/* Elementary functions for complex numbers German comments and names translated into English: Reini Urban 2007-11 Returns for the real numbers a and b /= Fixnum 0 the complex number a+bi. R_R_complex_C(a,b) can trigger GC */ #define R_R_complex_C make_complex /* Returns for the real numbers a and b the complex number a+bi. R_R_complex_N(a,b) can trigger GC */ local maygc object R_R_complex_N (object a, object b); /* Method: if b=0, only a. else create complex number. */ local maygc object R_R_complex_N (object a, object b) { return (eq(b,Fixnum_0) ? a : R_R_complex_C(a,b)); } /* (complex x (float 0 x)) */ global maygc object F_complex_C (object x) { #if SAFETY>=1 if (!floatp(x)) abort(); #endif pushSTACK(x); { var object zero = F_F_float_F(O(DF_zero),STACK_0); return R_R_complex_C(popSTACK(),zero); } } /* N_realpart_R(x) returns the realpart of the number x. */ local object N_realpart_R (object x); #if 0 local object N_realpart_R (object x) { return (N_realp(x) ? x : TheComplex(x)->c_real); } #else /* macro saves code */ #define N_realpart_R(x) (N_realp(x) ? x : (object)TheComplex(x)->c_real) #endif /* N_imagpart_R(x) returns the imagpart of the number x. */ local object N_imagpart_R (object x); #if 0 local object N_imagpart_R (object x) { return (N_realp(x) ? (R_floatp(x) ? RA_F_exact_contagion_R(Fixnum_0,x) : Fixnum_0) : TheComplex(x)->c_imag); } #else /* Macro saves Code */ #define N_imagpart_R(x) (N_realp(x) ? (R_floatp(x) ? RA_F_exact_contagion_R(Fixnum_0,x) : Fixnum_0) : (object)TheComplex(x)->c_imag) #endif /* N_conjugate_N(x) returns the conjugate complex number for the number x. can trigger GC */ local maygc object N_conjugate_N (object x) { if (N_realp(x)) { return x; } else { pushSTACK(TheComplex(x)->c_real); var object b = TheComplex(x)->c_imag; b = R_minus_R(b); /* flip sign at imagpart */ return R_R_complex_C(popSTACK(),b); } } /* N_minus_N(x) returns (- x), where x is a number. can trigger GC */ local maygc object N_minus_N (object x); /* Method: x real -> straightforward. x=a+bi -> (-a) + (-b) i */ local maygc object N_minus_N (object x) { if (N_realp(x)) { return R_minus_R(x); } else { pushSTACK(TheComplex(x)->c_real); var object b = R_minus_R(TheComplex(x)->c_imag); var object a = STACK_0; STACK_0 = b; a = R_minus_R(a); return R_R_complex_C(a,popSTACK()); } } /* N_N_plus_N(x) returns (+ x y), where x and y are numbers. can trigger GC */ local maygc object N_N_plus_N (object x, object y); /* Method: x,y both real -> straightforward. x=a, y=b+ci -> (a+b)+ci x=a+bi, y=c -> (a+c)+bi x=a+bi, y=c+di -> (a+c)+(b+d)i */ local maygc object N_N_plus_N (object x, object y) { if (N_realp(x)) { if (N_realp(y)) { return R_R_plus_R(x,y); } else { /* x=a, y=b+ci */ pushSTACK(TheComplex(y)->c_imag); /* c */ var object re = R_R_plus_R(x,TheComplex(y)->c_real); /* a+b */ return R_R_complex_C(re,popSTACK()); } } else { if (N_realp(y)) { /* x=a+bi, y=c */ pushSTACK(TheComplex(x)->c_imag); /* b */ var object re = R_R_plus_R(TheComplex(x)->c_real,y); /* a+c */ return R_R_complex_C(re,popSTACK()); } else { /* x=a+bi, y=c+di */ pushSTACK(TheComplex(x)->c_real); /* a */ pushSTACK(TheComplex(y)->c_real); /* c */ var object temp = /* b+d */ R_R_plus_R(TheComplex(x)->c_imag,TheComplex(y)->c_imag); y = STACK_0; STACK_0 = temp; temp = R_R_plus_R(STACK_1,y); /* a+c */ temp = R_R_complex_N(temp,STACK_0); skipSTACK(2); return temp; } } } /* N_N_minus_N(x) returns (- x y), where x and y are numbers. can trigger GC */ local maygc object N_N_minus_N (object x, object y); /* Method: x,y both real -> straightforward. x=a, y=b+ci -> (a-b)+(-c)i x=a+bi, y=c -> (a-c)+bi x=a+bi, y=c+di -> (a-c)+(b-d)i */ local maygc object N_N_minus_N (object x, object