libm_lgamma.s

glibc 2.9,最新版的C语言库函数

{ .mfi
      ldfe          FR_C71 = [GR_ad_Ce],32
(p14) fma.s1        FR_Xp1 = f1,f1,FR_NormX // x+1
      cmp.eq        p6,p7 = GR_ExpOf2,GR_SignExp
};;
.pred.rel "mutex",p6,p7
{ .mfi
      ldfe          FR_C81 = [GR_ad_Co],32
(p6)  fma.s1        FR_x = f0,f0,FR_NormX
      shladd        GR_Offs7 = GR_Ind,2,GR_Ind // (ind*16)*5
}
{ .mfi
      ldfe          FR_C91 = [GR_ad_Ce],32
(p7)  fms.s1        FR_x = FR_x,f1,f1
      add           GR_ad_Co7 = 0x800,GR_ad_Data
};;
{ .mfi
      ldfe          FR_CA1 = [GR_ad_Co],32
(p14) fma.s1        FR_3 = f1,f1,FR_2
      shladd        GR_Offs7 = GR_Ind,1,GR_Offs7 // (ind*16)*7
}
{ .mfi
      ldfe          FR_C00 = [GR_ad_Ce],32
(p14) fma.s1        FR_Xp4 = FR_2,FR_2,FR_NormX
      add           GR_ad_Ce7 = 0x810,GR_ad_Data
};;
{ .mfi
      ldfe          FR_C10 = [GR_ad_Co],32
(p6)  fms.s1        FR_Xm2 = FR_w,f1,f1
      add           GR_ad_Co7 = GR_ad_Co7,GR_Offs7
}
{ .mfi
      ldfe          FR_C20 = [GR_ad_Ce],32
(p14) fma.s1        FR_r2 = FR_r,FR_r,f0 // log(x)
      add           GR_ad_Ce7 = GR_ad_Ce7,GR_Offs7
};;
{ .mfi
      ldfe          FR_C30 = [GR_ad_Co],32
(p14) fms.s1        FR_Xf = FR_NormX,f1,FR_N  // xf = x - [x]
(p14) mov           GR_Arg17 = 0xC031 // -17
}
{ .mfi
      ldfe          FR_C40 = [GR_ad_Ce],32
(p14) fma.s1        FR_Xp5 = FR_5,f1,FR_NormX
(p14) sub           GR_Exp = GR_Exp,GR_ExpBias
};;
{ .mfi
      ldfe          FR_C50 = [GR_ad_Co7],32
(p14) fms.s1        FR_Xfr = FR_Xp1,f1,FR_Ntrunc // xfr = (x+1) - [x]
(p14) cmp.lt.unc    p13,p0 = GR_Arg,GR_Arg17
}
{ .mfb
      ldfe          FR_C60 = [GR_ad_Ce7],32
(p14) fma.s1        FR_Xp10 = FR_5,FR_2,FR_NormX
      // jump if the input argument is negative and great than -17.0
(p13) br.cond.spnt  lgamma_negrecursion
};;
{ .mfi
      ldfe          FR_C70 = [GR_ad_Co7],32
      fma.s1        FR_C01 = FR_x,f1,FR_C01
(p14) add           GR_ad_Ce = 0x1310,GR_ad_2
}
{ .mfi
      ldfe          FR_C80 = [GR_ad_Ce7],32
      fma.s1        FR_C11 = FR_x,f1,FR_C11
(p14) add           GR_ad_Co = 0x12C0,GR_ad_2
};;
{ .mfi
      ldfe          FR_C90 = [GR_ad_Co7],32
      fma.s1        FR_C21 = FR_x,f1,FR_C21
      nop.i         0
}
{ .mfi
      ldfe          FR_CA0 = [GR_ad_Ce7],32
      fma.s1        FR_C31 = FR_x,f1,FR_C31
      nop.i         0
};;
{ .mfi
      ldfe          FR_CN = [GR_ad_Co7],32
      fma.s1        FR_C41 = FR_x,f1,FR_C41
      nop.i         0
}
{ .mfi
(p14) ldfpd         FR_P5,FR_P4 = [GR_ad_1],16
      fma.s1        FR_C51 = FR_x,f1,FR_C51
      nop.i         0
};;
{ .mfi
(p14) ldfpd         FR_P3,FR_P2 = [GR_ad_2],16
      fma.s1        FR_C61 = FR_x,f1,FR_C61
      nop.i         0
}
{ .mfi
(p14) ldfe          FR_Ln2 = [GR_ad_1]
      fma.s1        FR_C71 = FR_x,f1,FR_C71
      nop.i         0
