e_acoshl.s

glibc 2.9,最新版的C语言库函数

};;

{ .mfi
      ldfps      FR_G, FR_H   = [GR_ad_tbl_1],8     // Load G_1, H_1
      fmerge.se  FR_S_hi      =  f1,FR_XLog_Hi            // Form |x|
      nop.i 0
};;

{ .mmi
      getf.exp   GR_N         =  FR_XLog_Hi         // Get N = exponent of x+1
      ldfd       FR_h         = [GR_ad_tbl_1] // Load h_1
      nop.i 0
};;

{ .mfi
      ldfe       FR_log2_hi   = [GR_ad_q],16      // Load log2_hi
      nop.f 0
      pmpyshr2.u GR_X_1       = GR_X_0,GR_Z_1,15  // Get bits 30-15 of X_0 * Z_1
};;

{ .mmi
      ldfe       FR_log2_lo   = [GR_ad_q],16     // Load log2_lo
      sub        GR_N         = GR_N, GR_Bias 
      mov        GR_exp_2tom80 = 0x0ffaf         // Exponent of 2^-80
};;

{ .mfi
      ldfe       FR_Q4        = [GR_ad_q],16          // Load Q4
      nop.f 0
      sub        GR_minus_N   = GR_Bias, GR_N         // Form exponent of 2^(-N)
};;

{ .mmf
      ldfe       FR_Q3        = [GR_ad_q],16   // Load Q3
      setf.sig   FR_float_N   = GR_N        // Put integer N into rightmost sign
      nop.f 0
};;

{ .mmi
      ldfe       FR_Q2        = [GR_ad_q],16      // Load Q2
	  nop.m 0
      extr.u     GR_Index2    = GR_X_1, 6, 4      // Extract bits 6-9 of X_1 
};;

{ .mmi
      ldfe       FR_Q1        = [GR_ad_q]                // Load Q1
      shladd     GR_ad_z_2    = GR_Index2, 2, GR_ad_z_2  // Point to Z_2
      nop.i 0
};;

{ .mmi
      ld4        GR_Z_2       = [GR_ad_z_2]                // Load Z_2
      shladd     GR_ad_tbl_2  = GR_Index2, 4, GR_ad_tbl_2  // Point to G_2
	  nop.i 0
};;

{ .mmi
      ldfps      FR_G2, FR_H2 = [GR_ad_tbl_2],8       // Load G_2, H_2
      nop.m 0
      nop.i 0
};;

{ .mmf
      ldfd       FR_h2        = [GR_ad_tbl_2]         // Load h_2
      setf.exp FR_2_to_minus_N = GR_minus_N   // Form 2^(-N)
      nop.f 0
};;

{ .mfi
      nop.m 0
      nop.f 0
      pmpyshr2.u GR_X_2       = GR_X_1,GR_Z_2,15   // Get bits 30-15 of X_1*Z_2
};;

// WE CANNOT USE GR_X_2 IN NEXT 3 CYCLES ("DEAD" ZONE!) 
// BECAUSE OF POSSIBLE 10 CLOCKS STALL!
// (Just nops added - nothing to do here)

{ .mfi
      nop.m 0
      nop.f 0
      nop.i 0
};;

{ .mfi
      nop.m 0
      nop.f 0
      nop.i 0
};;

{ .mfi
      nop.m 0
      nop.f 0
      nop.i 0
};;

{ .mfi
      nop.m 0
      nop.f 0
      extr.u     GR_Index3    = GR_X_2, 1, 5          // Extract bits 1-5 of X_2
};;

{ .mfi
      shladd     GR_ad_tbl_3  = GR_Index3, 4, GR_ad_tbl_3  // Point to G_3
	  fcvt.xf    FR_float_N   = FR_float_N
      nop.i 0
};;

{ .mfi
      ldfps      FR_G3, FR_H3 = [GR_ad_tbl_3],8   // Load G_3, H_3
      nop.f 0
      nop.i 0
};;

{ .mfi
      ldfd       FR_h3        = [GR_ad_tbl_3]            // Load h_3
      fmpy.s1    FR_G         = FR_G, FR_G2              // G = G_1 * G_2
      nop.i 0
}
{ .mfi
      nop.m 0
      fadd.s1    FR_H         = FR_H, FR_H2              // H = H_1 + H_2
      nop.i 0
};;

