mirror of
https://sourceware.org/git/glibc.git
synced 2024-11-26 15:00:06 +00:00
aa1142c593
Continuing the preparation for additional _FloatN / _FloatNx function aliases, this patch makes ia64 libm function implementations use libm_alias_float to define function aliases. The same approach is followed as with the corresponding long double and double patches: the ia64-specific macros are left unchanged, with calls to libm_alias_float_other being added in most cases and libm_alias_float itself being used in only a few places. Tested with build-many-glibcs.py for ia64-linux-gnu that installed stripped shared libraries are unchanged by the patch. * sysdeps/ia64/fpu/libm-symbols.h: Include <libm-alias-float.h>. * sysdeps/ia64/fpu/e_acosf.S (acosf): Use libm_alias_float_other. * sysdeps/ia64/fpu/e_acoshf.S (acoshf): Likewise. * sysdeps/ia64/fpu/e_asinf.S (asinf): Likewise. * sysdeps/ia64/fpu/e_atan2f.S (atan2f): Likewise. * sysdeps/ia64/fpu/e_atanhf.S (atanhf): Likewise. * sysdeps/ia64/fpu/e_coshf.S (coshf): Likewise. * sysdeps/ia64/fpu/e_exp10f.S (exp10f): Likewise. * sysdeps/ia64/fpu/e_exp2f.S (exp2f): Likewise. * sysdeps/ia64/fpu/e_expf.S (expf): Likewise. * sysdeps/ia64/fpu/e_fmodf.S (fmodf): Likewise. * sysdeps/ia64/fpu/e_hypotf.S (hypotf): Likewise. * sysdeps/ia64/fpu/e_lgammaf_r.c (lgammaf_r): Define using libm_alias_float_r. * sysdeps/ia64/fpu/e_log2f.S (log2f): Use libm_alias_float_other. * sysdeps/ia64/fpu/e_logf.S (log10f): Likewise. (logf): Likewise. * sysdeps/ia64/fpu/e_powf.S (powf): Likewise. * sysdeps/ia64/fpu/e_remainderf.S (remainderf): Likewise. * sysdeps/ia64/fpu/e_sinhf.S (sinhf): Likewise. * sysdeps/ia64/fpu/e_sqrtf.S (sqrtf): Likewise. * sysdeps/ia64/fpu/libm_sincosf.S (sincosf): Likewise. * sysdeps/ia64/fpu/s_asinhf.S (asinhf): Likewise. * sysdeps/ia64/fpu/s_atanf.S (atanf): Likewise. * sysdeps/ia64/fpu/s_cbrtf.S (cbrtf): Likewise. * sysdeps/ia64/fpu/s_ceilf.S (ceilf): Likewise. * sysdeps/ia64/fpu/s_copysign.S (copysignf): Define using libm_alias_float. * sysdeps/ia64/fpu/s_cosf.S (sinf): Use libm_alias_float_other. (cosf): Likewise. * sysdeps/ia64/fpu/s_erfcf.S (erfcf): Likewise. * sysdeps/ia64/fpu/s_erff.S (erff): Likewise. * sysdeps/ia64/fpu/s_expm1f.S (expm1f): Likewise. * sysdeps/ia64/fpu/s_fabsf.S (fabsf): Likewise. * sysdeps/ia64/fpu/s_fdimf.S (fdimf): Likewise. * sysdeps/ia64/fpu/s_floorf.S (floorf): Likewise. * sysdeps/ia64/fpu/s_fmaf.S (fmaf): Likewise. * sysdeps/ia64/fpu/s_fmaxf.S (fmaxf): Likewise. * sysdeps/ia64/fpu/s_frexpf.c (frexpf): Likewise. * sysdeps/ia64/fpu/s_ldexpf.c (ldexpf): Likewise. * sysdeps/ia64/fpu/s_log1pf.S (log1pf): Likewise. * sysdeps/ia64/fpu/s_logbf.S (logbf): Likewise. * sysdeps/ia64/fpu/s_modff.S (modff): Likewise. * sysdeps/ia64/fpu/s_nearbyintf.S (nearbyintf): Likewise. * sysdeps/ia64/fpu/s_nextafterf.S (nextafterf): Likewise. * sysdeps/ia64/fpu/s_rintf.S (rintf): Likewise. * sysdeps/ia64/fpu/s_roundf.S (roundf): Likewise. * sysdeps/ia64/fpu/s_scalblnf.c (scalblnf): Likewise. * sysdeps/ia64/fpu/s_scalbnf.c (scalbnf): Define using libm_alias_float. * sysdeps/ia64/fpu/s_tanf.S (tanf): Use libm_alias_float_other. * sysdeps/ia64/fpu/s_tanhf.S (tanhf): Likewise. * sysdeps/ia64/fpu/s_truncf.S (truncf): Likewise. * sysdeps/ia64/fpu/w_lgammaf_main.c [BUILD_LGAMMA && !USE_AS_COMPAT] (lgammaf): Likewise. * sysdeps/ia64/fpu/w_tgammaf_compat.S (tgammaf): Likewise.