};;
{ .mfi
(p14) ldfpd         FR_S28,FR_S26 = [GR_ad_Co],16
      fma.s1        FR_C81 = FR_x,f1,FR_C81
      add           GR_ad_2 = 0x60,GR_ad_2
}
{ .mfi
(p14) ldfpd         FR_S24,FR_S22 = [GR_ad_Ce],16
      fma.s1        FR_C91 = FR_x,f1,FR_C91
      nop.i         0
};;
{ .mfi
(p14) ldfpd         FR_S20,FR_S18 = [GR_ad_Co],16
      fma.s1        FR_CA1 = FR_x,f1,FR_CA1
      nop.i         0
}
{ .mfi
(p14) ldfpd         FR_S16,FR_S14 = [GR_ad_Ce],16
      fma.s1        FR_C01 = FR_C01,FR_x,FR_C00
      nop.i         0
};;
{ .mfi
(p14) getf.exp      GR_SignExp = FR_Xf
      fma.s1        FR_C11 = FR_C11,FR_x,FR_C10
      nop.i         0
}
{ .mfi
(p14) ldfe          FR_S12 = [GR_ad_Co],16
      fma.s1        FR_C21 = FR_C21,FR_x,FR_C20
      nop.i         0
};;
{ .mfi
(p14) getf.sig      GR_Sig = FR_Xf
(p14) frcpa.s1      FR_InvXf,p0 = f1,FR_Xf
      nop.i         0
}
{ .mfi
(p14) ldfe          FR_S10 = [GR_ad_Ce],16
      fma.s1        FR_C41 = FR_C41,FR_x,FR_C40
      nop.i         0
};;
{ .mfi
(p14) ldfe          FR_S8 = [GR_ad_Co],16
      fma.s1        FR_C51 = FR_C51,FR_x,FR_C50
      nop.i         0
}
{ .mfi
(p14) ldfe          FR_S6 = [GR_ad_Ce],16
      fma.s1        FR_C61 = FR_C61,FR_x,FR_C60
(p14) and           GR_Expf = GR_SignExp,GR_ExpMask
};;
{ .mfi
(p14) sub           GR_Expf = GR_Expf,GR_ExpBias
      fma.s1        FR_C71 = FR_C71,FR_x,FR_C70
(p14) shl           GR_Ind = GR_Sig,1
}
{ .mfi
(p14) ldfe          FR_S4 = [GR_ad_Co],16
      fma.s1        FR_C81 = FR_C81,FR_x,FR_C80
(p14) cmp.eq.unc    p8,p0 = 0,GR_Sig
};;
{ .mfi
(p14) setf.sig      FR_int_Nf = GR_Expf
      fma.s1        FR_C91 = FR_C91,FR_x,FR_C90
(p14) shr.u         GR_Ind = GR_Ind,56
}
{ .mfb
(p14) ldfe          FR_S2 = [GR_ad_Ce],16
      fma.s1        FR_CA1 = FR_CA1,FR_x,FR_CA0
      // jump if the input argument is integer number from range (-512.0;-17.0]
(p8)  br.cond.spnt  lgamma_singularity
};;
{ .mfi
(p14) getf.sig      GR_Sig = FR_int_Ntrunc
      fma.s1        FR_C01 = FR_C01,FR_C11,f0
      nop.i         0
}
{ .mfi
(p14) shladd        GR_ad_T = GR_Ind,4,GR_ad_2
      fma.s1        FR_C31 = FR_C31,FR_x,FR_C30
      nop.i         0
};;
{ .mfi
(p14) ldfe          FR_Tf = [GR_ad_T]
(p14) fms.s1        FR_rf = FR_InvXf,FR_Xf,f1 // reduced arg for log({x})
(p14) extr.u        GR_Ind = GR_ArgAsIs,44,8
}
{ .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_C21 = FR_C21,FR_C41,f0
      mov           GR_SignOfGamma = 1
};;
{ .mfi
      nop.m         0
      fma.s1        FR_C51 = FR_C51,FR_C61,f0
(p14) tbit.z.unc    p8,p0 = GR_Sig,0
}
{ .mfi
(p14) shladd        GR_ad_T = GR_Ind,4,GR_ad_2
(p6)  fma.s1        FR_CN = FR_CN,FR_Xm2,f0
      nop.i         0
};;
{ .mfi
(p14) setf.sig      FR_int_N = GR_Exp
      fma.s1        FR_C71 = FR_C71,FR_C81,f0