{ .mmf
      nop.m 0
      nop.m 0
      fadd.s1    FR_h         = FR_h, FR_h2              // h = h_1 + h_2
};;

{ .mfi
      nop.m 0
      fmpy.s1    FR_G         = FR_G, FR_G3              // G = (G_1 * G_2)*G_3
      nop.i 0
}
{ .mfi
      nop.m 0
      fadd.s1    FR_H         = FR_H, FR_H3              // H = (H_1 + H_2)+H_3
      nop.i 0
};;

{ .mfi
      nop.m 0
      fadd.s1    FR_h         = FR_h, FR_h3            // h = (h_1 + h_2) + h_3
      nop.i 0
};;

{ .mfi
      nop.m 0
      fms.s1     FR_r         = FR_G, FR_S_hi, f1           // r = G * S_hi - 1
      nop.i 0
}
{ .mfi
      nop.m 0
      fma.s1     FR_Y_hi      = FR_float_N, FR_log2_hi, FR_H // Y_hi=N*log2_hi+H
      nop.i 0
};;

{ .mfi
      nop.m 0
      fma.s1     FR_h         = FR_float_N, FR_log2_lo, FR_h  // h = N*log2_lo+h
      nop.i 0
};;

{ .mfi
      nop.m 0
      fma.s1     FR_poly_lo   = FR_r, FR_Q4, FR_Q3      // poly_lo = r * Q4 + Q3
      nop.i 0
}
{ .mfi
      nop.m 0
      fmpy.s1    FR_rsq       = FR_r, FR_r              // rsq = r * r
      nop.i 0
};;

{ .mfi
      nop.m 0
      fma.s1     FR_poly_lo   = FR_poly_lo, FR_r, FR_Q2 // poly_lo=poly_lo*r+Q2
      nop.i 0
}
{ .mfi
      nop.m 0
      fma.s1     FR_rcub      = FR_rsq, FR_r, f0        // rcub = r^3
      nop.i 0
};;

{ .mfi
      nop.m 0
      fma.s1     FR_poly_hi   = FR_Q1, FR_rsq, FR_r     // poly_hi = Q1*rsq + r
      nop.i 0
};;

{ .mfi
      nop.m 0
      fma.s1     FR_poly_lo   = FR_poly_lo, FR_rcub, FR_h//poly_lo=poly_lo*r^3+h
      nop.i 0
};;
{ .mfi
      nop.m 0
      fadd.s0    FR_Y_lo      = FR_poly_hi, FR_poly_lo  // Y_lo=poly_hi+poly_lo 
      nop.i 0
};;
{ .mfb
      nop.m 0
      fadd.s0    FR_Res       = FR_Y_lo,FR_Y_hi    // Result=Y_lo+Y_hi
      br.ret.sptk   b0                        // Common exit
};;


// NEAR ONE INTERVAL
near_1:
{ .mfi 
      nop.m 0  
      frsqrta.s1 FR_Rcp, p0   = FR_2XM1 // Rcp = 1/x reciprocal appr. &SQRT&
      nop.i 0  
};;

{ .mfi 
      nop.m 0  
      fma.s1     FR_PV6       = FR_PP5, FR_XM1, FR_PP4 // pv6 = P5*xm1+P4 $POLY$
      nop.i 0  
}
{ .mfi
      nop.m 0  
	  fma.s1     FR_QV6       = FR_QQ5, FR_XM1, FR_QQ4 // qv6 = Q5*xm1+Q4 $POLY$
      nop.i 0  
};;

{ .mfi 
      nop.m 0  
	  fma.s1     FR_PV4       = FR_PP3, FR_XM1, FR_PP2 // pv4 = P3*xm1+P2 $POLY$
      nop.i 0  
}
{ .mfi
      nop.m 0  
	  fma.s1     FR_QV4       = FR_QQ3, FR_XM1, FR_QQ2 // qv4 = Q3*xm1+Q2 $POLY$
      nop.i 0  
};;

