diff options
| author | Jakub Jelinek <jakub@redhat.com> | 2007-07-12 18:26:36 +0000 |
|---|---|---|
| committer | Jakub Jelinek <jakub@redhat.com> | 2007-07-12 18:26:36 +0000 |
| commit | 0ecb606cb6cf65de1d9fc8a919bceb4be476c602 (patch) | |
| tree | 2ea1f8305970753e4a657acb2ccc15ca3eec8e2c /sysdeps/ia64/fpu/e_sinhf.S | |
| parent | 7d58530341304d403a6626d7f7a1913165fe2f32 (diff) | |
| download | glibc-0ecb606cb6cf65de1d9fc8a919bceb4be476c602.tar.gz glibc-0ecb606cb6cf65de1d9fc8a919bceb4be476c602.tar.xz glibc-0ecb606cb6cf65de1d9fc8a919bceb4be476c602.zip | |
2.5-18.1
Diffstat (limited to 'sysdeps/ia64/fpu/e_sinhf.S')
| -rw-r--r-- | sysdeps/ia64/fpu/e_sinhf.S | 1614 |
1 files changed, 519 insertions, 1095 deletions
diff --git a/sysdeps/ia64/fpu/e_sinhf.S b/sysdeps/ia64/fpu/e_sinhf.S index d5aa2dca16..6d808cb478 100644 --- a/sysdeps/ia64/fpu/e_sinhf.S +++ b/sysdeps/ia64/fpu/e_sinhf.S @@ -1,10 +1,10 @@ .file "sinhf.s" -// Copyright (C) 2000, 2001, Intel Corporation + +// Copyright (c) 2000 - 2005, Intel Corporation // All rights reserved. -// -// Contributed 2/2/2000 by John Harrison, Ted Kubaska, Bob Norin, Shane Story, -// and Ping Tak Peter Tang of the Computational Software Lab, Intel Corporation. +// +// Contributed 2000 by the Intel Numerics Group, Intel Corporation // // Redistribution and use in source and binary forms, with or without // modification, are permitted provided that the following conditions are @@ -20,1305 +20,729 @@ // * The name of Intel Corporation may not be used to endorse or promote // products derived from this software without specific prior written // permission. -// -// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS -// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT + +// THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS +// "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT // LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR -// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS +// A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL OR ITS // CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, -// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, -// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR -// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY +// EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, +// PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR +// PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY // OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY OR TORT (INCLUDING -// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS -// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. -// -// Intel Corporation is the author of this code, and requests that all -// problem reports or change requests be submitted to it directly at -// http://developer.intel.com/opensource. +// NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS +// SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. // +// Intel Corporation is the author of this code, and requests that all +// problem reports or change requests be submitted to it directly at +// http://www.intel.com/software/products/opensource/libraries/num.htm. + // History -//============================================================== -// 2/02/00 Initial version -// 4/04/00 Unwind support added -// 8/15/00 Bundle added after call to __libm_error_support to properly +//********************************************************************* +// 02/02/00 Initial version +// 04/04/00 Unwind support added +// 08/15/00 Bundle added after call to __libm_error_support to properly // set [the previously overwritten] GR_Parameter_RESULT. // 10/12/00 Update to set denormal operand and underflow flags -// 1/22/01 Fixed to set inexact flag for small args. +// 01/22/01 Fixed to set inexact flag for small args. +// 05/02/01 Reworked to improve speed of all paths +// 05/20/02 Cleaned up namespace and sf0 syntax +// 11/20/02 Improved algorithm based on expf +// 03/31/05 Reformatted delimiters between data tables // // API -//============================================================== -// float = sinhf(float) -// input floating point f8 -// output floating point f8 -// -// Registers used -//============================================================== -// general registers: -// r32 -> r47 -// predicate registers used: -// p6 p7 p8 p9 -// floating-point registers used: -// f9 -> f15; f32 -> f45; -// f8 has input, then output +//********************************************************************* +// float sinhf(float) // // Overview of operation -//============================================================== -// There are four paths -// 1. |x| < 0.25 SINH_BY_POLY -// 2. |x| < 32 SINH_BY_TBL -// 3. |x| < 2^14 SINH_BY_EXP -// 4. |x_ >= 2^14 SINH_HUGE -// -// For double extended we get infinity for x >= 400c b174 ddc0 31ae c0ea -// >= 1.0110001.... x 2^13 -// >= 11357.2166 +//********************************************************************* +// Case 1: 0 < |x| < 2^-60 +// Result = x, computed by x+sgn(x)*x^2) to handle flags and rounding // -// But for double we get infinity for x >= 408633ce8fb9f87e -// >= 1.0110...x 2^9 -// >= +7.10476e+002 +// Case 2: 2^-60 < |x| < 0.25 +// Evaluate sinh(x) by a 9th order polynomial +// Care is take for the order of multiplication; and A2 is not exactly 1/5!, +// A3 is not exactly 1/7!, etc. +// sinh(x) = x + (A1*x^3 + A2*x^5 + A3*x^7 + A4*x^9) // -// And for single we get infinity for x >= 42b3a496 -// >= 1.0110... 