y) { if (N_realp(x)) { if (N_realp(y)) { return R_R_minus_R(x,y); } else { /* x=a, y=b+ci */ pushSTACK(TheComplex(y)->c_imag); /* c */ var object temp = R_R_minus_R(x,TheComplex(y)->c_real); /* a-b */ y = STACK_0; STACK_0 = temp; y = R_minus_R(y); /* -c */ return R_R_complex_C(popSTACK(),y); } } else { if (N_realp(y)) { /* x=a+bi, y=c */ pushSTACK(TheComplex(x)->c_imag); /* b */ var object re = R_R_minus_R(TheComplex(x)->c_real,y); /* a-c */ return R_R_complex_C(re,popSTACK()); } else { /* x=a+bi, y=c+di */ pushSTACK(TheComplex(x)->c_real); /* a */ pushSTACK(TheComplex(y)->c_real); /* c */ var object temp = /* b-d */ R_R_minus_R(TheComplex(x)->c_imag,TheComplex(y)->c_imag); y = STACK_0; STACK_0 = temp; temp = R_R_minus_R(STACK_1,y); /* a-c */ temp = R_R_complex_N(temp,STACK_0); skipSTACK(2); return temp; } } } /* N_1_plus_N(x) returns (1+ x), where x is a number. can trigger GC */ local maygc object N_1_plus_N (object x); /* Method: x real -> straightforward. x=a+bi -> (a+1)+bi */ local maygc object N_1_plus_N (object x) { if (N_realp(x)) return R_1_plus_R(x); else { pushSTACK(TheComplex(x)->c_imag); var object re = R_1_plus_R(TheComplex(x)->c_real); /* a+1 */ return R_R_complex_C(re,popSTACK()); } } /* N_minus1_plus_N(x) returns (1- x), where x is a number. can trigger GC */ local maygc object N_minus1_plus_N (object x); /* Method: x real -> straightforward. x=a+bi -> (a-1)+bi */ local maygc object N_minus1_plus_N (object x) { if (N_realp(x)) return R_minus1_plus_R(x); else { pushSTACK(TheComplex(x)->c_imag); var object re = R_minus1_plus_R(TheComplex(x)->c_real); /* a-1 */ return R_R_complex_C(re,popSTACK()); } } /* N_square_N(x) returns (* x x), where x is a number. can trigger GC */ local maygc object N_square_N (object x); /* Method: x real -> straightforward. x=a+bi -> (a^2-b^2)+(2*a*b)i */ local maygc object N_square_N (object x) { if (N_realp(x)) return R_square_R(x); else { var object a = TheComplex(x)->c_real; var object b = TheComplex(x)->c_imag; pushSTACK(a); pushSTACK(b); var object temp = R_R_mult_R(a,b); /* a*b */ temp = R_R_plus_R(temp,temp); /* 2*a*b */ a = STACK_1; STACK_1 = temp; temp = R_square_R(a); /* a*a */ b = STACK_0; STACK_0 = temp; temp = R_square_R(b); /* b*b */ temp = R_R_minus_R(popSTACK(),temp); /* a*a-b*b */ return R_R_complex_N(temp,popSTACK()); } } /* N_N_mult_N(x) returns (* x y), where x and y are numbers. can trigger GC */ local maygc object N_N_mult_N (object x, object y); /* Method: x and y both real -> straightforward. x=a, y=b+ci -> (a*b)+(a*c)i x=a+bi, y=c -> (a*c)+(b*c)i x=a+bi, y=c+di -> (a*c-b*d)+(a*d+b*c)i */ local maygc object N_N_mult_N (object x, object y) { if (N_realp(x)) { if (N_realp(y)) { return R_R_mult_R(x,y); } else { /* x=a, y=b+ci */ pushSTACK(x); /* a */ pushSTACK(TheComplex(y)->c_real); /* b */ var object temp = R_R_mult_R(x,TheComplex(y)->c_imag); /* a*c */ y = popSTACK(); x = STACK_0; STACK_0 = temp; y = R_R_mult_R(x,y); /* a*b */ return R_R_complex_N(y,popSTACK()); } } else { if (N_realp(y)) { /* x=a+bi, y=c */ pushSTACK(y); /* c */ pushSTACK(TheComplex(x)->c_real); /* a */ var object temp = R_R_mult_R(TheComplex(x)->c_imag,y); /* b*c */ x = popSTACK(); y = STACK_0; STACK_0 = temp; x = R_R_mult_R(x,y); /* a*c */ return R_R_complex_N(x,popSTACK()); } else { /* x=a+bi, y=c+di -> (a*c-b*d)+(a*d+b*c)i */ pushSTACK(TheComplex(x)->c_real); /* a */ pushSTACK(TheComplex(y)->c_real); /* c */ pushSTACK(TheComplex(x)->c_imag); /* b */ pushSTACK(y = TheComplex(y)->c_imag); /* d */ /* stack layout: a, c, b, d. */ y = R_R_mult_R(STACK_3,y); /* a*d */ x = STACK_3; STACK_3 = y; /* stack layout: a*d, c, b, d. */ x = R_R_mult_R(x,STACK_2); /* a*c */ y = STACK_2; STACK_2 = x; /* stack