{ .mfi 
      nop.m 0  
	  fma.s1     FR_XM12      = FR_XM1, FR_XM1, f0 // xm1^2 = xm1 * xm1 $POLY$
      nop.i 0  
};;

{ .mfi 
      nop.m 0  
	  fma.s1     FR_PV2       = FR_PP1, FR_XM1, FR_PP0 // pv2 = P1*xm1+P0 $POLY$
      nop.i 0  
}
{ .mfi
      nop.m 0  
	  fma.s1     FR_QV2       = FR_QQ1, FR_XM1, FR_QQ0 // qv2 = Q1*xm1+Q0 $POLY$
      nop.i 0  
};;

{ .mfi 
      nop.m 0  
      fma.s1     FR_GG        = FR_Rcp, FR_2XM1, f0 // g = Rcp * x &SQRT&                
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fma.s1     FR_HH        = FR_Half, FR_Rcp, f0 // h = 0.5 * Rcp &SQRT&
      nop.i 0  
};;


{ .mfi 
      nop.m 0  
	  fma.s1    FR_PV3       = FR_XM12, FR_PV6, FR_PV4//pv3=pv6*xm1^2+pv4 $POLY$
      nop.i 0  
}
{ .mfi
      nop.m 0  
	  fma.s1    FR_QV3       = FR_XM12, FR_QV6, FR_QV4//qv3=qv6*xm1^2+qv4 $POLY$
      nop.i 0  
};;


{ .mfi 
      nop.m 0  
      fnma.s1   FR_EE        = FR_GG, FR_HH, FR_Half   // e = 0.5 - g * h &SQRT&
      nop.i 0  
};;

{ .mfi 
      nop.m 0  
	  fma.s1    FR_PP        = FR_XM12, FR_PV3, FR_PV2 //pp=pv3*xm1^2+pv2 $POLY$
      nop.i 0  
}
{ .mfi
      nop.m 0  
	  fma.s1    FR_QQ        = FR_XM12, FR_QV3, FR_QV2 //qq=qv3*xm1^2+qv2 $POLY$
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_GG        = FR_GG, FR_EE, FR_GG  // g = g * e + g &SQRT&
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fma.s1     FR_HH        = FR_HH, FR_EE, FR_HH  // h = h * e + h &SQRT&
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      frcpa.s1   FR_Y0,p0     = f1,FR_QQ // y = frcpa(b)  #DIV#
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fnma.s1    FR_EE        = FR_GG, FR_HH, FR_Half // e = 0.5 - g*h &SQRT&
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_Q0        = FR_PP,FR_Y0,f0 // q = a*y  #DIV#
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fnma.s1    FR_E0        = FR_Y0,FR_QQ,f1 // e = 1 - b*y  #DIV#
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_GG        = FR_GG, FR_EE, FR_GG // g = g * e + g &SQRT&             
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fma.s1     FR_HH        = FR_HH, FR_EE, FR_HH // h = h * e + h &SQRT&
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_E2        = FR_E0,FR_E0,FR_E0 // e2 = e+e^2 #DIV#
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fma.s1     FR_E1        = FR_E0,FR_E0,f0 // e1 = e^2 #DIV#
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fnma.s1   FR_EE        = FR_GG, FR_HH, FR_Half   // e = 0.5 - g * h &SQRT&
      nop.i 0  
}
{ .mfi
      nop.m 0  
	  fnma.s1   FR_DD        = FR_GG, FR_GG, FR_2XM1   // d = x - g * g &SQRT&
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_Y1        = FR_Y0,FR_E2,FR_Y0 // y1 = y+y*e2 #DIV#
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fma.s1     FR_E3        = FR_E1,FR_E1,FR_E0 // e3 = e+e1^2 #DIV#
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_GG        = FR_DD, FR_HH, FR_GG // g = d * h + g &SQRT&
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fma.s1     FR_HH        = FR_HH, FR_EE, FR_HH // h = h * e + h &SQRT&
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_Y2        = FR_Y1,FR_E3,FR_Y0 // y2 = y+y1*e3 #DIV#
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fnma.s1    FR_R0        = FR_QQ,FR_Q0,FR_PP // r = a-b*q #DIV#
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fnma.s1    FR_DD        = FR_GG, FR_GG, FR_2XM1 // d = x - g * g &SQRT&             
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fnma.s1    FR_E4        = FR_QQ,FR_Y2,f1    // e4 = 1-b*y2 #DIV#
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fma.s1     FR_X_Hi      = FR_R0,FR_Y2,FR_Q0 // x = q+r*y2 #DIV#
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_GL        = FR_DD, FR_HH, f0   // gl = d * h &SQRT&
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_Y3        = FR_Y2,FR_E4,FR_Y2 // y3 = y2+y2*e4 #DIV#
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fnma.s1    FR_R1        = FR_QQ,FR_X_Hi,FR_PP // r1 = a-b*x #DIV#
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_HH        = FR_GG, FR_X_Hi, f0 // hh = gg * x_hi
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fma.s1     FR_LH        = FR_GL, FR_X_Hi, f0 // lh = gl * x_hi
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_X_lo      = FR_R1,FR_Y3,f0 // x_lo = r1*y3 #DIV#
      nop.i 0  
};;