902 lines
24 KiB
ArmAsm
902 lines
24 KiB
ArmAsm
.file "atan2f.s"
|
|
|
|
|
|
// Copyright (c) 2000 - 2003, Intel Corporation
|
|
// All rights reserved.
|
|
//
|
|
// 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
|
|
// met:
|
|
//
|
|
// * Redistributions of source code must retain the above copyright
|
|
// notice, this list of conditions and the following disclaimer.
|
|
//
|
|
// * Redistributions in binary form must reproduce the above copyright
|
|
// notice, this list of conditions and the following disclaimer in the
|
|
// documentation and/or other materials provided with the distribution.
|
|
//
|
|
// * 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
|
|
// LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
|
|
// 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
|
|
// 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://www.intel.com/software/products/opensource/libraries/num.htm.
|
|
|
|
// History
|
|
//==============================================================
|
|
// 06/01/00 Initial version
|
|
// 08/15/00 Bundle added after call to __libm_error_support to properly
|
|
// set [the previously overwritten] GR_Parameter_RESULT.
|
|
// 08/17/00 Changed predicate register macro-usage to direct predicate
|
|
// names due to an assembler bug.
|
|
// 01/05/01 Fixed flag settings for denormal input.
|
|
// 01/19/01 Added documentation
|
|
// 01/30/01 Improved speed
|
|
// 02/06/02 Corrected .section statement
|
|
// 05/20/02 Cleaned up namespace and sf0 syntax
|
|
// 02/06/03 Reordered header: .section, .global, .proc, .align
|
|
|
|
// Description
|
|
//=========================================
|
|
// The atan2 function computes the principle value of the arc tangent of y/x using
|
|
// the signs of both arguments to determine the quadrant of the return value.
|
|
// A domain error may occur if both arguments are zero.
|
|
|
|
// The atan2 function returns the arc tangent of y/x in the range [-pi,+pi] radians.
|
|
|
|
//..
|
|
//..Let (v,u) = (y,x) if |y| <= |x|, and (v,u) = (x,y) otherwise. Note that
|
|
//..v and u can be negative. We state the relationship between atan2(y,x) and
|
|
//..atan(v/u).
|
|
//..
|
|
//..Let swap = false if v = y, and swap = true if v = x.
|
|
//..Define C according to the matrix
|
|
//..
|
|
//.. TABLE FOR C
|
|
//.. x +ve x -ve
|
|
//.. no swap (swap = false) sgn(y)*0 sgn(y)*pi
|
|
//.. swap (swap = true ) sgn(y)*pi/2 sgn(y)*pi/2
|
|
//..
|
|
//.. atan2(y,x) = C + atan(v/u) if no swap
|
|
//.. atan2(y,x) = C - atan(v/u) if swap
|
|
//..
|
|
//..These relationship is more efficient to compute as we accommodate signs in v and u
|
|
//..saving the need to obtain the absolute value before computation can proceed.
|
|
//..
|
|
//..Suppose (v,u) = (y,x), we calculate atan(v/u) as follows:
|
|
//..A = y * frcpa(x) (so A = (y/x)(1 - beta))
|
|
//..atan(y/x) = atan(A) + atan( ((y/x)-A))/(1 + (y/x)A) ), the second term is
|
|
//..a correction.
|
|
//..atan(A) is approximated by a polynomial
|
|
//..A + p1 A^3 + p2 A^5 + ... + p10 A^21,
|
|
//..atan(G) is approximated as follows:
|
|
//..Let G = (y - Ax)/(x + Ay), atan(G) can be approximated by G + g * p1
|
|
//..where g is a limited precision approximation to G via g = (y - Ax)*frcpa(x + Ay).
|
|
//..