2^6 -// >= 89.8215 +// Case 3: 0.25 < |x| < 89.41598 +// Algorithm is based on the identity sinh(x) = ( exp(x) - exp(-x) ) / 2. +// The algorithm for exp is described as below. There are a number of +// economies from evaluating both exp(x) and exp(-x). Although we +// are evaluating both quantities, only where the quantities diverge do we +// duplicate the computations. The basic algorithm for exp(x) is described +// below. // -// SAFE: If there is danger of overflow set SAFE to 0 -// NOT implemented: if there is danger of underflow, set SAFE to 0 -// SAFE for all paths listed below -// -// 1. SINH_BY_POLY -// =============== -// If |x| is less than the tiny threshold, then clear SAFE -// For double, the tiny threshold is -1022 = -0x3fe => -3fe + ffff = fc01 -// register-biased, this is fc01 -// For single, the tiny threshold is -126 = -7e => -7e + ffff = ff81 -// If |x| < tiny threshold, set SAFE = 0 -// -// 2. SINH_BY_TBL -// ============= -// SAFE: SAFE is always 1 for TBL; +// Take the input x. w is "how many log2/128 in x?" +// w = x * 64/log2 +// NJ = int(w) +// x = NJ*log2/64 + R + +// NJ = 64*n + j +// x = n*log2 + (log2/64)*j + R // -// 3. SINH_BY_EXP -// ============== -// There is a danger of double-extended overflow if N-1 > 16382 = 0x3ffe -// r34 has N-1; 16382 is in register biased form, 0x13ffd -// There is danger of double overflow if N-1 > 0x3fe -// in register biased form, 0x103fd -// Analagously, there is danger of single overflow if N-1 > 0x7e -// in register biased form, 0x1007d -// SAFE: If there is danger of overflow set SAFE to 0 +// So, exp(x) = 2^n * 2^(j/64)* exp(R) // -// 4. SINH_HUGE -// ============ -// SAFE: SAFE is always 0 for HUGE +// T = 2^n * 2^(j/64) +// Construct 2^n +// Get 2^(j/64) table +// actually all the entries of 2^(j/64) table are stored in DP and +// with exponent bits set to 0 -> multiplication on 2^n can be +// performed by doing logical "or" operation with bits presenting 2^n + +// exp(R) = 1 + (exp(R) - 1) +// P = exp(R) - 1 approximated by Taylor series of 3rd degree +// P = A3*R^3 + A2*R^2 + R, A3 = 1/6, A2 = 1/2 // -#include "libm_support.h" - -// Assembly macros -//============================================================== -sinh_FR_X = f44 -sinh_FR_X2 = f9 -sinh_FR_X4 = f10 -sinh_FR_SGNX = f40 -sinh_FR_all_ones = f45 -sinh_FR_tmp = f42 - -sinh_FR_Inv_log2by64 = f9 -sinh_FR_log2by64_lo = f11 -sinh_FR_log2by64_hi = f10 - -sinh_FR_A1 = f9 -sinh_FR_A2 = f10 -sinh_FR_A3 = f11 - -sinh_FR_Rcub = f12 -sinh_FR_M_temp = f13 -sinh_FR_R_temp = f13 -sinh_FR_Rsq = f13 -sinh_FR_R = f14 - -sinh_FR_M = f38 - -sinh_FR_B1 = f15 -sinh_FR_B2 = f32 -sinh_FR_B3 = f33 +// The final result is reconstructed as follows +// exp(x) = T + T*P -sinh_FR_peven_temp1 = f34 -sinh_FR_peven_temp2 = f35 -sinh_FR_peven = f36 +// Special values +//********************************************************************* +// sinhf(+0) = +0 +// sinhf(-0) = -0 -sinh_FR_podd_temp1 = f34 -sinh_FR_podd_temp2 = f35 -sinh_FR_podd = f37 +// sinhf(+qnan) = +qnan +// sinhf(-qnan) = -qnan +// sinhf(+snan) = +qnan +// sinhf(-snan) = -qnan -sinh_FR_poly_podd_temp1 = f11 -sinh_FR_poly_podd_temp2 = f13 -sinh_FR_poly_peven_temp1 = f11 -sinh_FR_poly_peven_temp2 = f13 +// sinhf(-inf) = -inf +// sinhf(+inf) = +inf -sinh_FR_J_temp = f9 -sinh_FR_J = f10 - -sinh_FR_Mmj = f39 - -sinh_FR_N_temp1 = f11 -sinh_FR_N_temp2 = f12 -sinh_FR_N = f13 - -sinh_FR_spos = f14 -sinh_FR_sneg = f15 - -sinh_FR_Tjhi = f32 -sinh_FR_Tjlo = f33 -sinh_FR_Tmjhi = f34 -sinh_FR_Tmjlo = f35 - -sinh_GR_mJ = r35 -sinh_GR_J = r36 - -sinh_AD_mJ = r38 -sinh_AD_J = r39 -sinh_GR_all_ones = r40 - -sinh_FR_S_hi = f9 -sinh_FR_S_hi_temp = f10 -sinh_FR_S_lo_temp1 = f11 -sinh_FR_S_lo_temp2 = f12 -sinh_FR_S_lo_temp3 = f13 - -sinh_FR_S_lo = f38 -sinh_FR_C_hi = f39 - -sinh_FR_C_hi_temp1 = f10 -sinh_FR_Y_hi = f11 -sinh_FR_Y_lo_temp = f12 -sinh_FR_Y_lo = f13 -sinh_FR_SINH = f9 - -sinh_FR_P1 = f14 -sinh_FR_P2 = f15 -sinh_FR_P3 = f32 -sinh_FR_P4 = f33 -sinh_FR_P5 = f34 -sinh_FR_P6 = f35 - -sinh_FR_TINY_THRESH = f9 - -sinh_FR_SINH_temp = f10 -sinh_FR_SCALE = f11 - -sinh_FR_signed_hi_lo = f10 - - -GR_SAVE_PFS = r41 -GR_SAVE_B0 = r42 -GR_SAVE_GP = r43 +// Overflow and Underflow +//********************************************************************* +// sinhf(x) = largest single normal when +// x = 89.41598 = 0x42b2d4fc +// +// Underflow is handled as described in case 1 above -GR_Parameter_X = r44 -GR_Parameter_Y = r45 -GR_Parameter_RESULT = r46 +// Registers used +//********************************************************************* +// Floating Point registers used: +// f8 input, output +// f6,f7, f9 -> f15, f32 -> f45 -// Data tables -//============================================================== +// General registers used: +// r2, r3, r16 -> r38 -#ifdef _LIBC -.rodata -#else -.data -#endif +// Predicate registers used: +// p6 -> p15 +// Assembly macros +//********************************************************************* +// integer registers used +// scratch +rNJ = r2 +rNJ_neg = r3 + +rJ_neg = r16 +rN_neg = r17 +rSignexp_x = r18 +rExp_x = r18 +rExp_mask = r19 +rExp_bias = r20 +rAd1 = r21 +rAd2 = r22 +rJ = r23 +rN = r24 +rTblAddr = r25 +rA3 = r26 +rExpHalf = r27 +rLn2Div64 = r28 +rGt_ln = r29 +r17ones_m1 = r29 +rRightShifter = r30 +rJ_mask = r30 +r64DivLn2 = r31 +rN_mask = r31 +// stacked +GR_SAVE_PFS = r32 +GR_SAVE_B0 = r33 +GR_SAVE_GP = r34 +GR_Parameter_X = r35 +GR_Parameter_Y = r36 +GR_Parameter_RESULT = r37 +GR_Parameter_TAG = r38 + +// floating point registers used +FR_X = f10 +FR_Y = f1 +FR_RESULT = f8 +// scratch +fRightShifter = f6 +f64DivLn2 = f7 +fNormX = f9 +fNint = f10 +fN = f11 +fR = f12 +fLn2Div64 = f13 +fA2 = f14 +fA3 = f15 +// stacked +fP = f32 +fT = f33 +fMIN_SGL_OFLOW_ARG = f34 +fMAX_SGL_NORM_ARG = f35 +fRSqr = f36 +fA1 = f37 +fA21 = f37 +fA4 = f38 +fA43 = f38 +fA4321 = f38 +fX4 = f39 +fTmp = f39 +fGt_pln = f39 +fWre_urm_f8 = f40 +fXsq = f40 +fP_neg = f41 +fX3 = f41 +fT_neg = f42 +fExp = f43 +fExp_neg = f44 +fAbsX = f45 + + +RODATA .align 16 -double_sinh_arg_reduction: -ASM_TYPE_DIRECTIVE(double_sinh_arg_reduction,@object) - data8 0xB8AA3B295C17F0BC, 0x00004005 - data8 0xB17217F7D1000000, 0x00003FF8 - data8 0xCF79ABC9E3B39804, 0x00003FD0 -ASM_SIZE_DIRECTIVE(double_sinh_arg_reduction) - -double_sinh_p_table: -ASM_TYPE_DIRECTIVE(double_sinh_p_table,@object) - data8 0xAAAAAAAAAAAAAAAB, 0x00003FFC - data8 0x8888888888888412, 0x00003FF8 - data8 0xD00D00D00D4D39F2, 0x00003FF2 - data8 0xB8EF1D28926D8891, 0x00003FEC - data8 0xD732377688025BE9, 0x00003FE5 - data8 0xB08AF9AE78C1239F, 0x00003FDE -ASM_SIZE_DIRECTIVE(double_sinh_p_table) - -double_sinh_ab_table: -ASM_TYPE_DIRECTIVE(double_sinh_ab_table,@object) - data8 0xAAAAAAAAAAAAAAAC, 0x00003FFC - data8 0x88888888884ECDD5, 0x00003FF8 - data8 