpcklqdq %xmm13, %xmm2, %xmm6 + vandps _iAbsMask+__svml_scbrt_data_internal(%rip), %ymm5, %ymm3 + +/* Argument reduction */ + vandps _sMantissaMask+__svml_scbrt_data_internal(%rip), %ymm5, %ymm8 + vandps _sMantissaMask1+__svml_scbrt_data_internal(%rip), %ymm5, %ymm9 + vpsubd _iSubConst+__svml_scbrt_data_internal(%rip), %ymm3, %ymm7 + vorps _sExpMask+__svml_scbrt_data_internal(%rip), %ymm8, %ymm10 + vorps _sExpMask1+__svml_scbrt_data_internal(%rip), %ymm9, %ymm11 + +/* r=y-y` */ + vsubps %ymm11, %ymm10, %ymm15 + +/* Biased exponent-1 */ + vpand _iSignMask+__svml_scbrt_data_internal(%rip), %ymm0, %ymm8 + vpcmpgtd _iCmpConst+__svml_scbrt_data_internal(%rip), %ymm7, %ymm2 + vmovmskps %ymm2, %eax + vinsertf128 $1, %xmm6, %ymm12, %ymm14 + +/* Get absolute biased exponent */ + vpand _iBExpMask+__svml_scbrt_data_internal(%rip), %ymm0, %ymm6 + +/* r=(y-y`)*rcp_table(y`) */ + vmulps %ymm15, %ymm14, %ymm1 + vpsubd _iOne+__svml_scbrt_data_internal(%rip), %ymm6, %ymm10 + +/* + * Calculate exponent/3 + * i555Exp=(2^{12}-1)/3*exponent + */ + vpmulld _i555+__svml_scbrt_data_internal(%rip), %ymm6, %ymm3 + +/* Get K (exponent=3*k+j) */ + vpsrld $12, %ymm3, %ymm13 + +/* Get J */ + vpsubd %ymm13, %ymm10, %ymm11 + +/* Add 2/3*(bias-1)+1 to (k+1/3*(bias-1)) */ + vpaddd _iBias+__svml_scbrt_data_internal(%rip), %ymm13, %ymm7 + vpsubd %ymm13, %ymm11, %ymm12 + +/* Attach sign to exponent */ + vpor %ymm8, %ymm7, %ymm9 + vpsubd %ymm13, %ymm12, %ymm14 + vpslld $23, %ymm9, %ymm0 + +/* Get 128*J */ + vpslld $7, %ymm14, %ymm15 + +/* iCbrtIndex=4*l+128*j */ + vpaddd %ymm15, %ymm4, %ymm4 + +/* Zero index if callout expected */ + vpandn %ymm4, %ymm2, %ymm4 + +/* Load Cbrt table Hi & Lo values */ + vmovd %xmm4, %ecx + vextractf128 $1, %ymm4, %xmm13 + vpextrd $1, %xmm4, %esi + movslq %ecx, %rcx + movslq %esi, %rsi + vmovd %xmm13, %r9d + vmovd 128(%rdx,%rcx), %xmm2 + vpextrd $2, %xmm4, %edi + vpextrd $3, %xmm4, %r8d + vmovd 128(%rdx,%rsi), %xmm3 + vpextrd $1, %xmm13, %r10d + vpextrd $2, %xmm13, %ecx + vpextrd $3, %xmm13, %esi + movslq %edi, %rdi + movslq %r8d, %r8 + movslq %r9d, %r9 + movslq %r10d, %r10 + movslq %ecx, %rcx + movslq %esi, %rsi + vmovd 128(%rdx,%rdi), %xmm6 + vmovd 128(%rdx,%r8), %xmm7 + vmovd 128(%rdx,%r9), %xmm11 + vmovd 128(%rdx,%r10), %xmm12 + vmovd 128(%rdx,%rcx), %xmm14 + vmovd 128(%rdx,%rsi), %xmm15 + vpunpckldq %xmm3, %xmm2, %xmm8 + vpunpckldq %xmm7, %xmm6, %xmm9 + vpunpckldq %xmm12, %xmm11, %xmm4 + vpunpckldq %xmm15, %xmm14, %xmm11 + vpunpcklqdq %xmm9, %xmm8, %xmm10 + vpunpcklqdq %xmm11, %xmm4, %xmm2 + vinsertf128 $1, %xmm2, %ymm10, %ymm3 + +/* sCbrtHi *= 2^k */ + vmulps %ymm3, %ymm0, %ymm2 + +/* Polynomial: p1+r*(p2*r+r*(p3+r*p4)) */ + vmovups _sP2+__svml_scbrt_data_internal(%rip), %ymm0 + vfmadd213ps _sP1+__svml_scbrt_data_internal(%rip), %ymm1, %ymm0 + +/* T`*r */ + vmulps %ymm2, %ymm1, %ymm1 + +/* (T`*r)*P */ + vmulps %ymm1, %ymm0, %ymm0 + +/* + * T`*r*P+D` + * result = T`+(T`*r*P+D`) + */ + vaddps %ymm0, %ymm2, %ymm0 + testl %eax, %eax + +/* Go to special inputs processing branch */ + jne L(SPECIAL_VALUES_BRANCH) + # LOE