|
|
//..Suppose (v,u) = (x,y), we calculate atan(v/u) as follows:
|
|
//..Z = x * frcpa(y) (so Z = (x/y)(1 - beta))
|
|
//..atan(x/y) = atan(Z) + atan( ((x/y)-Z))/(1 + (x/y)Z) ), the second term is
|
|
//..a correction.
|
|
//..atan(Z) is approximated by a polynomial
|
|
//..Z + p1 Z^3 + p2 Z^5 + ... + p10 Z^21,
|
|
//..atan(T) is approximated as follows:
|
|
//..Let T = (x - Ay)/(y + Ax), atan(T) can be approximated by T + t * p1
|
|
//..where t is a limited precision approximation to T via t = (x - Ay)*frcpa(y + Ax).
|
|
//..
|
|
//..
|
|
//..A = y * frcpa(x)
|
|
//..atan(A) ~=~ A + p1 A^3 + ... + P10 A^21
|
|
//..
|
|
//..This polynomial is computed as follows:
|
|
//..Asq = A*A; Acub = A*Asq, A4 = Asq*Asq
|
|
//..A5 = Asq*Acub, A6 = Asq*A4; A11 = A5 * A6
|
|
//..
|
|
//..poly_A1 = p9 + Asq*p10, poly_A2 = p7 + Asq*p8, poly_A3 = p5 + Asq*p6
|
|
//..poly_A1 = poly_A2 + A4 * poly_A1
|
|
//..poly_A1 = poly_A3 + A4 * poly_A1
|
|
//..
|
|
//..poly_A4 = p1 * A
|
|
//,,poly_A5 = p3 + Asq * p4, poly_A4 = A + Asq*poly_A4
|
|
//..poly_A5 = p2 + Asq * poly_A5
|
|
//..poly_A4 = poly_A4 + A5 * poly_A5
|
|
//..
|
|
//..atan_A = poly_A4 + A11 * poly_A1
|
|
//..
|
|
//..atan(G) is approximated as follows:
|
|
//..G_numer = y - A*x, G_denom = x + A*y
|
|
//..H1 = frcpa(G_denom)
|
|
//..H_beta = 1 - H1 * G_denom
|
|
//..H2 = H1 + H1 * H_beta
|
|
//..H_beta2 = H_beta*H_beta
|
|
//..H3 = H2 + H2*H_beta2
|
|
//..g = H1 * G_numer; gsq = g*g; atan_G = g*p1, atan_G = atan_G*gsq
|
|
//..atan_G = G_numer*H3 + atan_G
|
|
//..
|
|
//..
|
|
//..A = y * frcpa(x)
|
|
//..atan(A) ~=~ A + p1 A^3 + ... + P10 A^21
|
|
//..
|
|
//..This polynomial is computed as follows:
|
|
//..Asq = A*A; Acub = A*Asq, A4 = Asq*Asq
|
|
//..A5 = Asq*Acub, A6 = Asq*A4; A11 = A5 * A6
|
|
//..
|
|
//..poly_A1 = p9 + Asq*p10, poly_A2 = p7 + Asq*p8, poly_A3 = p5 + Asq*p6
|
|
//..poly_A1 = poly_A2 + A4 * poly_A1
|
|
//..poly_A1 = poly_A3 + A4 * poly_A1
|
|
//..
|
|
//..poly_A4 = p1 * A
|
|
//,,poly_A5 = p3 + Asq * p4, poly_A4 = A + Asq*poly_A4
|
|
//..poly_A5 = p2 + Asq * poly_A5
|
|
//..poly_A4 = poly_A4 + A5 * poly_A5
|
|
//..
|
|
//..atan_A = poly_A4 + A11 * poly_A1
|
|
//..
|
|
//..