0xD00D0C6DCC26A86B, 0x00003FF2 - data8 0x8000000000000002, 0x00003FFE - data8 0xAAAAAAAAAA402C77, 0x00003FFA - data8 0xB60B6CC96BDB144D, 0x00003FF5 -ASM_SIZE_DIRECTIVE(double_sinh_ab_table) - -double_sinh_j_table: -ASM_TYPE_DIRECTIVE(double_sinh_j_table,@object) - data8 0xB504F333F9DE6484, 0x00003FFE, 0x1EB2FB13, 0x00000000 - data8 0xB6FD91E328D17791, 0x00003FFE, 0x1CE2CBE2, 0x00000000 - data8 0xB8FBAF4762FB9EE9, 0x00003FFE, 0x1DDC3CBC, 0x00000000 - data8 0xBAFF5AB2133E45FB, 0x00003FFE, 0x1EE9AA34, 0x00000000 - data8 0xBD08A39F580C36BF, 0x00003FFE, 0x9EAEFDC1, 0x00000000 - data8 0xBF1799B67A731083, 0x00003FFE, 0x9DBF517B, 0x00000000 - data8 0xC12C4CCA66709456, 0x00003FFE, 0x1EF88AFB, 0x00000000 - data8 0xC346CCDA24976407, 0x00003FFE, 0x1E03B216, 0x00000000 - data8 0xC5672A115506DADD, 0x00003FFE, 0x1E78AB43, 0x00000000 - data8 0xC78D74C8ABB9B15D, 0x00003FFE, 0x9E7B1747, 0x00000000 - data8 0xC9B9BD866E2F27A3, 0x00003FFE, 0x9EFE3C0E, 0x00000000 - data8 0xCBEC14FEF2727C5D, 0x00003FFE, 0x9D36F837, 0x00000000 - data8 0xCE248C151F8480E4, 0x00003FFE, 0x9DEE53E4, 0x00000000 - data8 0xD06333DAEF2B2595, 0x00003FFE, 0x9E24AE8E, 0x00000000 - data8 0xD2A81D91F12AE45A, 0x00003FFE, 0x1D912473, 0x00000000 - data8 0xD4F35AABCFEDFA1F, 0x00003FFE, 0x1EB243BE, 0x00000000 - data8 0xD744FCCAD69D6AF4, 0x00003FFE, 0x1E669A2F, 0x00000000 - data8 0xD99D15C278AFD7B6, 0x00003FFE, 0x9BBC610A, 0x00000000 - data8 0xDBFBB797DAF23755, 0x00003FFE, 0x1E761035, 0x00000000 - data8 0xDE60F4825E0E9124, 0x00003FFE, 0x9E0BE175, 0x00000000 - data8 0xE0CCDEEC2A94E111, 0x00003FFE, 0x1CCB12A1, 0x00000000 - data8 0xE33F8972BE8A5A51, 0x00003FFE, 0x1D1BFE90, 0x00000000 - data8 0xE5B906E77C8348A8, 0x00003FFE, 0x1DF2F47A, 0x00000000 - data8 0xE8396A503C4BDC68, 0x00003FFE, 0x1EF22F22, 0x00000000 - data8 0xEAC0C6E7DD24392F, 0x00003FFE, 0x9E3F4A29, 0x00000000 - data8 0xED4F301ED9942B84, 0x00003FFE, 0x1EC01A5B, 0x00000000 - data8 0xEFE4B99BDCDAF5CB, 0x00003FFE, 0x1E8CAC3A, 0x00000000 - data8 0xF281773C59FFB13A, 0x00003FFE, 0x9DBB3FAB, 0x00000000 - data8 0xF5257D152486CC2C, 0x00003FFE, 0x1EF73A19, 0x00000000 - data8 0xF7D0DF730AD13BB9, 0x00003FFE, 0x9BB795B5, 0x00000000 - data8 0xFA83B2DB722A033A, 0x00003FFE, 0x1EF84B76, 0x00000000 - data8 0xFD3E0C0CF486C175, 0x00003FFE, 0x9EF5818B, 0x00000000 - data8 0x8000000000000000, 0x00003FFF, 0x00000000, 0x00000000 - data8 0x8164D1F3BC030773, 0x00003FFF, 0x1F77CACA, 0x00000000 - data8 0x82CD8698AC2BA1D7, 0x00003FFF, 0x1EF8A91D, 0x00000000 - data8 0x843A28C3ACDE4046, 0x00003FFF, 0x1E57C976, 0x00000000 - data8 0x85AAC367CC487B15, 0x00003FFF, 0x9EE8DA92, 0x00000000 - data8 0x871F61969E8D1010, 0x00003FFF, 0x1EE85C9F, 0x00000000 - data8 0x88980E8092DA8527, 0x00003FFF, 0x1F3BF1AF, 0x00000000 - data8 0x8A14D575496EFD9A, 0x00003FFF, 0x1D80CA1E, 0x00000000 - data8 0x8B95C1E3EA8BD6E7, 0x00003FFF, 0x9D0373AF, 0x00000000 - data8 0x8D1ADF5B7E5BA9E6, 0x00003FFF, 0x9F167097, 0x00000000 - data8 0x8EA4398B45CD53C0, 0x00003FFF, 0x1EB70051, 0x00000000 - data8 0x9031DC431466B1DC, 0x00003FFF, 0x1F6EB029, 0x00000000 - data8 0x91C3D373AB11C336, 0x00003FFF, 0x1DFD6D8E, 0x00000000 - data8 0x935A2B2F13E6E92C, 0x00003FFF, 0x9EB319B0, 0x00000000 - data8 0x94F4EFA8FEF70961, 0x00003FFF, 0x1EBA2BEB, 0x00000000 - data8 0x96942D3720185A00, 0x00003FFF, 0x1F11D537, 0x00000000 - data8 0x9837F0518DB8A96F, 0x00003FFF, 0x1F0D5A46, 0x00000000 - data8 0x99E0459320B7FA65, 0x00003FFF, 0x9E5E7BCA, 0x00000000 - data8 0x9B8D39B9D54E5539, 0x00003FFF, 0x9F3AAFD1, 0x00000000 - data8 0x9D3ED9A72CFFB751, 0x00003FFF, 0x9E86DACC, 0x00000000 - data8 0x9EF5326091A111AE, 0x00003FFF, 0x9F3EDDC2, 0x00000000 - data8 0xA0B0510FB9714FC2, 0x00003FFF, 0x1E496E3D, 0x00000000 - data8 0xA27043030C496819, 0x00003FFF, 0x9F490BF6, 0x00000000 - data8 0xA43515AE09E6809E, 0x00003FFF, 0x1DD1DB48, 0x00000000 - data8 0xA5FED6A9B15138EA, 0x00003FFF, 0x1E65EBFB, 0x00000000 - data8 0xA7CD93B4E965356A, 0x00003FFF, 0x9F427496, 0x00000000 - data8 0xA9A15AB4EA7C0EF8, 0x00003FFF, 0x1F283C4A, 0x00000000 - data8 0xAB7A39B5A93ED337, 0x00003FFF, 0x1F4B0047, 0x00000000 - data8 0xAD583EEA42A14AC6, 0x00003FFF, 0x1F130152, 0x00000000 - data8 0xAF3B78AD690A4375, 0x00003FFF, 0x9E8367C0, 0x00000000 - data8 0xB123F581D2AC2590, 0x00003FFF, 0x9F705F90, 0x00000000 - data8 0xB311C412A9112489, 0x00003FFF, 0x1EFB3C53, 0x00000000 - data8 0xB504F333F9DE6484, 0x00003FFF, 0x1F32FB13, 0x00000000 -ASM_SIZE_DIRECTIVE(double_sinh_j_table) - -.align 32 -.global sinhf# - -.section .text -.proc sinhf# -.align 32 - -sinhf: -#ifdef _LIBC -.global __ieee754_sinhf -.type __ieee754_sinhf,@function -__ieee754_sinhf: -#endif - -// X infinity or NAN? -// Take invalid fault if enabled - -{ .mfi - alloc r32 = ar.pfs,0,12,4,0 -(p0) fclass.m.unc p6,p0 = f8, 0xe3 //@qnan | @snan | @inf - mov sinh_GR_all_ones = -1 -} -;; +LOCAL_OBJECT_START(_sinhf_table) +data4 0x42b2d4fd // Smallest single arg to overflow single result +data4 0x42b2d4fc // Largest single arg to give normal single result +data4 0x00000000 // pad +data4 0x00000000 // pad +// +// 2^(j/64) table, j goes from 0 to 63 +data8 0x0000000000000000 // 2^(0/64) +data8 0x00002C9A3E778061 // 2^(1/64) +data8 0x000059B0D3158574 // 2^(2/64) +data8 0x0000874518759BC8 // 2^(3/64) +data8 0x0000B5586CF9890F // 2^(4/64) +data8 0x0000E3EC32D3D1A2 // 2^(5/64) +data8 0x00011301D0125B51 // 2^(6/64) +data8 0x0001429AAEA92DE0 // 2^(7/64) +data8 0x000172B83C7D517B // 2^(8/64) +data8 0x0001A35BEB6FCB75 // 2^(9/64) +data8 0x0001D4873168B9AA // 2^(10/64) +data8 0x0002063B88628CD6 // 2^(11/64) +data8 0x0002387A6E756238 // 2^(12/64) +data8 0x00026B4565E27CDD // 2^(13/64) +data8 0x00029E9DF51FDEE1 // 2^(14/64) +data8 0x0002D285A6E4030B // 2^(15/64) +data8 0x000306FE0A31B715 // 2^(16/64) +data8 0x00033C08B26416FF // 2^(17/64) +data8 0x000371A7373AA9CB // 2^(18/64) +data8 0x0003A7DB34E59FF7 // 2^(19/64) +data8 0x0003DEA64C123422 // 2^(20/64) +data8 0x0004160A21F72E2A // 2^(21/64) +data8 0x00044E086061892D // 2^(22/64) +data8 0x000486A2B5C13CD0 // 2^(23/64) +data8 0x0004BFDAD5362A27 // 2^(24/64) +data8 0x0004F9B2769D2CA7 // 2^(25/64) +data8 0x0005342B569D4F82 // 2^(26/64) +data8 0x00056F4736B527DA // 2^(27/64) +data8 0x0005AB07DD485429 // 