diff options
Diffstat (limited to 'arch/x86/math-emu/poly_2xm1.c')
-rw-r--r-- | arch/x86/math-emu/poly_2xm1.c | 199 |
1 files changed, 94 insertions, 105 deletions
diff --git a/arch/x86/math-emu/poly_2xm1.c b/arch/x86/math-emu/poly_2xm1.c index 9766ad5e9743..d8f2be3c8383 100644 --- a/arch/x86/math-emu/poly_2xm1.c +++ b/arch/x86/math-emu/poly_2xm1.c | |||
@@ -17,21 +17,19 @@ | |||
17 | #include "control_w.h" | 17 | #include "control_w.h" |
18 | #include "poly.h" | 18 | #include "poly.h" |
19 | 19 | ||
20 | |||
21 | #define HIPOWER 11 | 20 | #define HIPOWER 11 |
22 | static const unsigned long long lterms[HIPOWER] = | 21 | static const unsigned long long lterms[HIPOWER] = { |
23 | { | 22 | 0x0000000000000000LL, /* This term done separately as 12 bytes */ |
24 | 0x0000000000000000LL, /* This term done separately as 12 bytes */ | 23 | 0xf5fdeffc162c7543LL, |
25 | 0xf5fdeffc162c7543LL, | 24 | 0x1c6b08d704a0bfa6LL, |
26 | 0x1c6b08d704a0bfa6LL, | 25 | 0x0276556df749cc21LL, |
27 | 0x0276556df749cc21LL, | 26 | 0x002bb0ffcf14f6b8LL, |
28 | 0x002bb0ffcf14f6b8LL, | 27 | 0x0002861225ef751cLL, |
29 | 0x0002861225ef751cLL, | 28 | 0x00001ffcbfcd5422LL, |
30 | 0x00001ffcbfcd5422LL, | 29 | 0x00000162c005d5f1LL, |
31 | 0x00000162c005d5f1LL, | 30 | 0x0000000da96ccb1bLL, |
32 | 0x0000000da96ccb1bLL, | 31 | 0x0000000078d1b897LL, |
33 | 0x0000000078d1b897LL, | 32 | 0x000000000422b029LL |
34 | 0x000000000422b029LL | ||
35 | }; | 33 | }; |
36 | 34 | ||
37 | static const Xsig hiterm = MK_XSIG(0xb17217f7, 0xd1cf79ab, 0xc8a39194); | 35 | static const Xsig hiterm = MK_XSIG(0xb17217f7, 0xd1cf79ab, 0xc8a39194); |
@@ -45,112 +43,103 @@ static const Xsig shiftterm2 = MK_XSIG(0xb504f333, 0xf9de6484, 0x597d89b3); | |||
45 | static const Xsig shiftterm3 = MK_XSIG(0xd744fcca, 0xd69d6af4, 0x39a68bb9); | 43 | static const Xsig shiftterm3 = MK_XSIG(0xd744fcca, 0xd69d6af4, 0x39a68bb9); |
46 | 44 | ||
47 | static const Xsig *shiftterm[] = { &shiftterm0, &shiftterm1, | 45 | static const Xsig *shiftterm[] = { &shiftterm0, &shiftterm1, |
48 | &shiftterm2, &shiftterm3 }; | 46 | &shiftterm2, &shiftterm3 |
49 | 47 | }; | |
50 | 48 | ||
51 | /*--- poly_2xm1() -----------------------------------------------------------+ | 49 | /*--- poly_2xm1() -----------------------------------------------------------+ |
52 | | Requires st(0) which is TAG_Valid and < 1. | | 50 | | Requires st(0) which is TAG_Valid and < 1. | |
53 | +---------------------------------------------------------------------------*/ | 51 | +---------------------------------------------------------------------------*/ |
54 | int poly_2xm1(u_char sign, FPU_REG *arg, FPU_REG *result) | 52 | int poly_2xm1(u_char sign, FPU_REG * arg, FPU_REG * result) |
55 | { | 53 | { |
56 | long int exponent, shift; | 54 | long int exponent, shift; |