layout: a*d, a*c, b, d. */ y = R_R_mult_R(STACK_1,y); /* b*c */ STACK_3 = R_R_plus_R(STACK_3,y); /* a*d+b*c */ /* stack layout: a*d+b*c, a*c, b, d. */ y = R_R_mult_R(STACK_1,STACK_0); /* b*d */ y = R_R_minus_R(STACK_2,y); /* a*c-b*d */ x = STACK_3; /* a*d+b*c */ skipSTACK(4); return R_R_complex_N(y,x); } } } /* N_zerop(x) checks, if (= x 0), where x is a number. */ local bool N_zerop (object x) { if (N_realp(x)) return R_zerop(x); else /* x complex, check if both are = 0. */ return (R_zerop(TheComplex(x)->c_real) && R_zerop(TheComplex(x)->c_imag)); } /* N_div_N(x) returns (/ x), where x is a number. can trigger GC */ local maygc object N_div_N (object x); /* Method: If x real -> straightforward. If x=a+bi: If a=0: 0+(-1/b)i. If a and b are both rational: c:=a*a+b*b, c:=1/c, return a*c+(-b*c)i. If a or b are floats: If one of each is rational, round it to the same selben float type and do the UP. If both are floats, extend to the more exact, do the UP and round back to the less exact. The result is a complex number, both components are floats. UP: [a,b floats of the same type] a=0.0 -> return the components a=0.0 and -1/b. b=0.0 -> return the components 1/a and b=0.0. e:=max(exponent(a),exponent(b)). a':=a/2^e resp. 0.0 with underflow possibility (by scaling a':=a/2^e or by squaring a'*a': 2*(e-exponent(a))>exp_mid-exp_low-1 i.e. exponent(b)-exponent(a)>floor((exp_mid-exp_low-1)/2) ). b':=b/2^e resp. 0.0 with underflow possibility (by scaling b':=b/2^e or by squaring b'*b': 2*(e-exponent(b))>exp_mid-exp_low-1 i.e. exponent(a)-exponent(b)>floor((exp_mid-exp_low-1)/2) ). c':=a'*a'+b'*b', return both components 2^(-e)*a'/c' and -2^(-e)*b'/c'. */ local maygc void SFC_div_SFC (object a, object b); local maygc void FFC_div_FFC (object a, object b); local maygc void DFC_div_DFC (object a, object b); local maygc void LFC_div_LFC (object a, object b); local maygc void SFC_div_SFC (object a, object b) { var sintWL a_exp; var sintWL b_exp; { /* Get exponents of a: */ var uintBWL uexp = SF_uexp(a); if (uexp==0) { /* a=0.0 -> return (complex a (- (/ b))) : */ pushSTACK(a); pushSTACK(SF_minus_SF(SF_div_SF(b))); return; } a_exp = (sintWL)((uintWL)uexp - SF_exp_mid); } { /* Get exponents of b: */ var uintBWL uexp = SF_uexp(b); if (uexp==0) { /* b=0.0 -> return (complex (/ a) b) : */ pushSTACK(SF_div_SF(a)); pushSTACK(b); return; } b_exp = (sintWL)((uintWL)uexp - SF_exp_mid); } /* Now a_exp = exponent(a), b_exp = exponent(b). */ var sintWL e = (a_exp > b_exp ? a_exp : b_exp); /* Maximum of the exponents */ var object delta = L_to_FN(-(sintL)e); /* -e as fixnum */ /* Divide a and b by 2^e: */ a = (b_exp-a_exp > floor(SF_exp_mid-SF_exp_low-1,2) ? SF_0 : SF_I_scale_float_SF(a,delta)); b = (a_exp-b_exp > floor(SF_exp_mid-SF_exp_low-1,2) ? SF_0 : SF_I_scale_float_SF(b,delta)); /* Calc c' := a'*a'+b'*b': */ var object c = SF_SF_plus_SF(SF_square_SF(a),SF_square_SF(b)); a = SF_I_scale_float_SF(SF_SF_div_SF(a,c),delta); /* 2^(-e)*a'/c' */ b = SF_I_scale_float_SF(SF_minus_SF(SF_SF_div_SF(b,c)),delta); /* -2^(-e)*b'/c' */ pushSTACK(a); pushSTACK(b); return; } local maygc void FFC_div_FFC (object a, object b) { var sintWL a_exp; var sintWL b_exp; { /* Get exponents of a: */ var uintBWL uexp = FF_uexp(ffloat_value(a)); if (uexp==0) { /* a=0.0 -> return (complex a (- (/ b))) : */ pushSTACK(a); pushSTACK(FF_minus_FF(FF_div_FF(b))); return; } a_exp = (sintWL)((uintWL)uexp - FF_exp_mid); } { /* Get exponents of a: */ var uintBWL uexp = FF_uexp(ffloat_value(b)); if (uexp==0) { /* b=0.0 -> return (complex (/ a) b) : */ pushSTACK(FF_div_FF(a)); pushSTACK(FF_0); return; } b_exp = (sintWL)((uintWL)uexp - FF_exp_mid); } pushSTACK(a); pushSTACK(b); /* stack layout: a, b. */ /* Now a_exp = exponent(a), b_exp = exponent(b). */ var sintWL e = (a_exp > b_exp ? a_exp : b_exp); /* Maximum of the exponents */ var object delta = L_to_FN(-(sintL)e); /* -e as fixnum */ /* Divide a and b by 2^e: */ STACK_1 = (b_exp-a_exp > floor(FF_exp_mid-FF_exp_low-1,2) ? FF_0 : FF_I_scale_float_FF(STACK_1,delta)); STACK_0 = (a_exp-b_exp > floor(FF_exp_mid-FF_exp_low-1,2) ? FF_0 : FF_I_scale_float_FF(STACK_0,delta)); /* stack layout: a', b'. */ /* Calc c' := a'*a'+b'*b': */ var object temp; temp = STACK_1; pushSTACK(FF_square_FF(temp)); /* a'*a' */ temp = STACK_1; temp = FF_square_FF(temp); /* b'*b' */ STACK_0 = temp = FF_FF_plus_FF(STACK_0,temp); /* c' = a'*a'+b'*b' */ STACK_2 = FF_I_scale_float_FF(FF_FF_div_FF(STACK_2,temp),delta); /* 2^(-e)*a'/c' */ STACK_1 = FF_I_scale_float_FF(FF_minus_FF(FF_FF_div_FF(STACK_1,STACK_0)),delta); /* -2^(-e)*b'/c' */ skipSTACK(1); return; } local maygc void DFC_div_DFC (object a, object b) { var sintWL a_exp; var sintWL b_exp; { /* get exponents of a: */ var uintWL uexp = DF_uexp(TheDfloat(a)->float_value_semhi); if (uexp==0) { /* a=0.0 -> return (complex a (- (/ b))) : */ pushSTACK(a); pushSTACK(DF_minus_DF(DF_div_DF(b))); return; } a_exp = (sintWL)(uexp - DF_exp_mid); } { /* get exponents of b: */ var uintWL uexp = DF_uexp(TheDfloat(b)->float_value_semhi); if (uexp==0) { /* b=0.0 -> return (complex (/ a) b) : */ pushSTACK(DF_div_DF(a)); pushSTACK(DF_0); return; } b_exp = (sintWL)(uexp - DF_exp_mid); } pushSTACK(a); pushSTACK(b); /* stack layout: a, b. */ /* Now a_exp = exponent(a), b_exp = exponent(b). */ var sintWL e = (a_exp > b_exp ? a_exp : b_exp); /* Maximum of the exponents */ var object delta = L_to_FN(-(sintL)e); /* -e as Fixnum */ /* Divide a and b by 2^e: */ STACK_1 = (b_exp-a_exp > floor(DF_exp_mid-DF_exp_low-1,2) ? DF_0 : DF_I_scale_float_DF(STACK_1,delta)); STACK_0 = (a_exp-b_exp > floor(DF_exp_mid-DF_exp_low-1,2) ? DF_0 : DF_I_scale_float_DF(STACK_0,delta)); /* stack layout: a', b'. */ /* calc c' := a'*a'+b'*b': */ var object temp; temp = STACK_1; pushSTACK(DF_square_DF(temp)); /* a'*a' */ temp = STACK_1; temp = DF_square_DF(temp); /* b'*b' */ STACK_0 = temp = DF_DF_plus_DF(STACK_0,temp); /* c' = a'*a'+b'*b' */ STACK_2 = DF_I_scale_float_DF(DF_DF_div_DF(STACK_2,temp),delta); /* 2^(-e)*a'/c' */ STACK_1 = DF_I_scale_float_DF(DF_minus_DF(DF_DF_div_DF(STACK_1,STACK_0)),delta); /* -2^(-e)*b'/c' */ skipSTACK(1); return; } local maygc void LFC_div_LFC (object a, object b) { var uintL a_exp; var uintL b_exp; { /* get exponents of a: */ a_exp = TheLfloat(a)->expo; if (a_exp==0) { /* a=0.0 -> return (complex a (- (/ b))) : */ pushSTACK(a); pushSTACK(LF_minus_LF(LF_div_LF(b))); return; } } { /* get exponents of b: */ b_exp = TheLfloat(b)->expo; if (b_exp==0) { /* b=0.0 -> return (complex (/ a) b) : */ pushSTACK(b); pushSTACK(b); STACK_1 = LF_div_LF(a); return; } } pushSTACK(a); pushSTACK(b); /* stack layout: a, b. */ /* Now a_exp = exponent(a)+LF_exp_mid, b_exp = exponent(b)+LF_exp_mid. */ { var uintL e = (a_exp > b_exp ? a_exp : b_exp); /* Maximum of the exponents */ pushSTACK(L_to_I(-(sintL)(e-LF_exp_mid))); /* -e as integer */ } /* Divide a and b by 2^e: */ if ((b_exp>a_exp) && (b_exp-a_exp > (uintL)floor((uintL)(LF_exp_mid-LF_exp_low-1),2))) { encode_LF0(Lfloat_length(STACK_2), STACK_2=); } else { STACK_2 = LF_I_scale_float_LF(STACK_2,STACK_0); } if ((a_exp>b_exp) && (a_exp-b_exp > (uintL)floor((uintL)(LF_exp_mid-LF_exp_low-1),2))) { encode_LF0(Lfloat_length(STACK_1), STACK_1=); } else { STACK_1 = LF_I_scale_float_LF(STACK_1,STACK_0); } /* stack layout: a', b', -e. */ /* Calc c' := a'*a'+b'*b': */ var object temp; temp = LF_square_LF(STACK_2); pushSTACK(temp); /* a'*a' */ temp = LF_square_LF(STACK_2); /* b'*b' */ STACK_0 = temp = LF_LF_plus_LF(STACK_0,temp); /* c' = a'*a'+b'*b' */ temp = LF_LF_div_LF(STACK_3,temp); STACK_3 = LF_I_scale_float_LF(temp,STACK_1); /* 2^(-e)*a'/c' */ temp = LF_minus_LF(LF_LF_div_LF(STACK_2,STACK_0)); STACK_2 = LF_I_scale_float_LF(temp,STACK_1); /* -2^(-e)*b'/c' */ skipSTACK(2); return; } local maygc object N_div_N (object x) { if (N_realp(x)) return R_div_R(x); else { var object a = TheComplex(x)->c_real; var object b = TheComplex(x)->c_imag; /* x = a+bi */ if (R_rationalp(a)) { if (eq(a,Fixnum_0)) return R_R_complex_C(Fixnum_0,R_minus_R(R_div_R(b))); /* (complex 0 (- (/ b))) */ if (R_rationalp(b)) { /* a,b both rational */ var object temp; pushSTACK(a); pushSTACK(b); temp = RA_square_RA(a); pushSTACK(temp); /* a*a */ temp = RA_square_RA(STACK_1); /* b*b */ temp = RA_RA_plus_RA(STACK_0,temp); /* a*a+b*b = c */ STACK_0 = temp = RA_div_RA(temp); /* c:=1/c */ STACK_2 = RA_RA_mult_RA(STACK_2,temp); /* a*c */ STACK_1 = RA_minus_RA(RA_RA_mult_RA(STACK_1,STACK_0)); /* -b*c */ temp = R_R_complex_C(STACK_2,STACK_1); /* (a*c) + (-b*c) i */ skipSTACK(3); return temp; } else { /* a rational, b float */ pushSTACK(b); floatcase(b, { a = RA_to_SF(a,true); SFC_div_SFC(a,popSTACK()); }, { a = RA_to_FF(a,true); FFC_div_FFC(a,popSTACK()); }, { a = RA_to_DF(a,true); DFC_div_DFC(a,popSTACK()); }, { a = RA_to_LF(a,Lfloat_length(b),true); LFC_div_LFC(a,popSTACK()); } ); } } else { if (R_rationalp(b)) { /* a float, b rational */ pushSTACK(a); floatcase(a, { b = RA_to_SF(b,true); SFC_div_SFC(popSTACK(),b); }, { b = RA_to_FF(b,true); FFC_div_FFC(popSTACK(),b); }, { b = RA_to_DF(b,true); DFC_div_DFC(popSTACK(),b); }, { b = RA_to_LF(b,Lfloat_length(a),true); LFC_div_LFC(popSTACK(),b); } ); } else { /* a,b floats */ GEN_F_op2(a,b,SFC_div_SFC,FFC_div_FFC,DFC_div_DFC,LFC_div_LFC,2,0,) } } /* merge both components to a complex number: */ a = STACK_1; b = STACK_0; skipSTACK(2); return R_R_complex_C(a,b); } } #define C_div_C N_div_N /* N_N_div_N(x,y) returns (/ x y), where x and y are numbers. can trigger GC */ local maygc object N_N_div_N (object x, object y); /* Method: x,y both real -> straightforward. x=a+bi, y=c real -> (a/c)+(b/c)i y complex -> (* x (/ y)) */ local maygc object N_N_div_N (object x, object y) { if (N_realp(y)) { /* y real */ if (N_realp(x)) { /* x,y both real */ return R_R_div_R(x,y); } else { /* x complex: x=a+bi, y=c */ pushSTACK(y); /* c */ pushSTACK(TheComplex(x)->c_real); /* a */ var object temp = R_R_div_R(TheComplex(x)->c_imag,y); /* b/c */ x = popSTACK(); y = STACK_0; STACK_0 = temp; x = R_R_div_R(x,y); /* a/c */ return R_R_complex_N(x,popSTACK()); } } else { /* y complex */ pushSTACK(x); y = C_div_C(y); /* with (/ y) */ return N_N_mult_N(popSTACK(),y); /* multiply */ } } /* R_R_hypot_R(a,b) returns sqrt(a^2+b^2), where a and b are real numbers. can trigger GC */ local maygc object R_R_hypot_R (object a, object b); /* Method: If a=0: (abs b). If b=0: (abs a). If a and b are both rational: c:=a*a+b*b, return (sqrt c). If a or b are floats: If one of both is rational, round it to the same float type as the other and do the UP. If both are floats, expand to the more precise, do the UP and round down to the imprecise. The result is a float >=0. UP: [a,b floats of the same type] a=0.0 -> return abs(b). b=0.0 -> return abs(a). e:=max(exponent(a),exponent(b)). a':=a/2^e resp. 0.0 with underflow possibility (by scaling a':=a/2^e or by squaring a'*a': 2*(e-exponent(a))>exp_mid-exp_low-1 i.e. exponent(b)-exponent(a)>floor((exp_mid-exp_low-1)/2) ). b':=b/2^e resp 0.0 with underflow possibility (by scaling b':=b/2^e or by squaring b'*b': 2*(e-exponent(b))>exp_mid-exp_low-1 