|
|
//..====================================================================
|
|
//.. COEFFICIENTS USED IN THE COMPUTATION
|
|
//..====================================================================
|
|
|
|
//coef_pj, j = 1,2,...,10; atan(A) ~=~ A + p1 A^3 + p2 A^5 + ... + p10 A^21
|
|
//
|
|
// coef_p1 = -.3333332707155439167401311806315789E+00
|
|
// coef_p1 in dbl = BFD5 5555 1219 1621
|
|
//
|
|
// coef_p2 = .1999967670926658391827857030875748E+00
|
|
// coef_p2 in dbl = 3FC9 997E 7AFB FF4E
|
|
//
|
|
// coef_p3 = -.1427989384500152360161563301087296E+00
|
|
// coef_p3 in dbl = BFC2 473C 5145 EE38
|
|
//
|
|
// coef_p4 = .1105852823460720770079031213661163E+00
|
|
// coef_p4 in dbl = 3FBC 4F51 2B18 65F5
|
|
//
|
|
// coef_p5 = -.8811839915595312348625710228448363E-01
|
|
// coef_p5 in dbl = BFB6 8EED 6A8C FA32
|
|
//
|
|
// coef_p6 = .6742329836955067042153645159059714E-01
|
|
// coef_p6 in dbl = 3FB1 42A7 3D7C 54E3
|
|
//
|
|
// coef_p7 = -.4468571068774672908561591262231909E-01
|
|
// coef_p7 in dbl = BFA6 E10B A401 393F
|
|
//
|
|
// coef_p8 = .2252333246746511135532726960586493E-01
|
|
// coef_p8 in dbl = 3F97 105B 4160 F86B
|
|
//
|
|
// coef_p9 = -.7303884867007574742501716845542314E-02
|
|
// coef_p9 in dbl = BF7D EAAD AA33 6451
|
|
//
|
|
// coef_p10 = .1109686868355312093949039454619058E-02
|
|
// coef_p10 in dbl = 3F52 2E5D 33BC 9BAA
|
|
//
|
|
|
|
// Special values
|
|
//==============================================================
|
|
// Y x Result
|
|
// +number +inf +0
|
|
// -number +inf -0
|
|
// +number -inf +pi
|
|
// -number -inf -pi
|
|
//
|
|
// +inf +number +pi/2
|
|
// -inf +number -pi/2
|
|
// +inf -number +pi/2
|
|
// -inf -number -pi/2
|
|
//
|
|
// +inf +inf +pi/4
|
|
// -inf +inf -pi/4
|
|
// +inf -inf +3pi/4
|
|
// -inf -inf -3pi/4
|
|
//
|
|
// +1 +1 +pi/4
|
|
// -1 +1 -pi/4
|
|
// +1 -1 +3pi/4
|
|
// -1 -1 -3pi/4
|
|
//
|
|
// +number +0 +pi/2 // does not raise DBZ
|
|
// -number +0 -pi/2 // does not raise DBZ
|
|
// +number -0 +pi/2 // does not raise DBZ
|
|
// -number -0 -pi/2 // does not raise DBZ
|
|
//
|
|
// +0 +number +0
|
|
// -0 +number -0
|
|
// +0 -number +pi
|
|
// -0 -number -pi
|
|
//
|
|
// +0 +0 +0 // does not raise invalid
|
|
// -0 +0 -0 // does not raise invalid
|
|
// +0 -0 +pi // does not raise invalid
|
|
// -0 -0 -pi // does not raise invalid
|
|
//
|
|
// Nan anything quiet Y
|
|
// anything NaN quiet X
|
|
|
|
// atan2(+-0/+-0) sets double error tag to 37
|
|
// atan2f(+-0/+-0) sets single error tag to 38
|
|
// These are domain errors.
|
|
|
|
|
|
//
|
|
// Assembly macros
|
|
//=========================================
|
|
|
|
|
|
// integer registers
|
|
atan2f_GR_Addr_1 = r33
|
|
atan2f_GR_Addr_2 = r34
|
|
GR_SAVE_B0 = r35
|
|
|
|
GR_SAVE_PFS = r36
|
|
GR_SAVE_GP = r37
|
|
|
|
GR_Parameter_X = r38
|
|
GR_Parameter_Y = r39
|
|
GR_Parameter_RESULT = r40
|
|
GR_Parameter_TAG = r41
|
|
|
|
// floating point registers
|
|
atan2f_coef_p1 = f32
|
|
atan2f_coef_p10 = f33
|
|
atan2f_coef_p7 = f34
|
|
atan2f_coef_p6 = f35
|
|
|
|
atan2f_coef_p3 = f36
|
|
atan2f_coef_p2 = f37
|
|
atan2f_coef_p9 = f38
|
|
atan2f_coef_p8 = f39
|
|
atan2f_coef_p5 = f40
|
|
|
|
atan2f_coef_p4 = f41
|
|