rbx r12 r13 r14 r15 eax 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): + vmovups %ymm5, 32(%rsp) + vmovups %ymm0, 64(%rsp) + # LOE rbx r12 r13 r14 r15 eax ymm0 + + xorl %edx, %edx + # 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 %edx, %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 %eax, %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 cbrtf@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_cbrtf_avx2) + + .section .rodata, "a" + .align 32 + +#ifdef __svml_scbrt_data_internal_typedef +typedef unsigned int VUINT32; +typedef struct +{ + __declspec(align(32)) VUINT32 _sRcp[32][1]; + __declspec(align(32)) VUINT32 _sCbrtHL[96][1]; + __declspec(align(32)) VUINT32 _sP2[8][1]; + __declspec(align(32)) VUINT32 _sP1[8][1]; + __declspec(align(32)) VUINT32 _sMantissaMask[8][1]; + __declspec(align(32)) VUINT32 _sMantissaMask1[8][1]; + __declspec(align(32)) VUINT32 _sExpMask[8][1]; + __declspec(align(32)) VUINT32 _sExpMask1[8][1]; + __declspec(align(32)) VUINT32 _iRcpIndexMask[8][1]; + __declspec(align(32)) VUINT32 _iBExpMask[8][1]; + __declspec(align(32)) VUINT32 _iSignMask[8][1]; + __declspec(align(32)) VUINT32 _iBias[8][1]; + __declspec(align(32)) VUINT32 _iOne[8][1]; + __declspec(align(32)) VUINT32 _i555[8][1]; + __declspec(align(32)) VUINT32 _iAbsMask[8][1]; + __declspec(align(32)) VUINT32 _iSubConst[8][1]; + __declspec(align(32)) VUINT32 _iCmpConst[8][1]; +} __svml_scbrt_data_internal; +#endif +__svml_scbrt_data_internal: + /*== _sRcp ==*/ + .long 0xBF7C0FC1 /* (1/(1+0/32+1/64)) = -.984615 */ + .long 0xBF74898D /* (1/(1+1/32+1/64)) = -.955224 */ + .long 0xBF6D7304 /* (1/(1+2/32+1/64)) = -.927536 */ + .long 0xBF66C2B4 /* (1/(1+3/32+1/64)) = -.901408 */ + .long 0xBF607038 /* (1/(1+4/32+1/64)) = -.876712 */ + .long 0xBF5A740E /* (1/(1+5/32+1/64)) = -.853333 */ + .long 0xBF54C77B /* (1/(1+6/32+1/64)) = -.831169 */ + .long 0xBF4F6475 /* (1/(1+7/32+1/64)) = -.810127 */ + .long 0xBF4A4588 /* (1/(1+8/32+1/64)) = -.790123 */ + .long 0xBF4565C8 /* (1/(1+9/32+1/64)) = -.771084 */ + .long 0xBF40C0C1 /* (1/(1+10/32+1/64)) = -.752941 */ + .long 0xBF3C5264 /* (1/(1+11/32+1/64)) = -.735632 */ + .long 0xBF381703 /* (1/(1+12/32+1/64)) = -.719101 */ + .long 0xBF340B41 /* (1/(1+13/32+1/64)) = -.703297 */ + .long 0xBF302C0B /* (1/(1+14/32+1/64)) = -.688172 */ + .long 0xBF2C7692 /* (1/(1+15/32+1/64)) = -.673684 */ + .long 0xBF28E83F /* (1/(1+16/32+1/64)) = -.659794 */ + .long 0xBF257EB5 /* (1/(1+17/32+1/64)) = -.646465 */ + .long 0xBF2237C3 /* (1/(1+18/32+1/64)) = -.633663 */ + .long 0xBF1F1166 /* (1/(1+19/32+1/64)) = -.621359 */ + .long 0xBF1C09C1 /* (1/(1+20/32+1/64)) = -.609524 */ + .long 0xBF191F1A /* (1/(1+21/32+1/64)) = -.598131 */ + .long 0xBF164FDA /* (1/(1+22/32+1/64)) = -.587156 */ + .long 0xBF139A86 /* (1/(1+23/32+1/64)) = -.576577 */ + .long 0xBF10FDBC /* (1/(1+24/32+1/64)) = -.566372 */ + .long 0xBF0E7835 /* (1/(1+25/32+1/64)) = -.556522 */ + .long 0xBF0C08C1 /* (1/(1+26/32+1/64)) = -.547009 */ + .long 0xBF09AE41 /* (1/(1+27/32+1/64)) = -.537815 */ + .long 0xBF0767AB /* (1/(1+28/32+1/64)) = -.528926 */ + .long 0xBF053408 /* (1/(1+29/32+1/64)) = -.520325 */ + .long 0xBF03126F /* (1/(1+30/32+1/64)) = -.512 */ + .long 0xBF010204 /* (1/(1+31/32+1/64)) = -.503937 */ + /*== _sCbrtHL ==*/ + .align 32 + .long 0x3F80A9C9 /* HI((2^0*(1+0/32+1/64))^(1/3)) = 1.005181 */ + .long 0x3F81F833 /* HI((2^0*(1+1/32+1/64))^(1/3)) = 1.015387 */ + .long 0x3F834007 /* HI((2^0*(1+2/32+1/64))^(1/3)) = 1.025391 */ + .long 0x3F848194 /* HI((2^0*(1+3/32+1/64))^(1/3)) = 1.035204 */ + .long 0x3F85BD25 /* HI((2^0*(1+4/32+1/64))^(1/3)) = 1.044835 */ + .long 0x3F86F300 /* HI((2^0*(1+5/32+1/64))^(1/3)) = 1.054291 */ + .long 0x3F882365 /* HI((2^0*(1+6/32+1/64))^(1/3)) = 1.06358 */ + .long 0x3F894E90 /* HI((2^0*(1+7/32+1/64))^(1/3)) = 1.07271 */ + .long 0x3F8A74B9 /* HI((2^0*(1+8/32+1/64))^(1/3)) = 1.081687 */ + .long 0x3F8B9615 /* HI((2^0*(1+9/32+1/64))^(1/3)) = 1.090518 */ + .long 0x3F8CB2D4 /* HI((2^0*(1+10/32+1/64))^(1/3)) = 1.099207 */ + .long 0x3F8DCB24 /* HI((2^0*(1+11/32+1/64))^(1/3)) = 1.107762 */ + .long 0x3F8EDF31 /* HI((2^0*(1+12/32+1/64))^(1/3)) = 1.116186 */ + .long 0x3F8FEF22 /* HI((2^0*(1+13/32+1/64))^(1/3)) = 1.124485 */ + .long 0x3F90FB1F /* HI((2^0*(1+14/32+1/64))^(1/3)) = 1.132664 */ + .long 0x3F92034C /* HI((2^0*(1+15/32+1/64))^(1/3)) = 1.140726 */ + .long 0x3F9307CA /* HI((2^0*(1+16/32+1/64))^(1/3)) = 1.148675 */ + .long 0x3F9408B9 /* HI((2^0*(1+17/32+1/64))^(1/3)) = 1.156516 */ + .long 0x3F950638 /* HI((2^0*(1+18/32+1/64))^(1/3)) = 1.164252 */ + .long 0x3F960064 /* HI((2^0*(1+19/32+1/64))^(1/3)) = 1.171887 */ + .long 0x3F96F759 /* HI((2^0*(1+20/32+1/64))^(1/3)) = 1.179423 */ + .long 0x3F97EB2F /* HI((2^0*(1+21/32+1/64))^(1/3)) = 1.186865 */ + .long 0x3F98DC01 /* HI((2^0*(1+22/32+1/64))^(1/3)) = 1.194214 */ + .long 0x3F99C9E5 /* HI((2^0*(1+23/32+1/64))^(1/3)) = 1.201474 */ + .long 0x3F9AB4F2 /* HI((2^0*(1+24/32+1/64))^(1/3)) = 1.208647 */ + .long 0x3F9B9D3D /* HI((2^0*(1+25/32+1/64))^(1/3)) = 1.215736 */ + .long 0x3F9C82DA /* HI((2^0*(1+26/32+1/64))^(1/3)) = 1.222743 */ + .long 0x3F9D65DD /* HI((2^0*(1+27/32+1/64))^(1/3)) = 1.229671 */ + .long 0x3F9E4659 /* HI((2^0*(1+28/32+1/64))^(1/3)) = 1.236522 */ + .long 0x3F9F245F /* HI((2^0*(1+29/32+1/64))^(1/3)) = 1.243297 */ + .long 0x3FA00000 /* HI((2^0*(1+30/32+1/64))^(1/3)) = 1.25 */ + .long 0x3FA0D94C /* HI((2^0*(1+31/32+1/64))^(1/3)) = 1.256631 */ + .long 0x3FA21B02 /* HI((2^1*(1+0/32+1/64))^(1/3)) = 1.266449 */ + .long 0x3FA3C059 /* HI((2^1*(1+1/32+1/64))^(1/3)) = 1.279307 */ + .long 0x3FA55D61 /* HI((2^1*(1+2/32+1/64))^(1/3)) = 1.291912 */ + .long 0x3FA6F282 /* HI((2^1*(1+3/32+1/64))^(1/3)) = 1.304276 */ + .long 0x3FA8801A /* HI((2^1*(1+4/32+1/64))^(1/3)) = 1.316409 */ + .long 0x3FAA067E /* HI((2^1*(1+5/32+1/64))^(1/3)) = 1.328323 */ + .long 0x3FAB8602 /* HI((2^1*(1+6/32+1/64))^(1/3)) = 1.340027 */ + .long 0x3FACFEEF /* HI((2^1*(1+7/32+1/64))^(1/3)) = 1.35153 */ + .long 0x3FAE718E /* HI((2^1*(1+8/32+1/64))^(1/3)) = 1.36284 */ + .long 0x3FAFDE1F /* HI((2^1*(1+9/32+1/64))^(1/3)) = 1.373966 */ + .long 0x3FB144E1 /* HI((2^1*(1+10/32+1/64))^(1/3)) = 1.384915 */ + .long 0x3FB2A60D /* HI((2^1*(1+11/32+1/64))^(1/3)) = 1.395692 */ + .long 0x3FB401DA /* HI((2^1*(1+12/32+1/64))^(1/3)) = 1.406307 */ + .long 0x3FB5587B /* HI((2^1*(1+13/32+1/64))^(1/3)) = 1.416763 */ + .long 0x3FB6AA20 /* HI((2^1*(1+14/32+1/64))^(1/3)) = 1.427067 */ + .long 0x3FB7F6F7 /* HI((2^1*(1+15/32+1/64))^(1/3)) = 1.437224 */ + .long 0x3FB93F29 /* HI((2^1*(1+16/32+1/64))^(1/3)) = 1.44724 */ + .long 0x3FBA82E1 /* HI((2^1*(1+17/32+1/64))^(1/3)) = 1.457119 */ + .long 0x3FBBC244 /* HI((2^1*(1+18/32+1/64))^(1/3)) = 1.466866 */ + .long 0x3FBCFD77 /* HI((2^1*(1+19/32+1/64))^(1/3)) = 1.476485 */ + .long 0x3FBE349B /* HI((2^1*(1+20/32+1/64))^(1/3)) = 1.48598 */ + .long 0x3FBF67D3 /* HI((2^1*(1+21/32+1/64))^(1/3)) = 1.495356 */ + .long 0x3FC0973C /* HI((2^1*(1+22/32+1/64))^(1/3)) = 1.504615 */ + .long 0x3FC1C2F6 /* HI((2^1*(1+23/32+1/64))^(1/3)) = 1.513762 */ + .long 0x3FC2EB1A /* HI((2^1*(1+24/32+1/64))^(1/3)) = 1.5228 */ + .long 0x3FC40FC6 /* HI((2^1*(1+25/32+1/64))^(1/3)) = 1.531731 */ + .long 0x3FC53112 /* HI((2^1*(1+26/32+1/64))^(1/3)) = 1.54056 */ + .long 0x3FC64F16 /* HI((2^1*(1+27/32+1/64))^(1/3)) = 1.549289 */ + .long 0x3FC769EB /* HI((2^1*(1+28/32+1/64))^(1/3)) = 1.55792 */ + .long 0x3FC881A6 /* HI((2^1*(1+29/32+1/64))^(1/3)) = 1.566457 */ + .long 0x3FC9965D /* HI((2^1*(1+30/32+1/64))^(1/3)) = 1.574901 */ + .long 0x3FCAA825 /* HI((2^1*(1+31/32+1/64))^(1/3)) = 1.583256 */ + .long 0x3FCC3D79 /* HI((2^2*(1+0/32+1/64))^(1/3)) = 1.595626 */ + .long 0x3FCE5054 /* HI((2^2*(1+1/32+1/64))^(1/3)) = 1.611826 */ + .long 0x3FD058B8 /* HI((2^2*(1+2/32+1/64))^(1/3)) = 1.627707 */ + .long 0x3FD25726 /* HI((2^2*(1+3/32+1/64))^(1/3)) = 1.643285 */ + .long 0x3FD44C15 /* HI((2^2*(1+4/32+1/64))^(1/3)) = 1.658572 */ + .long 0x3FD637F2 /* HI((2^2*(1+5/32+1/64))^(1/3)) = 1.673582 */ + .long 0x3FD81B24 /* HI((2^2*(1+6/32+1/64))^(1/3)) = 1.688328 */ + .long 0x3FD9F60B /* HI((2^2*(1+7/32+1/64))^(1/3)) = 1.702821 */ + .long 0x3FDBC8FE /* HI((2^2*(1+8/32+1/64))^(1/3)) = 1.717071 */ + .long 0x3FDD9452 /* HI((2^2*(1+9/32+1/64))^(1/3)) = 1.731089 */ + .long 0x3FDF5853 /* HI((2^2*(1+10/32+1/64))^(1/3)) = 1.744883 */ + .long 0x3FE1154B /* HI((2^2*(1+11/32+1/64))^(1/3)) = 1.758462 */ + .long 0x3FE2CB7F /* HI((2^2*(1+12/32+1/64))^(1/3)) = 1.771835 */ + .long 0x3FE47B2E /* HI((2^2*(1+13/32+1/64))^(1/3)) = 1.785009 */ + .long 0x3FE62496 /* HI((2^2*(1+14/32+1/64))^(1/3)) = 1.797992 */ + .long 0x3FE7C7F0 /* HI((2^2*(1+15/32+1/64))^(1/3)) = 1.810789 */ + .long 0x3FE96571 /* HI((2^2*(1+16/32+1/64))^(1/3)) = 1.823408 */ + .long 0x3FEAFD4C /* HI((2^2*(1+17/32+1/64))^(1/3)) = 1.835855 */ + .long 0x3FEC8FB3 /* HI((2^2*(1+18/32+1/64))^(1/3)) = 1.848135 */ + .long 0x3FEE1CD3 /* HI((2^2*(1+19/32+1/64))^(1/3)) = 1.860255 */ + .long 0x3FEFA4D7 /* HI((2^2*(1+20/32+1/64))^(1/3)) = 1.872218 */ + .long 0x3FF127E9 /* HI((2^2*(1+21/32+1/64))^(1/3)) = 1.88403 */ + .long 0x3FF2A62F /* HI((2^2*(1+22/32+1/64))^(1/3)) = 1.895697 */ + .long 0x3FF41FD0 /* HI((2^2*(1+23/32+1/64))^(1/3)) = 1.907221 */ + .long 0x3FF594EE /* HI((2^2*(1+24/32+1/64))^(1/3)) = 