libm_lgammaf.s

glibc 2.9,最新版的C语言库函数

      shladd        GR_ad_T = GR_Ind4T,3,GR_ad_Data
      fma.s1        FR_Xf4 = FR_Xf2,FR_Xf2,f0
      nop.i         0
};;
{ .mfi
(p14) sub           GR_SignOfGamma = r0,GR_SignOfGamma // set signgam to -1
      fma.s1        FR_S6 = FR_S8,FR_Xf2,FR_S6
      nop.i         0
}
{ .mfi
      // set p9  if signgum is 32-bit int
      // set p10 if signgum is 64-bit int
      cmp.eq        p10,p9 = 8,r34
      fma.s1        FR_S2 = FR_S4,FR_Xf2,FR_S2
      nop.i         0
};;
{ .mfi
      ldfd          FR_Tf = [GR_ad_T]
      fma.s1        FR_Ln = FR_P32,FR_r,FR_T
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_LnSqrt2Pi = FR_LnSqrt2Pi,f1,FR_NormX
      nop.i         0
};;
.pred.rel "mutex",p9,p10
{ .mfi
(p9)  st4           [r33] = GR_SignOfGamma  // as 32-bit int
      fma.s1        FR_rf2 = FR_rf,FR_rf,f0
      nop.i         0
}
{ .mfi
(p10) st8           [r33] = GR_SignOfGamma  // as 64-bit int
      fma.s1        FR_S10 = FR_S14,FR_Xf4,FR_S10
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_P32f = FR_P3,FR_rf,FR_P2
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_Xf8 = FR_Xf4,FR_Xf4,f0
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_P10f = FR_P1,FR_rf,f1
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_S2 = FR_S6,FR_Xf4,FR_S2
      nop.i         0
};;
{ .mfi
      nop.m         0
      fms.s1        FR_Ln = FR_Ln,FR_xm05,FR_LnSqrt2Pi
      nop.i         0
};;
{ .mfi
      nop.m         0
      fcvt.xf       FR_Nf = FR_int_Ln
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_S2 = FR_S10,FR_Xf8,FR_S2
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_Tf = FR_Nf,FR_Ln2,FR_Tf
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_P32f = FR_P32f,FR_rf2,FR_P10f // ??????
      nop.i         0
};;
{ .mfi
      nop.m         0
      fnma.s1       FR_Ln = FR_S2,FR_Xf2,FR_Ln
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_Lnf = FR_P32f,FR_rf,FR_Tf
      nop.i         0
};;
{ .mfb
      nop.m         0
      fms.s.s0      f8 = FR_Ln,f1,FR_Lnf
      br.ret.sptk   b0
};;
// branch for calculating of ln(GAMMA(x)) for -2^13 < x < -9
//---------------------------------------------------------------------
.align 32
lgammaf_negpoly:
{ .mfi
      getf.d        GR_Arg = FR_Xf
      frcpa.s1      FR_InvXf,p0 = f1,FR_Xf
      mov           GR_ExpBias = 0x3FF
}
{ .mfi
      nop.m         0
      fma.s1        FR_Xf2 = FR_Xf,FR_Xf,f0
      nop.i         0
};;
{ .mfi
      getf.sig      GR_N = FR_int_Ntrunc
      fcvt.xf       FR_N = FR_int_Ln
      mov           GR_SignOfGamma = 1
}
{ .mfi
      nop.m         0
      fma.s1        FR_A9 = FR_A10,FR_x,FR_A9
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_P10 = FR_P1,FR_r,f1
      extr.u        GR_Exp = GR_Arg,52,11
}
{ .mfi
      nop.m         0
      fma.s1        FR_x4 = FR_x2,FR_x2,f0
      nop.i         0
};;
{ .mfi
      sub           GR_PureExp = GR_Exp,GR_ExpBias
      fma.s1        FR_A7 = FR_A8,FR_x,FR_A7
      tbit.z        p14,p0 = GR_N,0
}
{ .mfi
      nop.m         0
      fma.s1        FR_A5 = FR_A6,FR_x,FR_A5
      nop.i         0
};;
{ .mfi
      setf.sig      FR_int_Ln = GR_PureExp
      fma.s1        FR_A3 = FR_A4,FR_x,FR_A3
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_A1 = FR_A2,FR_x,FR_A1
(p14) sub           GR_SignOfGamma = r0,GR_SignOfGamma
};;