{ .mfi
      nop.m 0  
      fma.s1     FR_LL        = FR_GL, FR_X_lo, f0 // ll = gl*x_lo
      nop.i 0  
}
{ .mfi
      nop.m 0  
      fma.s1     FR_HL        = FR_GG, FR_X_lo, f0 // hl = gg * x_lo
      nop.i 0  
};;

{ .mfi
      nop.m 0  
	  fms.s1     FR_Res       = FR_GL,  f1, FR_LL // res = gl + ll
      nop.i 0  
};;

{ .mfi
      nop.m 0  
	  fms.s1     FR_Res       = FR_Res, f1, FR_LH // res = res + lh
      nop.i 0  
};;

{ .mfi
      nop.m 0  
	  fms.s1     FR_Res       = FR_Res, f1, FR_HL // res = res + hl
      nop.i 0  
};;

{ .mfi
      nop.m 0  
	  fms.s1     FR_Res       = FR_Res, f1, FR_HH // res = res + hh
      nop.i 0  
};;

{ .mfb
      nop.m 0  
	  fma.s0     FR_Res       = FR_Res, f1, FR_GG  // result = res + gg
      br.ret.sptk   b0                     // Exit for near 1 path
};;
// NEAR ONE INTERVAL END




acoshl_lt_pone:
{ .mfi
      nop.m 0  
      fmerge.s   FR_Arg_X            = FR_Arg, FR_Arg
      nop.i 0  
};;
{ .mfb
      mov        GR_Parameter_TAG    = 135
      frcpa.s0   FR_Res,p0           = f0,f0 // get QNaN,and raise invalid
      br.cond.sptk  __libm_error_region      // exit if x < 1.0
};;

GLOBAL_LIBM_END(acoshl)



LOCAL_LIBM_ENTRY(__libm_error_region)
.prologue
{ .mfi
        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
}
{ .mfi
.fframe 64
        add      sp                  = -64,sp        // Create new stack
        nop.f 0
        mov      GR_SAVE_GP          = gp            // Save gp
};;

{ .mmi
        stfe     [GR_Parameter_Y]    = FR_Arg_Y,16   // Parameter 2 to stack
        add      GR_Parameter_X      = 16,sp         // Parameter 1 address
.save   b0,GR_SAVE_B0
        mov      GR_SAVE_B0          = b0            // Save b0
};;

.body
{ .mib
        stfe     [GR_Parameter_X]    = FR_Arg_X         // Parameter 1 to stack
        add      GR_Parameter_RESULT = 0,GR_Parameter_Y // Parameter 3 address
        nop.b 0                                 
}
{ .mib
        stfe     [GR_Parameter_Y]    = FR_Res        // Parameter 3 to stack
        add      GR_Parameter_Y      = -16,GR_Parameter_Y
        br.call.sptk b0 = __libm_error_support#      // Error handling function
};;

{ .mmi
        nop.m 0
        nop.m 0
        add      GR_Parameter_RESULT = 48,sp
};;

{ .mmi
        ldfe     f8                  = [GR_Parameter_RESULT]  // Get return res
.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#