atan2f_const_piby2 = f42
|
|
atan2f_const_pi = f43
|
|
atan2f_const_piby4 = f44
|
|
atan2f_const_3piby4 = f45
|
|
|
|
atan2f_xsq = f46
|
|
atan2f_ysq = f47
|
|
atan2f_xy = f48
|
|
atan2f_const_1 = f49
|
|
atan2f_sgn_Y = f50
|
|
|
|
atan2f_Z0 = f51
|
|
atan2f_A0 = f52
|
|
atan2f_Z = f53
|
|
atan2f_A = f54
|
|
atan2f_C = f55
|
|
|
|
atan2f_U = f56
|
|
atan2f_Usq = f57
|
|
atan2f_U4 = f58
|
|
atan2f_U6 = f59
|
|
atan2f_U8 = f60
|
|
|
|
atan2f_poly_u109 = f61
|
|
atan2f_poly_u87 = f62
|
|
atan2f_poly_u65 = f63
|
|
atan2f_poly_u43 = f64
|
|
atan2f_poly_u21 = f65
|
|
|
|
atan2f_poly_u10to7 = f66
|
|
atan2f_poly_u6to3 = f67
|
|
atan2f_poly_u10to3 = f68
|
|
atan2f_poly_u10to0 = f69
|
|
atan2f_poly_u210 = f70
|
|
|
|
atan2f_T_numer = f71
|
|
atan2f_T_denom = f72
|
|
atan2f_G_numer = f73
|
|
atan2f_G_denom = f74
|
|
atan2f_p1rnum = f75
|
|
|
|
atan2f_R_denom = f76
|
|
atan2f_R_numer = f77
|
|
atan2f_pR = f78
|
|
atan2f_pRC = f79
|
|
atan2f_pQRC = f80
|
|
|
|
atan2f_Q1 = f81
|
|
atan2f_Q_beta = f82
|
|
atan2f_Q2 = f83
|
|
atan2f_Q_beta2 = f84
|
|
atan2f_Q3 = f85
|
|
|
|
atan2f_r = f86
|
|
atan2f_rsq = f87
|
|
atan2f_poly_atan_U = f88
|
|
|
|
|
|
// predicate registers
|
|
//atan2f_Pred_Swap = p6 // |y| > |x|
|
|
//atan2f_Pred_noSwap = p7 // |y| <= |x|
|
|
//atan2f_Pred_Xpos = p8 // x >= 0
|
|
//atan2f_Pred_Xneg = p9 // x < 0
|
|
|
|
|
|
RODATA
|
|
|
|
.align 16
|
|
|
|
LOCAL_OBJECT_START(atan2f_coef_table1)
|
|
data8 0xBFD5555512191621 // p1
|
|
data8 0x3F522E5D33BC9BAA // p10
|
|
data8 0xBFA6E10BA401393F // p7
|
|
data8 0x3FB142A73D7C54E3 // p6
|
|
data8 0xBFC2473C5145EE38 // p3
|
|
data8 0x3FC9997E7AFBFF4E // p2
|
|
LOCAL_OBJECT_END(atan2f_coef_table1)
|
|
|
|
LOCAL_OBJECT_START(atan2f_coef_table2)
|
|
data8 0xBF7DEAADAA336451 // p9
|
|
data8 0x3F97105B4160F86B // p8
|
|
data8 0xBFB68EED6A8CFA32 // p5
|
|
data8 0x3FBC4F512B1865F5 // p4
|
|
data8 0x3ff921fb54442d18 // pi/2
|
|
data8 0x400921fb54442d18 // pi
|
|
data8 0x3fe921fb54442d18 // pi/4
|
|
data8 0x4002d97c7f3321d2 // 3pi/4
|
|
LOCAL_OBJECT_END(atan2f_coef_table2)
|
|
|
|
|
|
|
|
.section .text
|
|
GLOBAL_IEEE754_ENTRY(atan2f)
|
|
|
|
{ .mfi
|
|
alloc r32 = ar.pfs,1,5,4,0
|
|
frcpa.s1 atan2f_Z0,p0 = f1,f8 // Approx to 1/y
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
addl atan2f_GR_Addr_1 = @ltoff(atan2f_coef_table1),gp
|
|
fma.s1 atan2f_xsq = f9,f9,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
ld8 atan2f_GR_Addr_1 = [atan2f_GR_Addr_1]
|
|
frcpa.s1 atan2f_A0,p0 = f1,f9 // Approx to 1/x
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_ysq = f8,f8,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fcmp.ge.s1 p8,p9 = f9,f0 // Set p8 if x>=0, p9 if x<0
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_xy = f9,f8,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
add atan2f_GR_Addr_2 = 0x30, atan2f_GR_Addr_1
|
|
fmerge.s atan2f_sgn_Y = f8,f1
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mmf
|
|
ldfpd atan2f_coef_p1,atan2f_coef_p10 = [atan2f_GR_Addr_1],16
|
|
ldfpd atan2f_coef_p9,atan2f_coef_p8 = [atan2f_GR_Addr_2],16
|
|
fclass.m p10,p0 = f9,0xe7 // Test x @inf|@snan|@qnan|@zero
|
|
}
|
|
;;
|
|
|
|
{ .mfi
|
|
ldfpd atan2f_coef_p7,atan2f_coef_p6 = [atan2f_GR_Addr_1],16
|
|