2^(28/64) +data8 0x0005E76F15AD2148 // 2^(29/64) +data8 0x0006247EB03A5585 // 2^(30/64) +data8 0x0006623882552225 // 2^(31/64) +data8 0x0006A09E667F3BCD // 2^(32/64) +data8 0x0006DFB23C651A2F // 2^(33/64) +data8 0x00071F75E8EC5F74 // 2^(34/64) +data8 0x00075FEB564267C9 // 2^(35/64) +data8 0x0007A11473EB0187 // 2^(36/64) +data8 0x0007E2F336CF4E62 // 2^(37/64) +data8 0x00082589994CCE13 // 2^(38/64) +data8 0x000868D99B4492ED // 2^(39/64) +data8 0x0008ACE5422AA0DB // 2^(40/64) +data8 0x0008F1AE99157736 // 2^(41/64) +data8 0x00093737B0CDC5E5 // 2^(42/64) +data8 0x00097D829FDE4E50 // 2^(43/64) +data8 0x0009C49182A3F090 // 2^(44/64) +data8 0x000A0C667B5DE565 // 2^(45/64) +data8 0x000A5503B23E255D // 2^(46/64) +data8 0x000A9E6B5579FDBF // 2^(47/64) +data8 0x000AE89F995AD3AD // 2^(48/64) +data8 0x000B33A2B84F15FB // 2^(49/64) +data8 0x000B7F76F2FB5E47 // 2^(50/64) +data8 0x000BCC1E904BC1D2 // 2^(51/64) +data8 0x000C199BDD85529C // 2^(52/64) +data8 0x000C67F12E57D14B // 2^(53/64) +data8 0x000CB720DCEF9069 // 2^(54/64) +data8 0x000D072D4A07897C // 2^(55/64) +data8 0x000D5818DCFBA487 // 2^(56/64) +data8 0x000DA9E603DB3285 // 2^(57/64) +data8 0x000DFC97337B9B5F // 2^(58/64) +data8 0x000E502EE78B3FF6 // 2^(59/64) +data8 0x000EA4AFA2A490DA // 2^(60/64) +data8 0x000EFA1BEE615A27 // 2^(61/64) +data8 0x000F50765B6E4540 // 2^(62/64) +data8 0x000FA7C1819E90D8 // 2^(63/64) +LOCAL_OBJECT_END(_sinhf_table) + +LOCAL_OBJECT_START(sinh_p_table) +data8 0x3ec749d84bc96d7d // A4 +data8 0x3f2a0168d09557cf // A3 +data8 0x3f811111326ed15a // A2 +data8 0x3fc55555552ed1e2 // A1 +LOCAL_OBJECT_END(sinh_p_table) -{ .mfb - nop.m 999 -(p6) fma.s.s0 f8 = f8,f1,f8 -(p6) br.ret.spnt b0 ;; -} +.section .text +GLOBAL_IEEE754_ENTRY(sinhf) -// Put 0.25 in f9; p6 true if x < 0.25 -// Make constant that will generate inexact when squared { .mlx - setf.sig sinh_FR_all_ones = sinh_GR_all_ones -(p0) movl r32 = 0x000000000000fffd ;; -} - -{ .mfi -(p0) setf.exp f9 = r32 -(p0) fclass.m.unc p7,p0 = f8, 0x07 //@zero - nop.i 999 ;; -} - -{ .mfb - nop.m 999 -(p0) fmerge.s sinh_FR_X = f0,f8 -(p7) br.ret.spnt b0 ;; + getf.exp rSignexp_x = f8 // Must recompute if x unorm + movl r64DivLn2 = 0x40571547652B82FE // 64/ln(2) } - -// Identify denormal operands. -{ .mfi - nop.m 999 - fclass.m.unc p10,p0 = f8, 0x09 // + denorm - nop.i 999 -};; -{ .mfi - nop.m 999 - fclass.m.unc p11,p0 = f8, 0x0a // - denorm - nop.i 999 -} - -{ .mfi - nop.m 999 -(p0) fmerge.s sinh_FR_SGNX = f8,f1 - nop.i 999 ;; -} - -{ .mfi - nop.m 999 -(p0) fcmp.lt.unc.s1 p0,p7 = sinh_FR_X,f9 - nop.i 999 ;; -} - -{ .mib - nop.m 999 - nop.i 999 -(p7) br.cond.sptk L(SINH_BY_TBL) ;; -} - - -L(SINH_BY_POLY): - -// POLY cannot overflow so there is no need to call __libm_error_support -// Set tiny_SAFE (p7) to 1(0) if answer is not tiny -// Currently we do not use tiny_SAFE. So the setting of tiny_SAFE is -// commented out. -//(p0) movl r32 = 0x000000000000fc01 -//(p0) setf.exp f10 = r32 -//(p0) fcmp.lt.unc.s1 p6,p7 = f8,f10 -// Here is essentially the algorithm for SINH_BY_POLY. Care is take for the order -// of multiplication; and P_1 is not exactly 1/3!, P_2 is not exactly 1/5!, etc. -// Note that ax = |x| -// sinh(x) = sign * (series(e^x) - series(e^-x))/2 -// = sign * (ax + ax^3/3! + ax^5/5! + ax^7/7! + ax^9/9! + ax^11/11! + ax^13/13!) -// = sign * (ax + ax * ( ax^2 * (1/3! + ax^4 * (1/7! + ax^4*1/11!)) ) -// + ax * ( ax^4 * (1/5! + ax^4 * (1/9! + ax^4*1/13!)) ) ) -// = sign * (ax + ax*p_odd + (ax*p_even)) -// = sign * (ax + Y_lo) -// sinh(x) = sign * (Y_hi + Y_lo) -// Get the values of P_x from the table -{ .mfb -(p0) addl r34 = @ltoff(double_sinh_p_table), gp -(p10) fma.s.s0 f8 = f8,f8,f8 -(p10) br.ret.spnt b0 -} -;; - -{ .mfb - ld8 r34 = [r34] -(p11) fnma.s.s0 f8 = f8,f8,f8 -(p11) br.ret.spnt b0 +{ .mlx + addl rTblAddr = @ltoff(_sinhf_table),gp + movl rRightShifter = 0x43E8000000000000 // DP Right Shifter } ;; -// Calculate sinh_FR_X2 = ax*ax and sinh_FR_X4 = ax*ax*ax*ax -{ .mmf - nop.m 999 -(p0) ldfe sinh_FR_P1 = [r34],16 -(p0) fma.s1 sinh_FR_X2 = sinh_FR_X, sinh_FR_X, f0 ;; -} - -{ .mmi -(p0) ldfe sinh_FR_P2 = [r34],16 ;; -(p0) ldfe sinh_FR_P3 = [r34],16 - nop.i 999 ;; -} - -{ .mmi -(p0) ldfe sinh_FR_P4 = [r34],16 ;; -(p0) ldfe sinh_FR_P5 = [r34],16 - nop.i 999 ;; -} - -{ .mfi -(p0) ldfe sinh_FR_P6 = [r34],16 -(p0) fma.s1 sinh_FR_X4 = sinh_FR_X2, sinh_FR_X2, f0 - nop.i 999 ;; -} - -// Calculate sinh_FR_podd = p_odd and sinh_FR_peven = p_even { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_poly_podd_temp1 = sinh_FR_X4, sinh_FR_P5, sinh_FR_P3 - nop.i 999 ;; + // point to the beginning of the table + ld8 rTblAddr = [rTblAddr] + fclass.m p6, p0 = f8, 0x0b // Test for x=unorm + addl rA3 = 0x3E2AA, r0 // high bits of 1.0/6.0 rounded to SP } - { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_poly_podd_temp2 = sinh_FR_X4, sinh_FR_poly_podd_temp1, sinh_FR_P1 - nop.i 999 + nop.m 0 + fnorm.s1 fNormX = f8 // normalized x + addl rExpHalf = 0xFFFE, r0 // exponent of 1/2 } +;; { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_poly_peven_temp1 = sinh_FR_X4, sinh_FR_P6, sinh_FR_P4 - nop.i 999 ;; + setf.d f64DivLn2 = r64DivLn2 // load 64/ln(2) to FP reg + fclass.m p15, p0 = f8, 0x1e3 // test for NaT,NaN,Inf + nop.i 0 } - -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_podd = sinh_FR_X2, sinh_FR_poly_podd_temp2, f0 - nop.i 999 +{ .mlx + // load Right Shifter to FP reg + setf.d fRightShifter = rRightShifter + movl rLn2Div64 = 0x3F862E42FEFA39EF // DP ln(2)/64 in GR } +;; { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_poly_peven_temp2 = sinh_FR_X4, sinh_FR_poly_peven_temp1, sinh_FR_P2 - nop.i 999 ;; + mov rExp_mask = 0x1ffff + fcmp.eq.s1 p13, p0 = f0, f8 // test for x = 0.0 + shl rA3 = rA3, 12 // 0x3E2AA000, approx to 1.0/6.0 in SP } - -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_peven = sinh_FR_X4, sinh_FR_poly_peven_temp2, f0 - nop.i 999 ;; +{ .mfb + nop.m 0 + nop.f 0 +(p6) br.cond.spnt SINH_UNORM // Branch if x=unorm } +;; -// Calculate sinh_FR_Y_lo = ax*p_odd + (ax*p_even) +SINH_COMMON: { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_Y_lo_temp = sinh_FR_X, sinh_FR_peven, f0 - nop.i 999 ;; + setf.exp fA2 = rExpHalf // load A2 to FP reg + nop.f 0 + mov rExp_bias = 0xffff } - -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_Y_lo = sinh_FR_X, sinh_FR_podd, sinh_FR_Y_lo_temp - nop.i 999 ;; +{ .mfb + setf.d fLn2Div64 = rLn2Div64 // load ln(2)/64 to FP reg +(p15) fma.s.s0 f8 = f8, f1, f0 // result if x = NaT,NaN,Inf +(p15) br.ret.spnt b0 // exit here if x = NaT,NaN,Inf } +;; -// Calculate sinh_FR_SINH = Y_hi + Y_lo. Note that ax = Y_hi { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_SINH = sinh_FR_X, f1, sinh_FR_Y_lo - nop.i 999 ;; + // min overflow and max normal threshold + ldfps fMIN_SGL_OFLOW_ARG, fMAX_SGL_NORM_ARG = [rTblAddr], 8 + nop.f 0 + and rExp_x = rExp_mask, rSignexp_x // Biased exponent of x } -// Dummy multiply to generate inexact -{ .mfi - nop.m 999 -(p0) fmpy.s0 sinh_FR_tmp = sinh_FR_all_ones, sinh_FR_all_ones - nop.i 999 -} - -// Calculate f8 = sign * (Y_hi + Y_lo) -// Go to return { .mfb - nop.m 999 -(p0) fma.s.s0 f8 = sinh_FR_SGNX,sinh_FR_SINH,f0 -(p0) br.ret.sptk b0 ;; -} - - -L(SINH_BY_TBL): - -// Now that we are at TBL; so far all we know is that |x| >= 0.25. -// The first two steps are the same for TBL and EXP, but if we are HUGE -// we want to leave now. -// Double-extended: -// Go to HUGE if |x| >= 2^14, 1000d (register-biased) is e = 14 (true) -// Double -// Go to HUGE if |x| >= 2^10, 10009 (register-biased) is e = 10 (true) -// Single -// Go to HUGE if |x| >= 2^7, 10006 (register-biased) is e = 7 (true) - -{ .mlx - nop.m 999 -(p0) movl r32 = 0x0000000000010006 ;; + setf.s fA3 = rA3 // load A3 to FP reg + nop.f 0 +(p13) br.ret.spnt b0 // exit here if x=0.0, return x } +;; { .mfi -(p0) setf.exp f9 = r32 - nop.f 999 - nop.i 999 ;; + sub rExp_x = rExp_x, rExp_bias // True exponent of x + fmerge.s fAbsX = f0, fNormX // Form |x| + nop.i 0 } +;; { .mfi - nop.m 999 -(p0) fcmp.ge.unc.s1 p6,p7 = sinh_FR_X,f9 - nop.i 999 ;; + nop.m 0 + // x*(64/ln(2)) + Right Shifter + fma.s1 fNint = fNormX, f64DivLn2, fRightShifter + add rTblAddr = 8, rTblAddr } - -{ .mib - nop.m 999 - nop.i 999 -(p6) br.cond.spnt L(SINH_HUGE) ;; -} - -// r32 = 1 -// r34 = N-1 -// r35 = N -// r36 = j -// r37 = N+1 - -// TBL can never overflow -// sinh(x) = sinh(B+R) -// = sinh(B)cosh(R) + cosh(B)sinh(R) -// -// ax = |x| = M*log2/64 + R -// B = M*log2/64 -// M = 64*N + j -// We will calcualte M and get N as (M-j)/64 -// The division is a shift. -// exp(B) = exp(N*log2 + j*log2/64) -// = 2^N * 2^(j*log2/64) -// sinh(B) = 1/2(e^B -e^-B) -// = 1/2(2^N * 2^(j*log2/64) - 2^-N * 2^(-j*log2/64)) -// sinh(B) = (2^(N-1) * 2^(j*log2/64) - 2^(-N-1) * 2^(-j*log2/64)) -// cosh(B) = (2^(N-1) * 2^(j*log2/64) + 2^(-N-1) * 2^(-j*log2/64)) -// 2^(j*log2/64) is stored as Tjhi + Tjlo , j= -32,....,32 -// Tjhi is double-extended (80-bit) and Tjlo is single(32-bit) -// R = ax - M*log2/64 -// R = ax - M*log2_by_64_hi - M*log2_by_64_lo -// exp(R) = 1 + R +R^2(1/2! + R(1/3! + R(1/4! + ... + R(1/n!)...) -// = 1 + p_odd + p_even -// where the p_even uses the A coefficients and the p_even uses the B coefficients -// So sinh(R) = 1 + p_odd + p_even -(1 -p_odd -p_even)/2 = p_odd -// cosh(R) = 1 + p_even -// sinh(B) = S_hi + S_lo -// cosh(B) = C_hi -// sinh(x) = sinh(B)cosh(R) + cosh(B)sinh(R) -// ****************************************************** -// STEP 1 (TBL and EXP) -// ****************************************************** -// Get the following constants. -// f9 = Inv_log2by64 -// f10 = log2by64_hi -// f11 = log2by64_lo - -{ .mmi -(p0) adds r32 = 0x1,r0 -(p0) addl r34 = @ltoff(double_sinh_arg_reduction), gp - nop.i 999 -} -;; - -{ .mmi - ld8 r34 = [r34] - nop.m 999 - nop.i 999 +{ .mfb + cmp.gt p7, p0 = -2, rExp_x // Test |x| < 2^(-2) + fma.s1 fXsq = fNormX, fNormX, f0 // x*x for small path +(p7) br.cond.spnt SINH_SMALL // Branch if 0 < |x| < 2^-2 } ;; - -// We want 2^(N-1) and 2^(-N-1). So bias N-1 and -N-1 and -// put them in an exponent. -// sinh_FR_spos = 2^(N-1) and sinh_FR_sneg = 2^(-N-1) -// r39 = 0xffff + (N-1) = 0xffff +N -1 -// r40 = 0xffff - (N +1) = 0xffff -N -1 - -{ .mlx - nop.m 999 -(p0) movl r38 = 0x000000000000fffe ;; -} - -{ .mmi -(p0) ldfe sinh_FR_Inv_log2by64 = [r34],16 ;; -(p0) ldfe sinh_FR_log2by64_hi = [r34],16 - nop.i 999 ;; -} - -{ .mbb -(p0) ldfe sinh_FR_log2by64_lo = [r34],16 - nop.b 999 - nop.b 999 ;; -} - -// Get the A coefficients -// f9 = A_1 -// f10 = A_2 -// f11 = A_3 - -{ .mmi - nop.m 999 -(p0) addl r34 = @ltoff(double_sinh_ab_table), gp - nop.i 999 +{ .mfi + nop.m 0 + // check for overflow + fcmp.ge.s1 p12, p13 = fAbsX, fMIN_SGL_OFLOW_ARG + mov rJ_mask = 0x3f // 6-bit mask for J } ;; -{ .mmi - ld8 r34 = [r34] - nop.m 999 - nop.i 999 +{ .mfb + nop.m 0 + fms.s1 fN = fNint, f1, fRightShifter // n in FP register + // branch out if overflow +(p12) br.cond.spnt SINH_CERTAIN_OVERFLOW } ;; - -// Calculate M and keep it as integer and floating point. -// f38 = M = round-to-integer(x*Inv_log2by64) -// sinh_FR_M = M = truncate(ax/(log2/64)) -// Put the significand of M in r35 -// and the floating point representation of M in sinh_FR_M - { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_M = sinh_FR_X, sinh_FR_Inv_log2by64, f0 - nop.i 999 + getf.sig rNJ = fNint // bits of n, j + // check for possible overflow + fcmp.gt.s1 p13, p0 = fAbsX, fMAX_SGL_NORM_ARG + nop.i 0 } +;; { .mfi -(p0) ldfe sinh_FR_A1 = [r34],16 - nop.f 999 - nop.i 999 ;; + addl rN = 0xFFBF - 63, rNJ // biased and shifted n-1,j + fnma.s1 fR = fLn2Div64, fN, fNormX // R = x - N*ln(2)/64 + and rJ = rJ_mask, rNJ // bits of j } - { .mfi - nop.m 999 -(p0) fcvt.fx.s1 sinh_FR_M_temp = sinh_FR_M - nop.i 999 ;; + sub rNJ_neg = r0, rNJ // bits of n, j for -x + nop.f 0 + andcm rN_mask = -1, rJ_mask // 0xff...fc0 to mask N } +;; { .mfi - nop.m 999 -(p0) fnorm.s1 sinh_FR_M = sinh_FR_M_temp - nop.i 999 ;; + shladd rJ = rJ, 3, rTblAddr // address in the 2^(j/64) table + nop.f 0 + and rN = rN_mask, rN // biased, shifted n-1 } - { .mfi -(p0) getf.sig r35 = sinh_FR_M_temp - nop.f 999 - nop.i 999 ;; -} - -// M is still in r35. Calculate j. j is the signed extension of the six lsb of M. It -// has a range of -32 thru 31. -// r35 = M -// r36 = j - -{ .mii - nop.m 999 - nop.i 999 ;; -(p0) and r36 = 0x3f, r35 ;; + addl rN_neg = 0xFFBF - 63, rNJ_neg // -x biased, shifted n-1,j + nop.f 0 + and rJ_neg = rJ_mask, rNJ_neg // bits of j for -x } - -// Calculate R -// f13 = f44 - f12*f10 = ax - M*log2by64_hi -// f14 = f13 - f8*f11 = R = (ax - M*log2by64_hi) - M*log2by64_lo +;; { .mfi - nop.m 999 -(p0) fnma.s1 sinh_FR_R_temp = sinh_FR_M, sinh_FR_log2by64_hi, sinh_FR_X - nop.i 999 + ld8 rJ = [rJ] // Table value + nop.f 0 + shl rN = rN, 46 // 2^(n-1) bits in DP format } - { .mfi -(p0) ldfe sinh_FR_A2 = [r34],16 - nop.f 999 - nop.i 999 ;; + shladd rJ_neg = rJ_neg, 3, rTblAddr // addr in 2^(j/64) table -x + nop.f 0 + and rN_neg = rN_mask, rN_neg // biased, shifted n-1 for -x } +;; { .mfi - nop.m 999 -(p0) fnma.s1 sinh_FR_R = sinh_FR_M, sinh_FR_log2by64_lo, sinh_FR_R_temp - nop.i 999 + ld8 rJ_neg = [rJ_neg] // Table value for -x + nop.f 0 + shl rN_neg = rN_neg, 46 // 2^(n-1) bits in DP format for -x } - -// Get the B coefficients -// f15 = B_1 -// f32 = B_2 -// f33 = B_3 - -{ .mmi -(p0) ldfe sinh_FR_A3 = [r34],16 ;; -(p0) ldfe sinh_FR_B1 = [r34],16 - nop.i 999 ;; -} - -{ .mmi -(p0) ldfe sinh_FR_B2 = [r34],16 ;; -(p0) ldfe sinh_FR_B3 = [r34],16 - nop.i 999 ;; -} - -{ .mii - nop.m 999 -(p0) shl r34 = r36, 0x2 ;; -(p0) sxt1 r37 = r34 ;; -} - -// ****************************************************** -// STEP 2 (TBL and EXP) -// ****************************************************** -// Calculate Rsquared and Rcubed in preparation for p_even and p_odd -// f12 = R*R*R -// f13 = R*R -// f14 = R <== from above +;; { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_Rsq = sinh_FR_R, sinh_FR_R, f0 -(p0) shr r36 = r37, 0x2 ;; -} - -// r34 = M-j = r35 - r36 -// r35 = N = (M-j)/64 - -{ .mii -(p0) sub r34 = r35, r36 - nop.i 999 ;; -(p0) shr r35 = r34, 0x6 ;; -} - -{ .mii -(p0) sub r40 = r38, r35 -(p0) adds r37 = 0x1, r35 -(p0) add r39 = r38, r35 ;; -} - -// Get the address of the J table, add the offset, -// addresses are sinh_AD_mJ and sinh_AD_J, get the T value -// f32 = T(j)_hi -// f33 = T(j)_lo -// f34 = T(-j)_hi -// f35 = T(-j)_lo - -{ .mmi -(p0) sub r34 = r35, r32 -(p0) addl r37 = @ltoff(double_sinh_j_table), gp - nop.i 999 + or rN = rN, rJ // bits of 2^n * 2^(j/64) in DP format + nop.f 0 + nop.i 0 } ;; -{ .mmi - ld8 r37 = [r37] - nop.m 999 - nop.i 999 +{ .mmf + setf.d fT = rN // 2^(n-1) * 2^(j/64) + or rN_neg = rN_neg, rJ_neg // -x bits of 2^n * 2^(j/64) in DP + fma.s1 fRSqr = fR, fR, f0 // R^2 } ;; - -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_Rcub = sinh_FR_Rsq, sinh_FR_R, f0 - nop.i 999 -} - -// ****************************************************** -// STEP 3 Now decide if we need to branch to EXP -// ****************************************************** -// Put 32 in f9; p6 true if x < 32 -// Go to EXP if |x| >= 32 - -{ .mlx - nop.m 999 -(p0) movl r32 = 0x0000000000010004 ;; -} - -// Calculate p_even -// f34 = B_2 + Rsq *B_3 -// f35 = B_1 + Rsq*f34 = B_1 + Rsq * (B_2 + Rsq *B_3) -// f36 = p_even = Rsq * f35 = Rsq * (B_1 + Rsq * (B_2 + Rsq *B_3)) - -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_peven_temp1 = sinh_FR_Rsq, sinh_FR_B3, sinh_FR_B2 - nop.i 999 ;; -} - -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_peven_temp2 = sinh_FR_Rsq, sinh_FR_peven_temp1, sinh_FR_B1 - nop.i 999 -} - -// Calculate p_odd -// f34 = A_2 + Rsq *A_3 -// f35 = A_1 + Rsq * (A_2 + Rsq *A_3) -// f37 = podd = R + Rcub * (A_1 + Rsq * (A_2 + Rsq *A_3)) - -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_podd_temp1 = sinh_FR_Rsq, sinh_FR_A3, sinh_FR_A2 - nop.i 999 ;; -} - -{ .mfi -(p0) setf.exp sinh_FR_N_temp1 = r39 - nop.f 999 - nop.i 999 ;; -} - { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_peven = sinh_FR_Rsq, sinh_FR_peven_temp2, f0 - nop.i 999 + setf.d fT_neg = rN_neg // 2^(n-1) * 2^(j/64) for -x + fma.s1 fP = fA3, fR, fA2 // A3*R + A2 + nop.i 0 } - { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_podd_temp2 = sinh_FR_Rsq, sinh_FR_podd_temp1, sinh_FR_A1 - nop.i 999 ;; + nop.m 0 + fnma.s1 fP_neg = fA3, fR, fA2 // A3*R + A2 for -x + nop.i 0 } +;; { .mfi -(p0) setf.exp f9 = r32 - nop.f 999 - nop.i 999 ;; + nop.m 0 + fma.s1 fP = fP, fRSqr, fR // P = (A3*R + A2)*R^2 + R + nop.i 0 } - { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_podd = sinh_FR_podd_temp2, sinh_FR_Rcub, sinh_FR_R - nop.i 999 -} - -// sinh_GR_mj contains the table offset for -j -// sinh_GR_j contains the table offset for +j -// p6 is true when j <= 0 - -{ .mlx -(p0) setf.exp sinh_FR_N_temp2 = r40 -(p0) movl r40 = 0x0000000000000020 ;; + nop.m 0 + fms.s1 fP_neg = fP_neg, fRSqr, fR // P = (A3*R + A2)*R^2 + R, -x + nop.i 0 } +;; { .mfi -(p0) sub sinh_GR_mJ = r40, r36 -(p0) fmerge.se sinh_FR_spos = sinh_FR_N_temp1, f1 -(p0) adds sinh_GR_J = 0x20, r36 ;; -} - -{ .mii - nop.m 999 -(p0) shl sinh_GR_mJ = sinh_GR_mJ, 5 ;; -(p0) add sinh_AD_mJ = r37, sinh_GR_mJ ;; -} - -{ .mmi - nop.m 999 -(p0) ldfe sinh_FR_Tmjhi = [sinh_AD_mJ],16 -(p0) shl sinh_GR_J = sinh_GR_J, 5 ;; + nop.m 0 + fmpy.s0 fTmp = fLn2Div64, fLn2Div64 // Force inexact + nop.i 0 } +;; { .mfi -(p0) ldfs sinh_FR_Tmjlo = [sinh_AD_mJ],16 -(p0) fcmp.lt.unc.s1 p0,p7 = sinh_FR_X,f9 -(p0) add sinh_AD_J = r37, sinh_GR_J ;; + nop.m 0 + fma.s1 fExp = fP, fT, fT // exp(x)/2 + nop.i 0 } - -{ .mmi -(p0) ldfe sinh_FR_Tjhi = [sinh_AD_J],16 ;; -(p0) ldfs sinh_FR_Tjlo = [sinh_AD_J],16 - nop.i 999 ;; -} - { .mfb - nop.m 999 -(p0) fmerge.se sinh_FR_sneg = sinh_FR_N_temp2, f1 -(p7) br.cond.spnt L(SINH_BY_EXP) ;; -} - -{ .mfi - nop.m 999 - nop.f 999 - nop.i 999 ;; + nop.m 0 + fma.s1 fExp_neg = fP_neg, fT_neg, fT_neg // exp(-x)/2 + // branch out if possible overflow result +(p13) br.cond.spnt SINH_POSSIBLE_OVERFLOW } +;; -// ****************************************************** -// If NOT branch to EXP -// ****************************************************** -// Calculate S_hi and S_lo -// sinh_FR_S_hi_temp = sinh_FR_sneg * sinh_FR_Tmjhi -// sinh_FR_S_hi = sinh_FR_spos * sinh_FR_Tjhi - sinh_FR_S_hi_temp -// sinh_FR_S_hi = sinh_FR_spos * sinh_FR_Tjhi - (sinh_FR_sneg * sinh_FR_Tmjlo) - -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_S_hi_temp = sinh_FR_sneg, sinh_FR_Tmjhi, f0 - nop.i 999 ;; +{ .mfb + nop.m 0 + // final result in the absence of overflow + fms.s.s0 f8 = fExp, f1, fExp_neg // result = (exp(x)-exp(-x))/2 + // exit here in the absence of overflow + br.ret.sptk b0 // Exit main path, 0.25 <= |x| < 89.41598 } +;; +// Here if 0 < |x| < 0.25. Evaluate 9th order polynomial. +SINH_SMALL: { .mfi - nop.m 999 -(p0) fms.s1 sinh_FR_S_hi = sinh_FR_spos, sinh_FR_Tjhi, sinh_FR_S_hi_temp - nop.i 999 + add rAd1 = 0x200, rTblAddr + fcmp.lt.s1 p7, p8 = fNormX, f0 // Test sign of x + cmp.gt p6, p0 = -60, rExp_x // Test |x| < 2^(-60) } - -// Calculate C_hi -// sinh_FR_C_hi_temp1 = sinh_FR_sneg * sinh_FR_Tmjhi -// sinh_FR_C_hi = sinh_FR_spos * sinh_FR_Tjhi + sinh_FR_C_hi_temp1 - { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_C_hi_temp1 = sinh_FR_sneg, sinh_FR_Tmjhi, f0 - nop.i 999 ;; + add rAd2 = 0x210, rTblAddr + nop.f 0 + nop.i 0 } +;; -// sinh_FR_S_lo_temp1 = sinh_FR_spos * sinh_FR_Tjhi - sinh_FR_S_hi -// sinh_FR_S_lo_temp2 = -sinh_FR_sneg * sinh_FR_Tmjlo + (sinh_FR_spos * sinh_FR_Tjhi - sinh_FR_S_hi) -// sinh_FR_S_lo_temp2 = -sinh_FR_sneg * sinh_FR_Tmjlo + (sinh_FR_S_lo_temp1 ) - -{ .mfi - nop.m 999 -(p0) fms.s1 sinh_FR_S_lo_temp1 = sinh_FR_spos, sinh_FR_Tjhi, sinh_FR_S_hi - nop.i 999 +{ .mmb + ldfpd fA4, fA3 = [rAd1] + ldfpd fA2, fA1 = [rAd2] +(p6) br.cond.spnt SINH_VERY_SMALL // Branch if |x| < 2^(-60) } +;; { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_C_hi = sinh_FR_spos, sinh_FR_Tjhi, sinh_FR_C_hi_temp1 - nop.i 999 ;; + nop.m 0 + fma.s1 fX3 = fXsq, fNormX, f0 + nop.i 0 } - { .mfi - nop.m 999 -(p0) fnma.s1 sinh_FR_S_lo_temp2 = sinh_FR_sneg, sinh_FR_Tmjhi, sinh_FR_S_lo_temp1 - nop.i 999 + nop.m 0 + fma.s1 fX4 = fXsq, fXsq, f0 + nop.i 0 } - -// sinh_FR_S_lo_temp1 = sinh_FR_sneg * sinh_FR_Tmjlo -// sinh_FR_S_lo_temp3 = sinh_FR_spos * sinh_FR_Tjlo - sinh_FR_S_lo_temp1 -// sinh_FR_S_lo_temp3 = sinh_FR_spos * sinh_FR_Tjlo -(sinh_FR_sneg * sinh_FR_Tmjlo) -// sinh_FR_S_lo = sinh_FR_S_lo_temp3 + sinh_FR_S_lo_temp2 +;; { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_S_lo_temp1 = sinh_FR_sneg, sinh_FR_Tmjlo, f0 - nop.i 999 ;; + nop.m 0 + fma.s1 fA43 = fXsq, fA4, fA3 + nop.i 0 } - -/////////// BUG FIX fma to fms -TK { .mfi - nop.m 999 -(p0) fms.s1 sinh_FR_S_lo_temp3 = sinh_FR_spos, sinh_FR_Tjlo, sinh_FR_S_lo_temp1 - nop.i 999 ;; + nop.m 0 + fma.s1 fA21 = fXsq, fA2, fA1 + nop.i 0 } +;; { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_S_lo = sinh_FR_S_lo_temp3, f1, sinh_FR_S_lo_temp2 - nop.i 999 ;; + nop.m 0 + fma.s1 fA4321 = fX4, fA43, fA21 + nop.i 0 } +;; -// Y_hi = S_hi -// Y_lo = C_hi*p_odd + (S_hi*p_even + S_lo) -// sinh_FR_Y_lo_temp = sinh_FR_S_hi * sinh_FR_peven + sinh_FR_S_lo -// sinh_FR_Y_lo = sinh_FR_C_hi * sinh_FR_podd + sinh_FR_Y_lo_temp - +// Dummy multiply to generate inexact { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_Y_lo_temp = sinh_FR_S_hi, sinh_FR_peven, sinh_FR_S_lo - nop.i 999 ;; + nop.m 0 + fmpy.s0 fTmp = fA4, fA4 + nop.i 0 } - -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_Y_lo = sinh_FR_C_hi, sinh_FR_podd, sinh_FR_Y_lo_temp - nop.i 999 ;; +{ .mfb + nop.m 0 + fma.s.s0 f8 = fA4321, fX3, fNormX + br.ret.sptk b0 // Exit if 2^-60 < |x| < 0.25 } +;; -// sinh_FR_SINH = Y_hi + Y_lo -// f8 = answer = sinh_FR_SGNX * sinh_FR_SINH - -// Dummy multiply to generate inexact +SINH_VERY_SMALL: +// Here if 0 < |x| < 2^-60 +// Compute result by x + sgn(x)*x^2 to get properly rounded result +.pred.rel "mutex",p7,p8 { .mfi - nop.m 999 -(p0) fmpy.s0 sinh_FR_tmp = sinh_FR_all_ones, sinh_FR_all_ones - nop.i 999 + nop.m 0 +(p7) fnma.s.s0 f8 = fNormX, fNormX, fNormX // If x<0 result ~ x-x^2 + nop.i 0 } -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_SINH = sinh_FR_S_hi, f1, sinh_FR_Y_lo - nop.i 999 ;; -} - { .mfb - nop.m 999 -(p0) fma.s.s0 f8 = sinh_FR_SGNX, sinh_FR_SINH,f0 -(p0) br.ret.sptk b0 ;; + nop.m 0 +(p8) fma.s.s0 f8 = fNormX, fNormX, fNormX // If x>0 result ~ x+x^2 + br.ret.sptk b0 // Exit if |x| < 2^-60 } +;; +SINH_POSSIBLE_OVERFLOW: -L(SINH_BY_EXP): +// Here if fMAX_SGL_NORM_ARG < x < fMIN_SGL_OFLOW_ARG +// This cannot happen if input is a single, only if input higher precision. +// Overflow is a possibility, not a certainty. -// When p7 is true, we know that an overflow is not going to happen -// When p7 is false, we must check for possible overflow -// p7 is the over_SAFE flag -// Y_hi = Tjhi -// Y_lo = Tjhi * (p_odd + p_even) +Tjlo -// Scale = sign * 2^(N-1) -// sinh_FR_Y_lo = sinh_FR_Tjhi * (sinh_FR_peven + sinh_FR_podd) -// sinh_FR_Y_lo = sinh_FR_Tjhi * (sinh_FR_Y_lo_temp ) +// Recompute result using status field 2 with user's rounding mode, +// and wre set. If result is larger than largest single, then we have +// overflow { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_Y_lo_temp = sinh_FR_peven, f1, sinh_FR_podd - nop.i 999 -} - -// Now we are in EXP. This is the only path where an overflow is possible -// but not for certain. So this is the only path where over_SAFE has any use. -// r34 still has N-1 -// There is a danger of double-extended overflow if N-1 > 16382 = 0x3ffe -// There is a danger of double overflow if N-1 > 0x3fe = 1022 -// There is a danger of single overflow if N-1 > 0x7e = 126 -{ .mlx - nop.m 999 -(p0) movl r32 = 0x000000000000007e ;; -} - -{ .mfi -(p0) cmp.gt.unc p0,p7 = r34, r32 -(p0) fmerge.s sinh_FR_SCALE = sinh_FR_SGNX, sinh_FR_spos - nop.i 999 ;; -} - -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_Y_lo = sinh_FR_Tjhi, sinh_FR_Y_lo_temp, sinh_FR_Tjlo - nop.i 999 ;; + mov rGt_ln = 0x1007f // Exponent for largest single + 1 ulp + fsetc.s2 0x7F,0x42 // Get user's round mode, set wre + nop.i 0 } +;; -// f8 = answer = scale * (Y_hi + Y_lo) { .