i.e. exponent(a)-exponent(b)>floor((exp_mid-exp_low-1)/2) ). c':=a'*a'+b'*b', c':=sqrt(c'), return 2^e*c'. */ local maygc object SF_SF_hypot_SF (object a, object b); local maygc object FF_FF_hypot_FF (object a, object b); local maygc object DF_DF_hypot_DF (object a, object b); local maygc object LF_LF_hypot_LF (object a, object b); local maygc object SF_SF_hypot_SF (object a, object b) { var sintWL a_exp; var sintWL b_exp; { /* get exponents of a: */ var uintBWL uexp = SF_uexp(a); if (uexp==0) /* a=0.0 -> return (abs b) : */ return (R_minusp(b) ? SF_minus_SF(b) : b); a_exp = (sintWL)((uintWL)uexp - SF_exp_mid); } { /* get exponents of b: */ var uintBWL uexp = SF_uexp(b); if (uexp==0) /* b=0.0 -> return (abs a) : */ return (R_minusp(a) ? SF_minus_SF(a) : a); b_exp = (sintWL)((uintWL)uexp - SF_exp_mid); } /* Now a_exp = exponent(a), b_exp = exponent(b). */ var sintWL e = (a_exp > b_exp ? a_exp : b_exp); /* Maximum of the exponents */ var object delta = L_to_FN(-(sintL)e); /* -e as fixnum */ /* Divide a and b by 2^e: */ a = (b_exp-a_exp > floor(SF_exp_mid-SF_exp_low-1,2) ? SF_0 : SF_I_scale_float_SF(a,delta)); b = (a_exp-b_exp > floor(SF_exp_mid-SF_exp_low-1,2) ? SF_0 : SF_I_scale_float_SF(b,delta)); /* c' := a'*a'+b'*b': */ var object c = SF_SF_plus_SF(SF_square_SF(a),SF_square_SF(b)); c = SF_sqrt_SF(c); /* c':=2^e*c' */ return SF_I_scale_float_SF(c,L_to_FN((sintL)e)); /* 2^e*c' */ } local maygc object FF_FF_hypot_FF (object a, object b) { var sintWL a_exp; var sintWL b_exp; { /* get exponents of a: */ var uintBWL uexp = FF_uexp(ffloat_value(a)); if (uexp==0) /* a=0.0 -> return (abs b) : */ return (R_minusp(b) ? FF_minus_FF(b) : b); a_exp = (sintWL)((uintWL)uexp - FF_exp_mid); } { /* get exponents of b: */ var uintBWL uexp = FF_uexp(ffloat_value(b)); if (uexp==0) /* b=0.0 -> return (abs a) : */ return (R_minusp(a) ? FF_minus_FF(a) : a); b_exp = (sintWL)((uintWL)uexp - FF_exp_mid); } pushSTACK(a); pushSTACK(b); /* stack layout: a, b. */ /* Now a_exp = exponent(a), b_exp = exponent(b). */ var sintWL e = (a_exp > b_exp ? a_exp : b_exp); /* Maximum of the exponents */ var object delta = L_to_FN(-(sintL)e); /* -e as fixnum */ /* Divide a and b by 2^e: */ STACK_1 = (b_exp-a_exp > floor(FF_exp_mid-FF_exp_low-1,2) ? FF_0 : FF_I_scale_float_FF(STACK_1,delta)); STACK_0 = (a_exp-b_exp > floor(FF_exp_mid-FF_exp_low-1,2) ? FF_0 : FF_I_scale_float_FF(STACK_0,delta)); /* stack layout: a', b'. */ /* c' := a'*a'+b'*b': */ var object temp; temp = STACK_1; pushSTACK(FF_square_FF(temp)); /* a'*a' */ temp = STACK_1; temp = FF_square_FF(temp); /* b'*b' */ temp = FF_FF_plus_FF(STACK_0,temp); /* c' = a'*a'+b'*b' */ skipSTACK(3); temp = FF_sqrt_FF(temp); /* c':=2^e*c' */ return FF_I_scale_float_FF(temp,L_to_FN((sintL)e)); /* 2^e*c' */ } local maygc object DF_DF_hypot_DF (object a, object b) { var sintWL a_exp; var sintWL b_exp; { /* get exponents of a: */ var uintWL uexp = DF_uexp(TheDfloat(a)->float_value_semhi); if (uexp==0) /* a=0.0 -> return (abs b) : */ return (R_minusp(b) ? DF_minus_DF(b) : b); a_exp = (sintWL)(uexp - DF_exp_mid); } { /* get exponents of b: */ var uintWL uexp = DF_uexp(TheDfloat(b)->float_value_semhi); if (uexp==0) /* b=0.0 -> return (abs a) : */ return (R_minusp(a) ? DF_minus_DF(a) : a); b_exp = (sintWL)(uexp - DF_exp_mid); } pushSTACK(a); pushSTACK(b); /* stack layout: a, b. */ /* Now a_exp = exponent(a), b_exp = exponent(b). */ var sintWL e = (a_exp > b_exp ? a_exp : b_exp); /* Maximum of the exponents */ var object delta = L_to_FN(-(sintL)e); /* -e as fixnum */ /* Divide a and b by 2^e: */ STACK_1 = (b_exp-a_exp > floor(DF_exp_mid-DF_exp_low-1,2) ? DF_0 : DF_I_scale_float_DF(STACK_1,delta)); STACK_0 = (a_exp-b_exp > floor(DF_exp_mid-DF_exp_low-1,2) ? DF_0 : DF_I_scale_float_DF(STACK_0,delta)); /* stack layout: a', b'. */ /* c' := a'*a'+b'*b': */ var object temp; temp = STACK_1; pushSTACK(DF_square_DF(temp)); /* a'*a' */ temp = STACK_1; temp = DF_square_DF(temp); /* b'*b' */ temp = DF_DF_plus_DF(STACK_0,temp); /* c' = a'*a'+b'*b' */ skipSTACK(3); temp = DF_sqrt_DF(temp); /* c':=2^e*c' */ return DF_I_scale_float_DF(temp,L_to_FN((sintL)e)); /* 2^e*c' */ } local maygc object LF_LF_hypot_LF (object a, object b) { var uintL a_exp; var uintL b_exp; { /* get exponents of a: */ a_exp = TheLfloat(a)->expo; if (a_exp==0) /* a=0.0 -> return (abs b) : */ return (R_minusp(b) ? LF_minus_LF(b) : b); } { /* get exponents of b: */ b_exp = TheLfloat(b)->expo; if (b_exp==0) /* b=0.0 -> return (abs a) : */ return (R_minusp(a) ? LF_minus_LF(a) : a); } pushSTACK(a); pushSTACK(b); /* stack layout: a, b. */ /* a_exp = exponent(a)+LF_exp_mid, b_exp = exponent(b)+LF_exp_mid. */ { var uintL exp = (a_exp > b_exp ? a_exp : b_exp); /* Maximum of the exponents */ var sintL e = (sintL)(exp-LF_exp_mid); pushSTACK(L_to_I(e)); /* e as integer */ pushSTACK(L_to_I(-e)); /* -e as integer */ } /* Divide a and b by 2^e: */ if ((b_exp>a_exp) && (b_exp-a_exp > (uintL)floor((uintL)(LF_exp_mid-LF_exp_low-1),2))) { encode_LF0(Lfloat_length(STACK_3), STACK_3=); } else { STACK_3 = LF_I_scale_float_LF(STACK_3,STACK_0); } if ((a_exp>b_exp) && (a_exp-b_exp > (uintL)floor((uintL)(LF_exp_mid-LF_exp_low-1),2))) { encode_LF0(Lfloat_length(STACK_2), STACK_2=); } else { STACK_2 = LF_I_scale_float_LF(STACK_2,STACK_0); } /* stack layout: a', b', e, -e. */ /* c' := a'*a'+b'*b': */ var object temp; temp = LF_square_LF(STACK_3); pushSTACK(temp); /* a'*a' */ temp = LF_square_LF(STACK_3); /* b'*b' */ temp = LF_LF_plus_LF(STACK_0,temp); /* c' = a'*a'+b'*b' */ temp = LF_sqrt_LF(temp); /* c':=2^e*c' */ temp = LF_I_scale_float_LF(temp,STACK_2); /* 2^e*c' */ skipSTACK(5); return temp; } local maygc object R_R_hypot_R (object a, object b) { if (R_rationalp(a)) { if (eq(a,Fixnum_0)) /* a=0 -> (abs b) */ return R_abs_R(b); if (R_rationalp(b)) { /* a,b both rational */ var object temp; if (eq(b,Fixnum_0)) /* b=0 -> (abs a) */ return R_abs_R(a); pushSTACK(a); pushSTACK(b); temp = RA_square_RA(a); pushSTACK(temp); /* a*a */ temp = RA_square_RA(STACK_1); /* b*b */ temp = RA_RA_plus_RA(STACK_0,temp); /* a*a+b*b = c */ skipSTACK(3); return RA_sqrt_R(temp); /* (sqrt c) */ } else { /* a rational, b float */ pushSTACK(b); floatcase(b, { a = RA_to_SF(a,true); return SF_SF_hypot_SF(a,popSTACK()); }, { a = RA_to_FF(a,true); return FF_FF_hypot_FF(a,popSTACK()); }, { a = RA_to_DF(a,true); return DF_DF_hypot_DF(a,popSTACK()); }, { a = RA_to_LF(a,Lfloat_length(b),true); return LF_LF_hypot_LF(a,popSTACK()); } ); } } else { if (R_rationalp(b)) { /* a float, b rational */ if (eq(b,Fixnum_0)) /* b=0 -> (abs a) */ return R_abs_R(a); pushSTACK(a); floatcase(a, { b = RA_to_SF(b,true); return SF_SF_hypot_SF(popSTACK(),b); }, { b = RA_to_FF(b,true); return FF_FF_hypot_FF(popSTACK(),b); }, { b = RA_to_DF(b,true); return DF_DF_hypot_DF(popSTACK(),b); }, { b = RA_to_LF(b,Lfloat_length(a),true); return LF_LF_hypot_LF(popSTACK(),b); } ); } else { /* a,b Floats */ GEN_F_op2(a,b,SF_SF_hypot_SF,FF_FF_hypot_FF,DF_DF_hypot_DF,LF_LF_hypot_LF,1,0,return) } } } /* N_abs_R(x) returns (abs x), where x is a number. can trigger GC */ local maygc object N_abs_R (object x); /* Method: If x real -> straightforward. If x=a+bi: sqrt(a^2+b^2) */ local maygc object N_abs_R (object x) { if (N_realp(x)) return R_abs_R(x); else return R_R_hypot_R(TheComplex(x)->c_real,TheComplex(x)->c_imag); } #define C_abs_R(x) R_R_hypot_R(TheComplex(x)->c_real,TheComplex(x)->c_imag) /* N_signum_N(x) returns (signum x), where x is a number. can trigger GC */ local maygc object N_signum_N (object x); /* Method: x real -> straightforward. x complex -> if (zerop x), x as result, otherwise (/ x (abs x)). */ local maygc object N_signum_N (object x) { if (N_realp(x)) return R_signum_R(x); else { if (N_zerop(x)) return x; pushSTACK(x); x = C_abs_R(x); /* calculate (abs x) */ return N_N_div_N(popSTACK(),x); /* (/ x (abs x)) */ } } /* N_sqrt_N(x) returns (sqrt x), where x is any number. can trigger GC */ local maygc object N_sqrt_N (object x); /* Method: x real -> for x>=0 simple, for x<0: sqrt(-x)*i. x=a+bi -> Set r=abs(x)=sqrt(a*a+b*b). if a>=0: Set c:=sqrt((r+a)/2), d:=(b/(2*c) if c>0, c falls c=0). if a<0: Set d:=sqrt((r-a)/2)*(1 if b>=0, -1 if b<0), c:=b/(2*d). so we get for c>=0, 2*c*d=b, c*c=(r+a)/2, d*d=(r-a)/2, c*c-d*d=a, c*c+d*d=r, so c+di is the required square root */ local maygc object N_sqrt_N (object x) { if (N_realp(x)) { /* x real */ if (!R_minusp(x)) return R_sqrt_R(x); else return R_R_complex_C(Fixnum_0,R_sqrt_R(R_minus_R(x))); } else { /* x complex */ var object a = TheComplex(x)->c_real; var object b = TheComplex(x)->c_imag; pushSTACK(b); pushSTACK(a); var object r = R_R_hypot_R(a,b); /* r = (abs x) */ a = STACK_0; if (!R_minusp(a)) { /* a>=0 */ var object c = /* sqrt((r+a)/2) */ R_sqrt_R(R_R_div_R(R_R_plus_R(r,a),fixnum(2))); STACK_0 = c; if (!R_zerop(c)) { c = R_R_mult_R(fixnum(2),c); /* 2*c */ c = R_R_div_R(STACK_1,c); /* d:=b/(2*c) */ } c = R_R_complex_C(STACK_0,c); /* c+di */ skipSTACK(2); return c; /* as result */ } else { /* a<0 */ var object d = /* sqrt((r-a)/2) */ R_sqrt_R(R_R_div_R(R_R_minus_R(r,a),fixnum(2))); if (R_minusp(STACK_1)) /* if b<0 flip sign */ d = R_minus_R(d); STACK_0 = d; d = R_R_mult_R(fixnum(2),d); /* 2*d */ d = R_R_div_R(STACK_1,d); /* c:=b/(2*d) */ d = R_R_complex_C(d,STACK_0); /* c+di */ skipSTACK(2); return d; /* as result */ } } } /* N_N_equal(x,y) compares two real numbers x and y. Returns: true if x=y, false otherwise. */ global /* local */ bool N_N_equal (object x, object y); /* Method: both real => R_R_equal x real, y complex => (= x (realpart y)) && (zerop (imagpart y)) x complex, y real => (= y (realpart x)) && (zerop (imagpart x)) both complex: realpart & imagpart both equal */ global /* local */ bool N_N_equal (object x, object y) { return N_realp(x) ? (N_realp(y) ? /* x,y both real */ R_R_equal(x,y) : /* x real, y complex */ (R_zerop(TheComplex(y)->c_imag) ? R_R_equal(x,TheComplex(y)->c_real) : false)) : (N_realp(y) ? /* x complex, y real */ (R_zerop(TheComplex(x)->c_imag) ? R_R_equal(TheComplex(x)->c_real,y) : false) : /* x,y both complex */ (R_R_equal(TheComplex(x)->c_real,TheComplex(y)->c_real) ? R_R_equal(TheComplex(x)->c_imag,TheComplex(y)->c_imag) : false)); } /* ffloat_N_equal(x,y) compares a single-float x with a number y, Returns: true if x=y, false otherwise. */ local bool ffloat_N_equal (ffloat x, object y); /* Method: y real => ffloat_R_equal y complex => (= x (realpart y)) && (zerop (imagpart y)) */ local bool ffloat_N_equal (ffloat x, object y) { return N_realp(y) ? /* y real */ ffloat_R_equal(x,y) : /* y complex */ (R_zerop(TheComplex(y)->c_imag) ? ffloat_R_equal(x,TheComplex(y)->c_real) : false); } /* dfloat_N_equal(x,y) compares a double-float x with a number y, Returns: true if x=y, false otherwise. */ local bool dfloat_N_equal (dfloat x, object y); /* Method: y real => dfloat_R_equal y complex => (= x (realpart y)) && (zerop (imagpart y)) */ local bool dfloat_N_equal (dfloat x, object y) { return N_realp(y) ? /* y real */ dfloat_R_equal(x,y) : /* y complex */ (R_zerop(TheComplex(y)->c_imag) ? dfloat_R_equal(x,TheComplex(y)->c_real) : false); }