(p8)  sub           GR_SignOfGamma = r0,GR_SignOfGamma
}
{ .mfi
      nop.m         0
(p14) fma.s1        FR_Xf2 = FR_Xf,FR_Xf,f0
      nop.i         0
};;
{ .mfi
(p14) ldfe          FR_T = [GR_ad_T]
      fma.s1        FR_C91 = FR_C91,FR_CA1,f0
      nop.i         0
}
{ .mfi
      nop.m         0
(p14) fma.s1        FR_r2 = FR_r,FR_r,f0
      nop.i         0
};;
.pred.rel "mutex",p9,p10
{ .mfi
      // store sign of gamma(x) as 32-bit int
(p9)  st4           [r33] = GR_SignOfGamma
      fma.s1        FR_C01 = FR_C01,FR_C31,f0
      nop.i         0
}
{ .mfi
      // store sign of gamma(x) as 64-bit int
(p10) st8           [r33] = GR_SignOfGamma
(p14) fma.s1        FR_P54 = FR_P5,FR_r,FR_P4
      nop.i         0
};;
{ .mfi
      nop.m         0
(p14) fma.s1        FR_P32 = FR_P3,FR_r,FR_P2
      nop.i         0
}
{ .mfb
      nop.m         0
(p14) fma.s1        FR_P54f = FR_P5,FR_rf,FR_P4
      // jump if the input argument is non-integer from range (-512.0;-17.0]
(p14) br.cond.spnt  lgamma_negpoly
};;
{ .mfi
      nop.m         0
      fma.s1        FR_C21 = FR_C21,FR_C51,f0
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_C71 = FR_C71,FR_C91,f0
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_CN  = FR_C01,FR_CN,f0
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_C21 = FR_C21,FR_C71,f0
      nop.i         0
};;
{ .mfb
      nop.m         0
      fma.d.s0      f8 = FR_C21,FR_CN,f0
      br.ret.sptk   b0 // exit for arguments from range [2.25; 512.0)
};;
// branch for calculating of ln(GAMMA(x)) for -512 < x < -17
//---------------------------------------------------------------------
.align 32
lgamma_negpoly:
{ .mfi
      nop.m         0
      fma.s1        FR_Xf4 = FR_Xf2,FR_Xf2,f0
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_S28 = FR_S28,FR_Xf2,FR_S26
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_S24 = FR_S24,FR_Xf2,FR_S22
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_S20 = FR_S20,FR_Xf2,FR_S18
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_S16 = FR_S16,FR_Xf2,FR_S14
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_S12 = FR_S12,FR_Xf2,FR_S10
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_S8 = FR_S8,FR_Xf2,FR_S6
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_S4 = FR_S4,FR_Xf2,FR_S2
      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 // log(x)
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_r3 = FR_r2,FR_r,f0 // log(x)
      nop.i         0
}
{ .mfi
      nop.m         0
      fcvt.xf       FR_Nf = FR_int_Nf // log({x})
      nop.i         0
};;
{ .mfi
      nop.m         0
      fma.s1        FR_S28 = FR_S28,FR_Xf4,FR_S24
      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_S20 = FR_S20,FR_Xf4,FR_S16
      nop.i         0
}
{ .mfi
      nop.m         0
      fma.s1        FR_C21 = FR_C21