diff options
Diffstat (limited to 'target/hexagon/arch.c')
| -rw-r--r-- | target/hexagon/arch.c | 181 |
1 files changed, 146 insertions, 35 deletions
diff --git a/target/hexagon/arch.c b/target/hexagon/arch.c index 09de124..68a55b3 100644 --- a/target/hexagon/arch.c +++ b/target/hexagon/arch.c @@ -27,6 +27,97 @@ #define SF_MANTBITS 23 #define float32_nan make_float32(0xffffffff) +/* + * These three tables are used by the cabacdecbin instruction + */ +const uint8_t rLPS_table_64x4[64][4] = { + {128, 176, 208, 240}, + {128, 167, 197, 227}, + {128, 158, 187, 216}, + {123, 150, 178, 205}, + {116, 142, 169, 195}, + {111, 135, 160, 185}, + {105, 128, 152, 175}, + {100, 122, 144, 166}, + {95, 116, 137, 158}, + {90, 110, 130, 150}, + {85, 104, 123, 142}, + {81, 99, 117, 135}, + {77, 94, 111, 128}, + {73, 89, 105, 122}, + {69, 85, 100, 116}, + {66, 80, 95, 110}, + {62, 76, 90, 104}, + {59, 72, 86, 99}, + {56, 69, 81, 94}, + {53, 65, 77, 89}, + {51, 62, 73, 85}, + {48, 59, 69, 80}, + {46, 56, 66, 76}, + {43, 53, 63, 72}, + {41, 50, 59, 69}, + {39, 48, 56, 65}, + {37, 45, 54, 62}, + {35, 43, 51, 59}, + {33, 41, 48, 56}, + {32, 39, 46, 53}, + {30, 37, 43, 50}, + {29, 35, 41, 48}, + {27, 33, 39, 45}, + {26, 31, 37, 43}, + {24, 30, 35, 41}, + {23, 28, 33, 39}, + {22, 27, 32, 37}, + {21, 26, 30, 35}, + {20, 24, 29, 33}, + {19, 23, 27, 31}, + {18, 22, 26, 30}, + {17, 21, 25, 28}, + {16, 20, 23, 27}, + {15, 19, 22, 25}, + {14, 18, 21, 24}, + {14, 17, 20, 23}, + {13, 16, 19, 22}, + {12, 15, 18, 21}, + {12, 14, 17, 20}, + {11, 14, 16, 19}, + {11, 13, 15, 18}, + {10, 12, 15, 17}, + {10, 12, 14, 16}, + {9, 11, 13, 15}, + {9, 11, 12, 14}, + {8, 10, 12, 14}, + {8, 9, 11, 13}, + {7, 9, 11, 12}, + {7, 9, 10, 12}, + {7, 8, 10, 11}, + {6, 8, 9, 11}, + {6, 7, 9, 10}, + {6, 7, 8, 9}, + {2, 2, 2, 2} +}; + +const uint8_t AC_next_state_MPS_64[64] = { + 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, + 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, + 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, + 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, + 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, + 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, + 61, 62, 62, 63 +}; + + +const uint8_t AC_next_state_LPS_64[64] = { + 0, 0, 1, 2, 2, 4, 4, 5, 6, 7, + 8, 9, 9, 11, 11, 12, 13, 13, 15, 15, + 16, 16, 18, 18, 19, 19, 21, 21, 22, 22, + 23, 24, 24, 25, 26, 26, 27, 27, 28, 29, + 29, 30, 30, 30, 31, 32, 32, 33, 33, 33, + 34, 34, 35, 35, 35, 36, 36, 36, 37, 37, + 37, 38, 38, 63 +}; + #define BITS_MASK_8 0x5555555555555555ULL #define PAIR_MASK_8 0x3333333333333333ULL #define NYBL_MASK_8 0x0f0f0f0f0f0f0f0fULL @@ -76,19 +167,6 @@ uint64_t deinterleave(uint64_t src) return myeven | (myodd << 32); } -uint32_t