fma.s1 atan2f_T_denom = atan2f_Z0,atan2f_xsq,f8
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
ldfpd atan2f_coef_p5,atan2f_coef_p4 = [atan2f_GR_Addr_2],16
|
|
fma.s1 atan2f_Z = atan2f_Z0,f9,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
ldfpd atan2f_coef_p3,atan2f_coef_p2 = [atan2f_GR_Addr_1],16
|
|
fma.s1 atan2f_G_denom = atan2f_A0,atan2f_ysq,f9
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
ldfpd atan2f_const_piby2,atan2f_const_pi = [atan2f_GR_Addr_2],16
|
|
fma.s1 atan2f_A = atan2f_A0,f8,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
ldfpd atan2f_const_piby4,atan2f_const_3piby4 = [atan2f_GR_Addr_2]
|
|
fclass.m p11,p0 = f8,0xe7 // Test y @inf|@snan|@qnan|@zero
|
|
nop.i 999
|
|
}
|
|
{ .mfb
|
|
nop.m 999
|
|
fnma.s1 atan2f_T_numer = atan2f_Z0,atan2f_xy,f9
|
|
(p10) br.cond.spnt ATAN2F_XY_INF_NAN_ZERO ;; // Branch on x nan,inf,zero
|
|
}
|
|
|
|
|
|
// p6 if |y|>|x|, p7 if |x|>=|y| , use xsq and ysq for test
|
|
{ .mfi
|
|
nop.m 999
|
|
fcmp.gt.s1 p6,p7 = atan2f_ysq,atan2f_xsq
|
|
nop.i 999
|
|
}
|
|
{ .mfb
|
|
nop.m 999
|
|
fnma.s1 atan2f_G_numer = atan2f_A0,atan2f_xy,f8
|
|
(p11) br.cond.spnt ATAN2F_XY_INF_NAN_ZERO ;; // Branch on y nan,inf,zero
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p8) fma.s1 atan2f_const_1 = atan2f_sgn_Y,f0,f0
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
(p9) fma.s1 atan2f_const_1 = atan2f_sgn_Y,f1,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p6) fnma.s1 atan2f_U = atan2f_Z,f1,f0
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
(p6) fma.s1 atan2f_Usq = atan2f_Z,atan2f_Z,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p7) fma.s1 atan2f_U = atan2f_A,f1,f0
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
(p7) fma.s1 atan2f_Usq = atan2f_A,atan2f_A,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p6) frcpa.s1 atan2f_Q1,p0 = f1,atan2f_T_denom
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
(p6) fma.s1 atan2f_R_denom = atan2f_T_denom,f1,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p7) frcpa.s1 atan2f_Q1,p0 = f1,atan2f_G_denom
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
(p7) fma.s1 atan2f_R_denom = atan2f_G_denom,f1,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p6) fnma.s1 atan2f_R_numer = atan2f_T_numer,f1,f0
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
(p7) fma.s1 atan2f_R_numer = atan2f_G_numer,f1,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p6) fnma.s1 atan2f_p1rnum = atan2f_T_numer,atan2f_coef_p1,f0
|
|
nop.i 999 ;;
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
(p7) fma.s1 atan2f_p1rnum = atan2f_G_numer,atan2f_coef_p1,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_U4 = atan2f_Usq,atan2f_Usq,f0
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_poly_u109 = atan2f_Usq,atan2f_coef_p10,atan2f_coef_p9
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_poly_u87 = atan2f_Usq,atan2f_coef_p8,atan2f_coef_p7
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_poly_u65 = atan2f_Usq,atan2f_coef_p6,atan2f_coef_p5
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_poly_u43 = atan2f_Usq,atan2f_coef_p4,atan2f_coef_p3
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fnma.s1 atan2f_Q_beta = atan2f_Q1,atan2f_R_denom,f1