57 | unsigned long long Xll; | 55 | unsigned long long Xll; |
58 | Xsig accumulator, Denom, argSignif; | 56 | Xsig accumulator, Denom, argSignif; |
59 | u_char tag; | 57 | u_char tag; |
60 | 58 | ||
61 | exponent = exponent16(arg); | 59 | exponent = exponent16(arg); |
62 | 60 | ||
63 | #ifdef PARANOID | 61 | #ifdef PARANOID |
64 | if ( exponent >= 0 ) /* Don't want a |number| >= 1.0 */ | 62 | if (exponent >= 0) { /* Don't want a |number| >= 1.0 */ |
65 | { | 63 | /* Number negative, too large, or not Valid. */ |
66 | /* Number negative, too large, or not Valid. */ | 64 | EXCEPTION(EX_INTERNAL | 0x127); |
67 | EXCEPTION(EX_INTERNAL|0x127); | 65 | return 1; |
68 | return 1; | 66 | } |
69 | } | ||
70 | #endif /* PARANOID */ | 67 | #endif /* PARANOID */ |
71 | 68 | ||
72 | argSignif.lsw = 0; | 69 | argSignif.lsw = 0; |
73 | XSIG_LL(argSignif) = Xll = significand(arg); | 70 | XSIG_LL(argSignif) = Xll = significand(arg); |
74 | 71 | ||
75 | if ( exponent == -1 ) | 72 | if (exponent == -1) { |
76 | { | 73 | shift = (argSignif.msw & 0x40000000) ? 3 : 2; |
77 | shift = (argSignif.msw & 0x40000000) ? 3 : 2; | 74 | /* subtract 0.5 or 0.75 */ |
78 | /* subtract 0.5 or 0.75 */ | 75 | exponent -= 2; |
79 | exponent -= 2; | 76 | XSIG_LL(argSignif) <<= 2; |
80 | XSIG_LL(argSignif) <<= 2; | 77 | Xll <<= 2; |
81 | Xll <<= 2; | 78 | } else if (exponent == -2) { |
82 | } | 79 | shift = 1; |
83 | else if ( exponent == -2 ) | 80 | /* subtract 0.25 */ |
84 | { | 81 | exponent--; |
85 | shift = 1; | 82 | XSIG_LL(argSignif) <<= 1; |
86 | /* subtract 0.25 */ | 83 | Xll <<= 1; |
87 | exponent--; | 84 | } else |
88 | XSIG_LL(argSignif) <<= 1; | 85 | shift = 0; |
89 | Xll <<= 1; | 86 | |
90 | } | 87 | if (exponent < -2) { |
91 | else | 88 | /* Shift the argument right by the required places. */ |
92 | shift = 0; | 89 | if (FPU_shrx(&Xll, -2 - exponent) >= 0x80000000U) |
93 | 90 | Xll++; /* round up */ | |
94 | if ( exponent < -2 ) | 91 | } |
95 | { | 92 | |
96 | /* Shift the argument right by the required places. */ | 93 | accumulator.lsw = accumulator.midw = accumulator.msw = 0; |
97 | if ( FPU_shrx(&Xll, -2-exponent) >= 0x80000000U ) | 94 | polynomial_Xsig(&accumulator, &Xll, lterms, HIPOWER - 1); |
98 | Xll++; /* round up */ | 95 | mul_Xsig_Xsig(&accumulator, &argSignif); |
99 | } | 96 | shr_Xsig(&accumulator, 3); |
100 | 97 | ||
101 | accumulator.lsw = accumulator.midw = accumulator.msw = 0; | 98 | mul_Xsig_Xsig(&argSignif, &hiterm); /* The leading term */ |
102 | polynomial_Xsig(&accumulator, &Xll, lterms, HIPOWER-1); | 99 | add_two_Xsig(&accumulator, &argSignif, &exponent); |
103 | mul_Xsig_Xsig(&accumulator, &argSignif); | 100 | |
104 | shr_Xsig(&accumulator, 3); | 101 | if (shift) { |
105 | 102 | /* The argument is large, use the identity: | |
106 | mul_Xsig_Xsig(&argSignif, &hiterm); /* The leading term */ | 103 | f(x+a) = f(a) * (f(x) + 1) - 1; |