1.918607 */ + .long 0x3FF705AC /* HI((2^2*(1+25/32+1/64))^(1/3)) = 1.929861 */ + .long 0x3FF8722A /* HI((2^2*(1+26/32+1/64))^(1/3)) = 1.940984 */ + .long 0x3FF9DA86 /* HI((2^2*(1+27/32+1/64))^(1/3)) = 1.951981 */ + .long 0x3FFB3EDE /* HI((2^2*(1+28/32+1/64))^(1/3)) = 1.962856 */ + .long 0x3FFC9F4E /* HI((2^2*(1+29/32+1/64))^(1/3)) = 1.973612 */ + .long 0x3FFDFBF2 /* HI((2^2*(1+30/32+1/64))^(1/3)) = 1.984251 */ + .long 0x3FFF54E3 /* HI((2^2*(1+31/32+1/64))^(1/3)) = 1.994778 */ + .align 32 + .long 0xBDE3A962, 0xBDE3A962, 0xBDE3A962, 0xBDE3A962, 0xBDE3A962, 0xBDE3A962, 0xBDE3A962, 0xBDE3A962 /* _sP2 */ + .align 32 + .long 0x3EAAAC91, 0x3EAAAC91, 0x3EAAAC91, 0x3EAAAC91, 0x3EAAAC91, 0x3EAAAC91, 0x3EAAAC91, 0x3EAAAC91 /* _sP1 */ + .align 32 + .long 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff, 0x007fffff /* _sMantissaMask (EXP_MSK3) */ + .align 32 + .long 0x007e0000, 0x007e0000, 0x007e0000, 0x007e0000, 0x007e0000, 0x007e0000, 0x007e0000, 0x007e0000 /* _sMantissaMask1 (SIG_MASK) */ + .align 32 + .long 0xBF800000, 0xBF800000, 0xBF800000, 0xBF800000, 0xBF800000, 0xBF800000, 0xBF800000, 0xBF800000 /* _sExpMask (EXP_MASK) */ + .align 32 + .long 0xBF820000, 0xBF820000, 0xBF820000, 0xBF820000, 0xBF820000, 0xBF820000, 0xBF820000, 0xBF820000 /* _sExpMask1 (EXP_MASK2) */ + .align 32 + .long 0x0000007c, 0x0000007c, 0x0000007c, 0x0000007c, 0x0000007c, 0x0000007c, 0x0000007c, 0x0000007c /* _iRcpIndexMask */ + .align 32 + .long 0x000000ff, 0x000000ff, 0x000000ff, 0x000000ff, 0x000000ff, 0x000000ff, 0x000000ff, 0x000000ff /* _iBExpMask */ + .align 32 + .long 0x00000100, 0x00000100, 0x00000100, 0x00000100, 0x00000100, 0x00000100, 0x00000100, 0x00000100 /* _iSignMask */ + .align 32 + .long 0x00000055, 0x00000055, 0x00000055, 0x00000055, 0x00000055, 0x00000055, 0x00000055, 0x00000055 /* _iBias */ + .align 32 + .long 0x00000001, 0x00000001, 0x00000001, 0x00000001, 0x00000001, 0x00000001, 0x00000001, 0x00000001 /* _iOne */ + .align 32 + .long 0x00000555, 0x00000555, 0x00000555, 0x00000555, 0x00000555, 0x00000555, 0x00000555, 0x00000555 /* _i555 */ + .align 32 + .long 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff, 0x7fffffff /* _iAbsMask */ + .align 32 + .long 0x80800000, 0x80800000, 0x80800000, 0x80800000, 0x80800000, 0x80800000, 0x80800000, 0x80800000 /* _iSubConst */ + .align 32 + .long 0xFEFFFFFF, 0xFEFFFFFF, 0xFEFFFFFF, 0xFEFFFFFF, 0xFEFFFFFF, 0xFEFFFFFF, 0xFEFFFFFF, 0xFEFFFFFF /* _iCmpConst */ + .align 32 + .type __svml_scbrt_data_internal,@object + .size __svml_scbrt_data_internal,.-__svml_scbrt_data_internal diff --git a/sysdeps/x86_64/fpu/svml_d_cbrt2_core.S b/sysdeps/x86_64/fpu/svml_d_cbrt2_core.S new file mode 100644 index 0000000000..4bf546564b --- /dev/null +++ b/sysdeps/x86_64/fpu/svml_d_cbrt2_core.S @@ -0,0 +1,29 @@ +/* Function cbrt 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_cbrt) +WRAPPER_IMPL_SSE2 cbrt +END (_ZGVbN2v_cbrt) + +#ifndef USE_MULTIARCH + libmvec_hidden_def (_ZGVbN2v_cbrt) +#endif diff --git a/sysdeps/x86_64/fpu/svml_d_cbrt4_core.S b/sysdeps/x86_64/fpu/svml_d_cbrt4_core.S