{ .mfi
      nop.m         0
      fms.s1        FR_rf = FR_InvXf,FR_Xf,f1
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_Xf4 = FR_Xf2,FR_Xf2,f0
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_S14 = FR_S16,FR_Xf2,FR_S14
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_S10 = FR_S12,FR_Xf2,FR_S10
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_T = FR_N,FR_Ln2,FR_T
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_P32 = FR_P32,FR_r2,FR_P10
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_S6 = FR_S8,FR_Xf2,FR_S6
      extr.u        GR_Ind4T = GR_Arg,44,8
}
{ .mfi
      nop.m         0
      fma.s1        FR_S2 = FR_S4,FR_Xf2,FR_S2
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_A7 = FR_A9,FR_x2,FR_A7
      nop.i         0
}
{ .mfi
      shladd        GR_ad_T = GR_Ind4T,3,GR_ad_Data
      fma.s1        FR_A3 = FR_A5,FR_x2,FR_A3
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_Xf8 = FR_Xf4,FR_Xf4,f0
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_rf2 = FR_rf,FR_rf,f0
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_P32f = FR_P3,FR_rf,FR_P2
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_P10f = FR_P1,FR_rf,f1
      nop.i         0
};;
{ .mfi
      ldfd          FR_Tf = [GR_ad_T]
      fma.s1        FR_Ln = FR_P32,FR_r,FR_T
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_A0 = FR_A1,FR_x,FR_A0
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_S10 = FR_S14,FR_Xf4,FR_S10
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_S2 = FR_S6,FR_Xf4,FR_S2
      nop.i         0
};;
{ .mfi
      nop.m         0
      fcvt.xf       FR_Nf = FR_int_Ln
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_A3 = FR_A7,FR_x4,FR_A3
      nop.i         0
};;
{ .mfi
      nop.m         0
      fcmp.eq.s1    p13,p0 = FR_NormX,FR_Ntrunc
      nop.i         0
}
{ .mfi
      nop.m         0
      fnma.s1       FR_x3 = FR_x2,FR_x,f0 // -x^3
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_P32f = FR_P32f,FR_rf2,FR_P10f
      nop.i         0
};;
{ .mfb
      // set p9  if signgum is 32-bit int
      // set p10 if signgum is 64-bit int
      cmp.eq        p10,p9 = 8,r34
      fma.s1        FR_S2 = FR_S10,FR_Xf8,FR_S2
(p13) br.cond.spnt  lgammaf_singularity
};;
.pred.rel "mutex",p9,p10
{ .mmf
(p9)  st4           [r33] = GR_SignOfGamma  // as 32-bit int
(p10) st8           [r33] = GR_SignOfGamma  // as 64-bit int
      fms.s1        FR_A0 = FR_A3,FR_x3,FR_A0 // -A3*x^3-A0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_Tf = FR_Nf,FR_Ln2,FR_Tf
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_Ln = FR_S2,FR_Xf2,FR_Ln // S2*Xf^2+Ln
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_Lnf = FR_P32f,FR_rf,FR_Tf
      nop.i         0
};;
{ .mfi
      nop.m         0
      fms.s1        FR_Ln = FR_A0,f1,FR_Ln
      nop.i         0
};;
{ .mfb
      nop.m         0
      fms.s.s0      f8 = FR_Ln,f1,FR_Lnf
      br.ret.sptk   b0
};;
// branch for handling +/-0, NaT, QNaN, +/-INF and denormalised numbers
//---------------------------------------------------------------------
.align 32
lgammaf_spec:
{ .mfi
      getf.exp      GR_SignExp = FR_NormX
      fclass.m      p6,p0 = f8,0x21 // is arg +INF?
      mov           GR_SignOfGamma = 1 // set signgam to 1
};;
{ .mfi
      getf.sig      GR_Sig = FR_NormX
      fclass.m      p7,p0 = f8,0xB // is x deno?