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_poly_u21 = atan2f_Usq,atan2f_coef_p2,atan2f_coef_p1
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_r = atan2f_Q1,atan2f_R_numer,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p6) fma.s1 atan2f_C = atan2f_sgn_Y,atan2f_const_piby2,f0
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
(p7) fma.s1 atan2f_C = atan2f_const_1,atan2f_const_pi,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_U6 = atan2f_U4,atan2f_Usq,f0
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_U8 = atan2f_U4,atan2f_U4,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_poly_u10to7 = atan2f_U4,atan2f_poly_u109,atan2f_poly_u87
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_pR = atan2f_p1rnum,atan2f_Q1,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_poly_u6to3 = atan2f_U4,atan2f_poly_u65,atan2f_poly_u43
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_Q2 = atan2f_Q1,atan2f_Q_beta,atan2f_Q1
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_Q_beta2 = atan2f_Q_beta,atan2f_Q_beta,f0
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_rsq = atan2f_r,atan2f_r,f0
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_poly_u210 = atan2f_Usq,atan2f_poly_u21,f1
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fcmp.eq.s0 p8,p0 = f8,f9 // Dummy op to set flag on denormal inputs
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_poly_u10to3 = atan2f_U8,atan2f_poly_u10to7,atan2f_poly_u6to3
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_Q3 = atan2f_Q2,atan2f_Q_beta2,atan2f_Q2
|
|
nop.i 999
|
|
}
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_pRC = atan2f_rsq,atan2f_pR,atan2f_C
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_poly_u10to0 = atan2f_U6,atan2f_poly_u10to3,atan2f_poly_u210
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fma.s1 atan2f_pQRC = atan2f_R_numer,atan2f_Q3,atan2f_pRC
|
|
nop.i 999 ;;
|
|
}
|
|
|
|
{ .mfb
|
|
nop.m 999
|
|
fma.s.s0 f8 = atan2f_U,atan2f_poly_u10to0,atan2f_pQRC
|
|
br.ret.sptk b0 ;;
|
|
}
|
|
|
|
|
|
|
|
ATAN2F_XY_INF_NAN_ZERO:
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fclass.m p10,p0 = f8,0xc3 // Is y nan
|
|
nop.i 999
|
|
}
|
|
;;
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fclass.m p12,p0 = f9,0xc3 // Is x nan
|
|
nop.i 999
|
|
}
|
|
;;
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fclass.m p6,p0 = f9,0x21 // Is x +inf
|
|
nop.i 999
|
|
}
|
|
{ .mfb
|
|
nop.m 999
|
|
(p10) fma.s.s0 f8 = f9,f8,f0 // Result quietized y if y is nan
|
|
(p10) br.ret.spnt b0 // Exit if y is nan
|
|
}
|
|
;;
|
|
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p6) fclass.m.unc p7,p8 = f8,0x23 // x +inf, is y inf
|
|
nop.i 999
|
|
}
|
|
{ .mfb
|
|
nop.m 999
|
|
(p12) fnorm.s.s0 f8 = f9 // Result quietized x if x is nan, y not nan
|
|
(p12) br.ret.spnt b0 // Exit if x is nan, y not nan
|
|
}
|
|
;;
|
|
|
|
// Here if x or y inf, or x or y zero
|
|
{ .mfi
|
|
nop.m 999
|
|
fcmp.eq.s0 p15,p0 = f8,f9 // Dummy op to set flag on denormal inputs
|
|
nop.i 999
|
|
}
|
|
;;
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fclass.m p11,p12 = f9,0x22 // Is x -inf
|
|
nop.i 999
|
|
}
|
|
{ .mfb
|
|