107 | add_two_Xsig(&accumulator, &argSignif, &exponent); | 104 | */ |
108 | 105 | shr_Xsig(&accumulator, -exponent); | |
109 | if ( shift ) | 106 | accumulator.msw |= 0x80000000; /* add 1.0 */ |
110 | { | 107 | mul_Xsig_Xsig(&accumulator, shiftterm[shift]); |
111 | /* The argument is large, use the identity: | 108 | accumulator.msw &= 0x3fffffff; /* subtract 1.0 */ |
112 | f(x+a) = f(a) * (f(x) + 1) - 1; | 109 | exponent = 1; |
113 | */ | 110 | } |
114 | shr_Xsig(&accumulator, - exponent); | 111 | |
115 | accumulator.msw |= 0x80000000; /* add 1.0 */ | 112 | if (sign != SIGN_POS) { |
116 | mul_Xsig_Xsig(&accumulator, shiftterm[shift]); | 113 | /* The argument is negative, use the identity: |
117 | accumulator.msw &= 0x3fffffff; /* subtract 1.0 */ | 114 | f(-x) = -f(x) / (1 + f(x)) |
118 | exponent = 1; | 115 | */ |
119 | } | 116 | Denom.lsw = accumulator.lsw; |
120 | 117 | XSIG_LL(Denom) = XSIG_LL(accumulator); | |
121 | if ( sign != SIGN_POS ) | 118 | if (exponent < 0) |
122 | { | 119 | shr_Xsig(&Denom, -exponent); |
123 | /* The argument is negative, use the identity: | 120 | else if (exponent > 0) { |
124 | f(-x) = -f(x) / (1 + f(x)) | 121 | /* exponent must be 1 here */ |
125 | */ | 122 | XSIG_LL(Denom) <<= 1; |
126 | Denom.lsw = accumulator.lsw; | 123 | if (Denom.lsw & 0x80000000) |
127 | XSIG_LL(Denom) = XSIG_LL(accumulator); | 124 | XSIG_LL(Denom) |= 1; |
128 | if ( exponent < 0 ) | 125 | (Denom.lsw) <<= 1; |
129 | shr_Xsig(&Denom, - exponent); | 126 | } |
130 | else if ( exponent > 0 ) | 127 | Denom.msw |= 0x80000000; /* add 1.0 */ |
131 | { | 128 | div_Xsig(&accumulator, &Denom, &accumulator); |
132 | /* exponent must be 1 here */ | ||
133 | XSIG_LL(Denom) <<= 1; | ||
134 | if ( Denom.lsw & 0x80000000 ) | ||
135 | XSIG_LL(Denom) |= 1; | ||
136 | (Denom.lsw) <<= 1; | ||
137 | } | 129 | } |
138 | Denom.msw |= 0x80000000; /* add 1.0 */ | ||
139 | div_Xsig(&accumulator, &Denom, &accumulator); | ||
140 | } | ||
141 | 130 | ||
142 | /* Convert to 64 bit signed-compatible */ | 131 | /* Convert to 64 bit signed-compatible */ |
143 | exponent += round_Xsig(&accumulator); | 132 | exponent += round_Xsig(&accumulator); |
144 | 133 | ||
145 | result = &st(0); | 134 | result = &st(0); |
146 | significand(result) = XSIG_LL(accumulator); | 135 | significand(result) = XSIG_LL(accumulator); |
147 | setexponent16(result, exponent); | 136 | setexponent16(result, exponent); |
148 | 137 | ||
149 | tag = FPU_round(result, 1, 0, FULL_PRECISION, sign); | 138 | tag = FPU_round(result, 1, 0, FULL_PRECISION, sign); |
150 | 139 | ||
151 | setsign(result, sign); | 140 | setsign(result, sign); |
152 | FPU_settag0(tag); | 141 | FPU_settag0(tag); |
153 | 142 | ||
154 | return 0; | 143 | return 0; |
155 | 144 | ||
156 | } | 145 | } |