carry_from_add64(uint64_t a, uint64_t b, uint32_t c) -{ - uint64_t tmpa, tmpb, tmpc; - tmpa = fGETUWORD(0, a); - tmpb = fGETUWORD(0, b); - tmpc = tmpa + tmpb + c; - tmpa = fGETUWORD(1, a); - tmpb = fGETUWORD(1, b); - tmpc = tmpa + tmpb + fGETUWORD(1, tmpc); - tmpc = fGETUWORD(1, tmpc); - return tmpc; -} - int32_t conv_round(int32_t a, int n) { int64_t val; @@ -108,7 +186,7 @@ int32_t conv_round(int32_t a, int n) /* Floating Point Stuff */ -static const int softfloat_roundingmodes[] = { +static const FloatRoundMode softfloat_roundingmodes[] = { float_round_nearest_even, float_round_to_zero, float_round_down, @@ -156,12 +234,6 @@ void arch_fpop_end(CPUHexagonState *env) } } -static float32 float32_mul_pow2(float32 a, uint32_t p, float_status *fp_status) -{ - float32 b = make_float32((SF_BIAS + p) << SF_MANTBITS); - return float32_mul(a, b, fp_status); -} - int arch_sf_recip_common(float32 *Rs, float32 *Rt, float32 *Rd, int *adjust, float_status *fp_status) { @@ -200,12 +272,13 @@ int arch_sf_recip_common(float32 *Rs, float32 *Rt, float32 *Rd, int *adjust, /* or put Inf in num fixup? */ uint8_t RsV_sign = float32_is_neg(RsV); uint8_t RtV_sign = float32_is_neg(RtV); + /* Check that RsV is NOT infinite before we overwrite it */ + if (!float32_is_infinity(RsV)) { + float_raise(float_flag_divbyzero, fp_status); + } RsV = infinite_float32(RsV_sign ^ RtV_sign); RtV = float32_one; RdV = float32_one; - if (float32_is_infinity(RsV)) { - float_raise(float_flag_divbyzero, fp_status); - } } else if (float32_is_infinity(RtV)) { RsV = make_float32(0x80000000 & (RsV ^ RtV)); RtV = float32_one; @@ -230,22 +303,22 @@ int arch_sf_recip_common(float32 *Rs, float32 *Rt, float32 *Rd, int *adjust, if ((n_exp - d_exp + SF_BIAS) <= SF_MANTBITS) { /* Near quotient underflow / inexact Q */ PeV = 0x80; - RtV = float32_mul_pow2(RtV, -64, fp_status); - RsV = float32_mul_pow2(RsV, 64, fp_status); + RtV = float32_scalbn(RtV, -64, fp_status); + RsV = float32_scalbn(RsV, 64, fp_status); } else if ((n_exp - d_exp + SF_BIAS) > (SF_MAXEXP - 24)) { /* Near quotient overflow */ PeV = 0x40; - RtV = float32_mul_pow2(RtV, 32, fp_status); - RsV = float32_mul_pow2(RsV, -32, fp_status); + RtV = float32_scalbn(RtV, 32, fp_status); + RsV = float32_scalbn(RsV, -32, fp_status); } else if (n_exp <= SF_MANTBITS + 2) { - RtV = float32_mul_pow2(RtV, 64, fp_status); - RsV = float32_mul_pow2(RsV, 64, fp_status); + RtV = float32_scalbn(RtV, 64, fp_status); + RsV = float32_scalbn(RsV, 64, fp_status); } else if (d_exp <= 1) { - RtV = float32_mul_pow2(RtV, 32, fp_status); - RsV = float32_mul_pow2(RsV, 32, fp_status); + RtV = float32_scalbn(RtV, 32, fp_status); + RsV = float32_scalbn(RsV, 32, fp_status); } else if (d_exp > 252) { - RtV = float32_mul_pow2(RtV, -32, fp_status); - RsV = float32_mul_pow2(RsV, -32, fp_status); + RtV = float32_scalbn(RtV, -32, fp_status); + RsV = float32_scalbn(RsV, -32, fp_status); } RdV = 0; ret = 1; @@ -265,7 +338,7 @@ int arch_sf_invsqrt_common(float32 *Rs, float32 *Rd, int *adjust, int r_exp; int ret = 0; RsV = *Rs; - if (float32_is_infinity(RsV)) { + if (float32_is_any_nan(RsV)) { if (extract32(RsV, 22, 1) == 0) { float_raise(float_flag_invalid, fp_status); } @@ -287,7 +360,7 @@ int arch_sf_invsqrt_common(float32 *Rs, float32 *Rd, int *adjust, /* Basic checks passed */ r_exp = float32_getexp(RsV); if (r_exp <= 24) { - RsV = float32_mul_pow2(RsV, 64, fp_status); + RsV = float32_scalbn(RsV, 64, fp_status); PeV = 0xe0; } RdV = 0; @@ -298,3 +371,41 @@ int arch_sf_invsqrt_common(float32 *Rs, float32 *Rd, int *adjust, *adjust = PeV; return ret; } + +const uint8_t recip_lookup_table[128] = { + 0x0fe, 0x0fa, 0x0f6, 0x0f2, 0x0ef, 0x0eb, 0x0e7, 0x0e4, + 0x0e0, 0x0dd, 0x0d9, 0x0d6, 0x0d2, 0x0cf, 0x0cc, 0x0c9, + 0x0c6, 0x0c2, 0x0bf, 0x0bc, 0x0b9, 0x0b6, 0x0b3, 0x0b1, + 0x0ae, 0x0ab, 0x0a8, 0x0a5, 0x0a3, 0x0a0, 0x09d, 0x09b, + 0x098, 0x096, 0x093, 0x091, 0x08e, 0x08c, 0x08a, 0x087, + 0x085, 0x083, 0x080, 0x07e, 0x07c, 0x07a, 0x078, 0x075, + 0x073, 0x071, 0x06f, 0x06d, 0x06b, 0x069, 0x067, 0x065, + 0x063, 0x061, 0x05f, 0x05e, 0x05c, 0x05a, 0x058, 0x056, + 0x054, 0x053, 0x051, 0x04f, 0x04e, 0x04c, 0x04a, 0x049, + 0x047, 0x045, 0x044, 0x042, 0x040, 0x03f, 0x03d, 0x03c, + 0x03a, 0x039, 0x037, 0x036, 0x034, 0x033, 0x032, 0x030, + 0x02f, 0x02d, 0x02c, 0x02b, 0x029, 0x028, 0x027, 0x025, + 0x024, 0x023, 0x021, 0x020, 0x01f, 0x01e, 0x01c, 0x01b, + 0x01a, 0x019, 0x017, 0x016, 0x015, 0x014, 0x013, 0x012, + 0x011, 0x00f, 0x00e, 0x00d, 0x00c, 0x00b, 0x00a, 0x009, + 0x008, 0x007, 0x006, 0x005, 0x004, 0x003, 0x002, 0x000, +}; + +const uint8_t invsqrt_lookup_table[128] = { + 0x069, 0x066, 0x063, 0x061, 0x05e, 0x05b, 0x059, 0x057, + 0x054, 0x052, 0x050, 0x04d, 0x04b, 0x049, 0x047, 0x045, + 0x043, 0x041, 0x03f, 0x03d, 0x03b, 0x039, 0x037, 0x036, + 0x034, 0x032, 0x030, 0x02f, 0x02d, 0x02c, 0x02a, 0x028, + 0x027, 0x025, 0x024, 0x022, 0x021, 0x01f, 0x01e, 0x01d, + 0x01b, 0x01a, 0x019, 0x017, 0x016, 0x015, 0x014, 0x012, + 0x011, 0x010, 0x00f, 0x00d, 0x00c, 0x00b, 0x00a, 0x009, + 0x008, 0x007, 0x006, 0x005, 0x004, 0x003, 0x002, 0x001, + 0x0fe, 0x0fa, 0x0f6, 0x0f3, 0x0ef, 0x0eb, 0x0e8, 0x0e4, + 0x0e1, 0x0de, 0x0db, 0x0d7, 0x0d4, 0x0d1, 0x0ce, 0x0cb, + 0x0c9, 0x0c6, 0x0c3, 0x0c0, 0x0be, 0x0bb, 0x0b8, 0x0b6, + 0x0b3, 0x0b1, 0x0af, 0x0ac, 0x0aa, 0x0a8, 0x0a5, 0x0a3, + 0x0a1, 0x09f, 0x09d, 0x09b, 0x099, 0x097, 0x095, 0x093, + 0x091, 0x08f, 0x08d, 0x08b, 0x089, 0x087, 0x086, 0x084, + 0x082, 0x080, 0x07f, 0x07d, 0x07b, 0x07a, 0x078, 0x077, + 0x075, 0x074, 0x072, 0x071, 0x06f, 0x06e, 0x06c, 0x06b, +}; |