nop.m 999
|
|
(p7) fma.s.s0 f8 = atan2f_sgn_Y, atan2f_const_piby4,f0 // Result +-pi/4
|
|
(p7) br.ret.spnt b0 // Exit if x +inf and y inf
|
|
}
|
|
;;
|
|
|
|
{ .mfb
|
|
nop.m 999
|
|
(p8) fmerge.s f8 = f8,f0 // If x +inf and y not inf, result +-0
|
|
(p8) br.ret.spnt b0 // Exit if x +inf and y not inf
|
|
}
|
|
;;
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p12) fclass.m.unc p13,p0 = f8,0x23 // x not -inf, is y inf
|
|
nop.i 999
|
|
}
|
|
;;
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p11) fclass.m.unc p14,p15 = f8,0x23 // x -inf, is y inf
|
|
nop.i 999
|
|
}
|
|
;;
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
fclass.m p6,p7 = f9,0x7 // Is x zero
|
|
nop.i 999
|
|
}
|
|
{ .mfb
|
|
nop.m 999
|
|
(p13) fma.s.s0 f8 = atan2f_sgn_Y, atan2f_const_piby2,f0 // Result +-pi/2
|
|
(p13) br.ret.spnt b0 // Exit if x not -inf and y inf
|
|
}
|
|
;;
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p14) fma.s.s0 f8 = atan2f_sgn_Y, atan2f_const_3piby4,f0 // Result +-3pi/4
|
|
nop.i 999
|
|
}
|
|
{ .mfb
|
|
nop.m 999
|
|
(p15) fma.s.s0 f8 = atan2f_sgn_Y, atan2f_const_pi,f0 // Result +-pi
|
|
(p11) br.ret.spnt b0 // Exit if x -inf
|
|
}
|
|
;;
|
|
|
|
// Here if x or y zero
|
|
{ .mfi
|
|
nop.m 999
|
|
(p7) fclass.m.unc p8,p9 = f9,0x19 // x not zero, y zero, is x > zero
|
|
nop.i 999
|
|
}
|
|
;;
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p6) fclass.m.unc p10,p11 = f8,0x7 // x zero, is y zero
|
|
nop.i 999
|
|
}
|
|
;;
|
|
|
|
{ .mfi
|
|
nop.m 999
|
|
(p8) fmerge.s f8 = f8, f0 // x > zero and y zero, result is +-zero
|
|
nop.i 999
|
|
}
|
|
{ .mfb
|
|
nop.m 999
|
|
(p9) fma.s.s0 f8 = atan2f_sgn_Y, atan2f_const_pi,f0 // x < 0, y 0, result +-pi
|
|
(p10) br.cond.spnt __libm_error_region // Branch if x zero and y zero
|
|
}
|
|
;;
|
|
|
|
{ .mfb
|
|
nop.m 999
|
|
(p11) fma.s.s0 f8 = atan2f_sgn_Y, atan2f_const_piby2,f0 // x zero, y not zero
|
|
br.ret.sptk b0 // Final special case exit
|
|
}
|
|
;;
|
|
|
|
|
|
GLOBAL_IEEE754_END(atan2f)
|
|
libm_alias_float_other (__atan2, atan2)
|
|
|
|
|
|
LOCAL_LIBM_ENTRY(__libm_error_region)
|
|
.prologue
|
|
mov GR_Parameter_TAG = 38
|
|
fclass.m p10,p11 = f9,0x5 // @zero | @pos
|
|
;;
|
|
(p10) fmerge.s f10 = f8, f0
|
|
(p11) fma.s.s0 f10 = atan2f_sgn_Y, atan2f_const_pi,f0
|
|
;;
|
|
|
|
{ .mfi
|
|
add GR_Parameter_Y=-32,sp // Parameter 2 value
|
|
nop.f 999
|
|
.save ar.pfs,GR_SAVE_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
|
|
}
|
|
;;
|
|
|
|
{ .mmi
|
|
stfs [GR_Parameter_Y] = f9,16 // 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
|
|
}
|
|
;;
|
|
|
|
|
|
.body
|
|
{ .mib
|
|
stfs [GR_Parameter_X] = f8 // Store Parameter 1 on stack
|
|
add GR_Parameter_RESULT = 0,GR_Parameter_Y
|
|
nop.b 0 // Parameter 3 address
|
|
}
|
|
{ .mib
|
|
stfs [GR_Parameter_Y] = f10 // 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
|
|
};;
|
|
|
|
{ .mmi
|
|
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
|
|
}
|
|
;;
|
|
|
|
{ .mib
|
|
mov gp = GR_SAVE_GP // Restore gp
|
|
mov ar.pfs = GR_SAVE_PFS // Restore ar.pfs
|
|
br.ret.sptk b0 // Return
|
|
}
|
|
;;
|
|
|
|
LOCAL_LIBM_END(__libm_error_region)
|
|
|
|
.type __libm_error_support#,@function
|
|
.global __libm_error_support#
|