new file mode 100644 index 0000000000..e6d1003e27 --- /dev/null +++ b/sysdeps/x86_64/fpu/svml_d_cbrt4_core.S @@ -0,0 +1,29 @@ +/* Function cbrt 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_cbrt) +WRAPPER_IMPL_AVX _ZGVbN2v_cbrt +END (_ZGVdN4v_cbrt) + +#ifndef USE_MULTIARCH + libmvec_hidden_def (_ZGVdN4v_cbrt) +#endif diff --git a/sysdeps/x86_64/fpu/svml_d_cbrt4_core_avx.S b/sysdeps/x86_64/fpu/svml_d_cbrt4_core_avx.S new file mode 100644 index 0000000000..70632869ac --- /dev/null +++ b/sysdeps/x86_64/fpu/svml_d_cbrt4_core_avx.S @@ -0,0 +1,25 @@ +/* Function cbrt 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_cbrt) +WRAPPER_IMPL_AVX _ZGVbN2v_cbrt +END (_ZGVcN4v_cbrt) diff --git a/sysdeps/x86_64/fpu/svml_d_cbrt8_core.S b/sysdeps/x86_64/fpu/svml_d_cbrt8_core.S new file mode 100644 index 0000000000..37571673a7 --- /dev/null +++ b/sysdeps/x86_64/fpu/svml_d_cbrt8_core.S @@ -0,0 +1,25 @@ +/* Function cbrt 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_cbrt) +WRAPPER_IMPL_AVX512 _ZGVdN4v_cbrt +END (_ZGVeN8v_cbrt) diff --git a/sysdeps/x86_64/fpu/svml_s_cbrtf16_core.S b/sysdeps/x86_64/fpu/svml_s_cbrtf16_core.S new file mode 100644 index 0000000000..1be6294026 --- /dev/null +++ b/sysdeps/x86_64/fpu/svml_s_cbrtf16_core.S @@ -0,0 +1,25 @@ +/* Function cbrtf 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_cbrtf) +WRAPPER_IMPL_AVX512 _ZGVdN8v_cbrtf +END (_ZGVeN16v_cbrtf) diff --git a/sysdeps/x86_64/fpu/svml_s_cbrtf4_core.S b/sysdeps/x86_64/fpu/svml_s_cbrtf4_core.S new file mode 100644 index 0000000000..2469a100f4 --- /dev/null +++ b/sysdeps/x86_64/fpu/svml_s_cbrtf4_core.S @@ -0,0 +1,29 @@ +/* Function cbrtf 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_cbrtf) +WRAPPER_IMPL_SSE2 cbrtf +END (_ZGVbN4v_cbrtf) + +#ifndef USE_MULTIARCH + libmvec_hidden_def (_ZGVbN4v_cbrtf) +#endif diff --git a/sysdeps/x86_64/fpu/svml_s_cbrtf8_core.S b/sysdeps/x86_64/fpu/svml_s_cbrtf8_core.S new file mode 100644 index 0000000000..efedc22323 --- /dev/null +++ b/sysdeps/x86_64/fpu/svml_s_cbrtf8_core.S @@ -0,0 +1,29 @@ +/* Function cbrtf 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_cbrtf) +WRAPPER_IMPL_AVX _ZGVbN4v_cbrtf +END (_ZGVdN8v_cbrtf) + +#ifndef USE_MULTIARCH + libmvec_hidden_def (_ZGVdN8v_cbrtf) +#endif diff --git a/sysdeps/x86_64/fpu/svml_s_cbrtf8_core_avx.S b/sysdeps/x86_64/fpu/svml_s_cbrtf8_core_avx.S new file mode 100644 index 0000000000..b5acc62426 --- /dev/null +++ b/sysdeps/x86_64/fpu/svml_s_cbrtf8_core_avx.S @@ -0,0 +1,25 @@ +/* Function cbrtf 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_cbrtf) +WRAPPER_IMPL_AVX _ZGVbN4v_cbrtf +END (_ZGVcN8v_cbrtf) diff --git a/sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx.c b/sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx.c new file mode 100644 index 0000000000..c8bc643c99 --- /dev/null +++ b/sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx.c @@ -0,0 +1 @@ +#include "test-double-libmvec-cbrt.c" diff --git a/sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx2.c b/sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx2.c