      // set p11 if signgum is 32-bit int
      // set p12 if signgum is 64-bit int
      cmp.eq        p12,p11 = 8,r34
};;
.pred.rel "mutex",p11,p12
{ .mfi
      // store sign of gamma(x) as 32-bit int
(p11) st4           [r33] = GR_SignOfGamma
      fclass.m      p8,p0 = f8,0x1C0 // is arg NaT or NaN?
      dep.z         GR_Ind = GR_SignExp,3,4
}
{ .mib
      // store sign of gamma(x) as 64-bit int
(p12) st8           [r33] = GR_SignOfGamma
      and           GR_Exp = GR_ExpMask,GR_SignExp
(p6)  br.ret.spnt   b0 // exit for +INF
};;
{ .mfi
      sub           GR_PureExp = GR_Exp,GR_ExpBias
      fclass.m      p9,p0 = f8,0x22 // is arg -INF?
      extr.u        GR_Ind4T = GR_Sig,55,8
}
{ .mfb
      nop.m         0
(p7)  fma.s0        FR_tmp = f1,f1,f8
(p7)  br.cond.sptk  lgammaf_core
};;
{ .mfb
      nop.m         0
(p8)  fms.s.s0      f8 = f8,f1,f8
(p8)  br.ret.spnt   b0 // exit for NaT and NaN
};;
{ .mfb
      nop.m         0
(p9)  fmerge.s      f8 = f1,f8
(p9)  br.ret.spnt   b0 // exit -INF
};;
// branch for handling negative integers and +/-0
//---------------------------------------------------------------------
.align 32
lgammaf_singularity:
{ .mfi
      mov           GR_SignOfGamma = 1 // set signgam to 1
      fclass.m      p6,p0 = f8,0x6 // is x -0?
      mov           GR_TAG = 109 // negative
}
{ .mfi
      mov           GR_ad_SignGam = r33
      fma.s1        FR_X = f0,f0,f8
      nop.i         0
};;
{ .mfi
      nop.m         0
      frcpa.s0      f8,p0 = f1,f0
      // set p9  if signgum is 32-bit int
      // set p10 if signgum is 64-bit int
      cmp.eq        p10,p9 = 8,r34
}
{ .mib
      nop.m         0
(p6)  sub           GR_SignOfGamma = r0,GR_SignOfGamma
      br.cond.sptk  lgammaf_libm_err
};;
// overflow (x > OVERFLOV_BOUNDARY)
//---------------------------------------------------------------------
.align 32
lgammaf_overflow:
{ .mfi
      nop.m         0
      nop.f         0
      mov           r8 = 0x1FFFE
};;
{ .mfi
      setf.exp      f9 = r8
      fmerge.s      FR_X = f8,f8
      mov           GR_TAG = 108 // overflow
};;
{ .mfi
      mov           GR_ad_SignGam = r33
      nop.f         0
      // set p9  if signgum is 32-bit int
      // set p10 if signgum is 64-bit int
      cmp.eq        p10,p9 = 8,r34
}
{ .mfi
      nop.m         0
      fma.s.s0      f8 = f9,f9,f0 // Set I,O and +INF result
      nop.i         0
};;
// gate to __libm_error_support#
//---------------------------------------------------------------------
.align 32
lgammaf_libm_err:
{ .mmi
      alloc        r32 = ar.pfs,1,4,4,0
      mov          GR_Parameter_TAG = GR_TAG
      nop.i        0
};;
.pred.rel "mutex",p9,p10
{ .mmi
      // store sign of gamma(x) as 32-bit int
(p9)  st4          [GR_ad_SignGam] = GR_SignOfGamma
      // store sign of gamma(x) as 64-bit int
(p10) st8          [GR_ad_SignGam] = GR_SignOfGamma
      nop.i        0
};;
GLOBAL_LIBM_END(__libm_lgammaf)


LOCAL_LIBM_