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_SINH_temp = sinh_FR_Y_lo, f1, sinh_FR_Tjhi - nop.i 999 ;; + setf.exp fGt_pln = rGt_ln // Create largest single + 1 ulp + fma.s.s2 fWre_urm_f8 = fP, fT, fT // Result with wre set + nop.i 0 } +;; { .mfi - nop.m 999 -(p0) fma.s.s0 f44 = sinh_FR_SCALE, sinh_FR_SINH_temp, f0 - nop.i 999 ;; + nop.m 0 + fsetc.s2 0x7F,0x40 // Turn off wre in sf2 + nop.i 0 } +;; -// Dummy multiply to generate inexact { .mfi - nop.m 999 -(p7) fmpy.s0 sinh_FR_tmp = sinh_FR_all_ones, sinh_FR_all_ones - nop.i 999 ;; + nop.m 0 + fcmp.ge.s1 p6, p0 = fWre_urm_f8, fGt_pln // Test for overflow + nop.i 0 } +;; -// If over_SAFE is set, return { .mfb - nop.m 999 -(p7) fmerge.s f8 = f44,f44 -(p7) br.ret.sptk b0 ;; + nop.m 0 + nop.f 0 +(p6) br.cond.spnt SINH_CERTAIN_OVERFLOW // Branch if overflow } +;; -// Else see if we overflowed -// S0 user supplied status -// S2 user supplied status + WRE + TD (Overflows) -// If WRE is set then an overflow will not occur in EXP. -// The input value that would cause a register (WRE) value to overflow is about 2^15 -// and this input would go into the HUGE path. -// Answer with WRE is in f43. - -{ .mfi - nop.m 999 -(p0) fsetc.s2 0x7F,0x42 - nop.i 999;; +{ .mfb + nop.m 0 + fma.s.s0 f8 = fP, fT, fT + br.ret.sptk b0 // Exit if really no overflow } +;; +// here if overflow +SINH_CERTAIN_OVERFLOW: { .mfi - nop.m 999 -(p0) fma.s.s2 f43 = sinh_FR_SCALE, sinh_FR_SINH_temp, f0 - nop.i 999 ;; -} - -// 1007F => 1007F -FFFF = 80(true) -// 80 + 7F = FF, which is 1 more that the exponent of the largest -// double (FE). So 0 1007F 8000000000000000 is one ulp more than -// largest single in register bias -// Now set p8 if the answer with WRE is greater than or equal this value -// Also set p9 if the answer with WRE is less than or equal to negative this value - -{ .mlx - nop.m 999 -(p0) movl r32 = 0x0000000001007F ;; + addl r17ones_m1 = 0x1FFFE, r0 + fcmp.lt.s1 p6, p7 = fNormX, f0 // Test for x < 0 + nop.i 0 } +;; { .mmf - nop.m 999 -(p0) setf.exp f41 = r32 -(p0) fsetc.s2 0x7F,0x40 ;; -} - -{ .mfi - nop.m 999 -(p0) fcmp.ge.unc.s1 p8, p0 = f43, f41 - nop.i 999 + alloc r32 = ar.pfs, 0, 3, 4, 0 // get some registers + setf.exp fTmp = r17ones_m1 + fmerge.s FR_X = f8,f8 } +;; { .mfi - nop.m 999 -(p0) fmerge.ns f42 = f41, f41 - nop.i 999 ;; + mov GR_Parameter_TAG = 128 +(p6) fnma.s.s0 FR_RESULT = fTmp, fTmp, f0 // Set I,O and -INF result + nop.i 0 } - -// The error tag for overflow is 128 -{ .mii - nop.m 999 - nop.i 999 ;; -(p8) mov r47 = 128 ;; -} - { .mfb - nop.m 999 -(p0) fcmp.le.unc.s1 p9, p0 = f43, f42 -(p8) br.cond.spnt L(SINH_ERROR_SUPPORT) ;; -} - -{ .mii - nop.m 999 - nop.i 999 ;; -(p9) mov r47 = 128 -} - -{ .mib - nop.m 999 - nop.i 999 -(p9) br.cond.spnt L(SINH_ERROR_SUPPORT) ;; -} - -// Dummy multiply to generate inexact -{ .mfi - nop.m 999 -(p0) fmpy.s0 sinh_FR_tmp = sinh_FR_all_ones, sinh_FR_all_ones - nop.i 999 ;; + nop.m 0 +(p7) fma.s.s0 FR_RESULT = fTmp, fTmp, f0 // Set I,O and +INF result + br.cond.sptk __libm_error_region } +;; +// Here if x unorm +SINH_UNORM: { .mfb - nop.m 999 -(p0) fmerge.s f8 = f44,f44 -(p0) br.ret.sptk b0 ;; -} - -L(SINH_HUGE): - -// for SINH_HUGE, put 24000 in exponent; take sign from input; add 1 -// SAFE: SAFE is always 0 for HUGE - -{ .mlx - nop.m 999 -(p0) movl r32 = 0x0000000000015dbf ;; + getf.exp rSignexp_x = fNormX // Must recompute if x unorm + fcmp.eq.s0 p6, p0 = f8, f0 // Set D flag + br.cond.sptk SINH_COMMON // Return to main path } +;; -{ .mfi -(p0) setf.exp f9 = r32 - nop.f 999 - nop.i 999 ;; -} +GLOBAL_IEEE754_END(sinhf) -{ .mfi - nop.m 999 -(p0) fma.s1 sinh_FR_signed_hi_lo = sinh_FR_SGNX, f9, f1 - nop.i 999 ;; -} -{ .mfi - nop.m 999 -(p0) fma.s.s0 f44 = sinh_FR_signed_hi_lo, f9, f0 -(p0) mov r47 = 128 -} -.endp sinhf -ASM_SIZE_DIRECTIVE(sinhf) -#ifdef _LIBC -ASM_SIZE_DIRECTIVE(__ieee754_sinhf) -#endif - -// Stack operations when calling error support. -// (1) (2) (3) (call) (4) -// sp -> + psp -> + psp -> + sp -> + -// | | | | -// | | <- GR_Y R3 ->| <- GR_RESULT | -> f8 -// | | | | -// | <-GR_Y Y2->| Y2 ->| <- GR_Y | -// | | | | -// | | <- GR_X X1 ->| | -// | | | | -// sp-64 -> + sp -> + sp -> + + -// save ar.pfs save b0 restore gp -// save gp restore ar.pfs - -.proc __libm_error_region -__libm_error_region: -L(SINH_ERROR_SUPPORT): +LOCAL_LIBM_ENTRY(__libm_error_region) .prologue - -// (1) { .mfi - add GR_Parameter_Y=-32,sp // Parameter 2 value - nop.f 0 + add GR_Parameter_Y=-32,sp // Parameter 2 value + nop.f 0 .save ar.pfs,GR_SAVE_PFS - mov GR_SAVE_PFS=ar.pfs // Save ar.pfs + mov GR_SAVE_PFS=ar.pfs // Save ar.pfs } { .mfi .fframe 64 - add sp=-64,sp // Create new stack - nop.f 0 - mov GR_SAVE_GP=gp // Save gp + add sp=-64,sp // Create new stack + nop.f 0 + mov GR_SAVE_GP=gp // Save gp };; - - -// (2) { .mmi - stfs [GR_Parameter_Y] = f0,16 // STORE Parameter 2 on stack - add GR_Parameter_X = 16,sp // Parameter 1 address + stfs [GR_Parameter_Y] = FR_Y,16 // Store Parameter 2 on stack + add GR_Parameter_X = 16,sp // Parameter 1 address .save b0, GR_SAVE_B0 - mov GR_SAVE_B0=b0 // Save b0 + mov GR_SAVE_B0=b0 // Save b0 };; - .body -// (3) -{ .mib - stfs [GR_Parameter_X] = f8 // STORE Parameter 1 on stack - add GR_Parameter_RESULT = 0,GR_Parameter_Y // Parameter 3 address - nop.b 0 +{ .mfi + stfs [GR_Parameter_X] = FR_X // Store Parameter 1 on stack + nop.f 0 + add GR_Parameter_RESULT = 0,GR_Parameter_Y // Parameter 3 address } { .mib - stfs [GR_Parameter_Y] = f44 // STORE Parameter 3 on stack - add GR_Parameter_Y = -16,GR_Parameter_Y - br.call.sptk b0=__libm_error_support# // Call error handling function + stfs [GR_Parameter_Y] = FR_RESULT // Store Parameter 3 on stack + add GR_Parameter_Y = -16,GR_Parameter_Y + br.call.sptk b0=__libm_error_support# // Call error handling function };; + { .mmi - nop.m 0 - nop.m 0 - add GR_Parameter_RESULT = 48,sp + add GR_Parameter_RESULT = 48,sp + nop.m 0 + nop.i 0 };; -// (4) { .mmi - ldfs f8 = [GR_Parameter_RESULT] // Get return result off stack + ldfs f8 = [GR_Parameter_RESULT] // Get return result off stack .restore sp - add sp = 64,sp // Restore stack pointer - mov b0 = GR_SAVE_B0 // Restore return address + add sp = 64,sp // Restore stack pointer + mov b0 = GR_SAVE_B0 // Restore return address };; { .mib - mov gp = GR_SAVE_GP // Restore gp - mov ar.pfs = GR_SAVE_PFS // Restore ar.pfs - br.ret.sptk b0 // Return + mov gp = GR_SAVE_GP // Restore gp + mov ar.pfs = GR_SAVE_PFS // Restore ar.pfs + br.ret.sptk b0 // Return };; -.endp __libm_error_region -ASM_SIZE_DIRECTIVE(__libm_error_region) +LOCAL_LIBM_END(__libm_error_region) + .type __libm_error_support#,@function .global __libm_error_support# |