new file mode 100644 index 0000000000..c8bc643c99 --- /dev/null +++ b/sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx2.c @@ -0,0 +1 @@ +#include "test-double-libmvec-cbrt.c" diff --git a/sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx512f.c b/sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx512f.c new file mode 100644 index 0000000000..c8bc643c99 --- /dev/null +++ b/sysdeps/x86_64/fpu/test-double-libmvec-cbrt-avx512f.c @@ -0,0 +1 @@ +#include "test-double-libmvec-cbrt.c" diff --git a/sysdeps/x86_64/fpu/test-double-libmvec-cbrt.c b/sysdeps/x86_64/fpu/test-double-libmvec-cbrt.c new file mode 100644 index 0000000000..fb3684b18c --- /dev/null +++ b/sysdeps/x86_64/fpu/test-double-libmvec-cbrt.c @@ -0,0 +1,3 @@ +#define LIBMVEC_TYPE double +#define LIBMVEC_FUNC cbrt +#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 db136cc901..b1981ac7e4 100644 --- a/sysdeps/x86_64/fpu/test-double-vlen2-wrappers.c +++ b/sysdeps/x86_64/fpu/test-double-vlen2-wrappers.c @@ -36,6 +36,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp10), _ZGVbN2v_exp10) VECTOR_WRAPPER (WRAPPER_NAME (cosh), _ZGVbN2v_cosh) VECTOR_WRAPPER (WRAPPER_NAME (expm1), _ZGVbN2v_expm1) VECTOR_WRAPPER (WRAPPER_NAME (sinh), _ZGVbN2v_sinh) +VECTOR_WRAPPER (WRAPPER_NAME (cbrt), _ZGVbN2v_cbrt) #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 5fc09ac8c0..47915a7e59 100644 --- a/sysdeps/x86_64/fpu/test-double-vlen4-avx2-wrappers.c +++ b/sysdeps/x86_64/fpu/test-double-vlen4-avx2-wrappers.c @@ -39,6 +39,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp10), _ZGVdN4v_exp10) VECTOR_WRAPPER (WRAPPER_NAME (cosh), _ZGVdN4v_cosh) VECTOR_WRAPPER (WRAPPER_NAME (expm1), _ZGVdN4v_expm1) VECTOR_WRAPPER (WRAPPER_NAME (sinh), _ZGVdN4v_sinh) +VECTOR_WRAPPER (WRAPPER_NAME (cbrt), _ZGVdN4v_cbrt) #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 26ef7fb365..5cd5049807 100644 --- a/sysdeps/x86_64/fpu/test-double-vlen4-wrappers.c +++ b/sysdeps/x86_64/fpu/test-double-vlen4-wrappers.c @@ -36,6 +36,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp10), _ZGVcN4v_exp10) VECTOR_WRAPPER (WRAPPER_NAME (cosh), _ZGVcN4v_cosh) VECTOR_WRAPPER (WRAPPER_NAME (expm1), _ZGVcN4v_expm1) VECTOR_WRAPPER (WRAPPER_NAME (sinh), _ZGVcN4v_sinh) +VECTOR_WRAPPER (WRAPPER_NAME (cbrt), _ZGVcN4v_cbrt) #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 c7055fca76..83970739ab 100644 --- a/sysdeps/x86_64/fpu/test-double-vlen8-wrappers.c +++ b/sysdeps/x86_64/fpu/test-double-vlen8-wrappers.c @@ -36,6 +36,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp10), _ZGVeN8v_exp10) VECTOR_WRAPPER (WRAPPER_NAME (cosh), _ZGVeN8v_cosh) VECTOR_WRAPPER (WRAPPER_NAME (expm1), _ZGVeN8v_expm1) VECTOR_WRAPPER (WRAPPER_NAME (sinh), _ZGVeN8v_sinh) +VECTOR_WRAPPER (WRAPPER_NAME (cbrt), _ZGVeN8v_cbrt) #ifndef __ILP32__ # define VEC_INT_TYPE __m512i diff --git a/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx.c b/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx.c new file mode 100644 index 0000000000..59b8d77f71 --- /dev/null +++ b/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx.c @@ -0,0 +1 @@ +#include "test-float-libmvec-cbrtf.c" diff --git a/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx2.c b/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx2.c new file mode 100644 index 0000000000..59b8d77f71 --- /dev/null +++ b/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx2.c @@ -0,0 +1 @@ +#include "test-float-libmvec-cbrtf.c" diff --git a/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx512f.c b/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx512f.c new file mode 100644 index 0000000000..59b8d77f71 --- /dev/null +++ b/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf-avx512f.c @@ -0,0 +1 @@ +#include "test-float-libmvec-cbrtf.c" diff --git a/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf.c b/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf.c new file mode 100644 index 0000000000..3a06ba79e0 --- /dev/null +++ b/sysdeps/x86_64/fpu/test-float-libmvec-cbrtf.c @@ -0,0 +1,3 @@ +#define LIBMVEC_TYPE float +#define LIBMVEC_FUNC cbrtf +#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 d353bcb0f2..0420f11c28 100644 --- a/sysdeps/x86_64/fpu/test-float-vlen16-wrappers.c +++ b/sysdeps/x86_64/fpu/test-float-vlen16-wrappers.c @@ -36,6 +36,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp10f), _ZGVeN16v_exp10f) VECTOR_WRAPPER (WRAPPER_NAME (coshf), _ZGVeN16v_coshf) VECTOR_WRAPPER (WRAPPER_NAME (expm1f), _ZGVeN16v_expm1f) VECTOR_WRAPPER (WRAPPER_NAME (sinhf), _ZGVeN16v_sinhf) +VECTOR_WRAPPER (WRAPPER_NAME (cbrtf), _ZGVeN16v_cbrtf) #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 5e59117626..c8f7580265 100644 --- a/sysdeps/x86_64/fpu/test-float-vlen4-wrappers.c +++ b/sysdeps/x86_64/fpu/test-float-vlen4-wrappers.c @@ -36,6 +36,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp10f), _ZGVbN4v_exp10f) VECTOR_WRAPPER (WRAPPER_NAME (coshf), _ZGVbN4v_coshf) VECTOR_WRAPPER (WRAPPER_NAME (expm1f), _ZGVbN4v_expm1f) VECTOR_WRAPPER (WRAPPER_NAME (sinhf), _ZGVbN4v_sinhf) +VECTOR_WRAPPER (WRAPPER_NAME (cbrtf), _ZGVbN4v_cbrtf) #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 e884a5f4df..b581796b88 100644 --- a/sysdeps/x86_64/fpu/test-float-vlen8-avx2-wrappers.c +++ b/sysdeps/x86_64/fpu/test-float-vlen8-avx2-wrappers.c @@ -39,6 +39,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp10f), _ZGVdN8v_exp10f) VECTOR_WRAPPER (WRAPPER_NAME (coshf), _ZGVdN8v_coshf) VECTOR_WRAPPER (WRAPPER_NAME (expm1f), _ZGVdN8v_expm1f) VECTOR_WRAPPER (WRAPPER_NAME (sinhf), _ZGVdN8v_sinhf) +VECTOR_WRAPPER (WRAPPER_NAME (cbrtf), _ZGVdN8v_cbrtf) /* 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 95910d39e9..f16789e5ff 100644 --- a/sysdeps/x86_64/fpu/test-float-vlen8-wrappers.c +++ b/sysdeps/x86_64/fpu/test-float-vlen8-wrappers.c @@ -36,6 +36,7 @@ VECTOR_WRAPPER (WRAPPER_NAME (exp10f), _ZGVcN8v_exp10f) VECTOR_WRAPPER (WRAPPER_NAME (coshf), _ZGVcN8v_coshf) VECTOR_WRAPPER (WRAPPER_NAME (expm1f), _ZGVcN8v_expm1f) VECTOR_WRAPPER (WRAPPER_NAME (sinhf), _ZGVcN8v_sinhf) +VECTOR_WRAPPER (WRAPPER_NAME (cbrtf), _ZGVcN8v_cbrtf) #define VEC_INT_TYPE __m128i