arm/insts/fplib.cc

2379 fplibAdd(uint16_t op1, uint16_t op2, FPSCR &fpscr)
2382     uint16_t result = fp16_add(op1, op2, 0, modeConv(fpscr), &flags);
2389 fplibAdd(uint32_t op1, uint32_t op2, FPSCR &fpscr)
2392     uint32_t result = fp32_add(op1, op2, 0, modeConv(fpscr), &flags);
2399 fplibAdd(uint64_t op1, uint64_t op2, FPSCR &fpscr)
2402     uint64_t result = fp64_add(op1, op2, 0, modeConv(fpscr), &flags);
2409 fplibCompare(uint16_t op1, uint16_t op2, bool signal_nans, FPSCR &fpscr)
2416     fp16_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
2425         if (op1 == op2 || (!mnt1 && !mnt2)) {
2443 fplibCompare(uint32_t op1, uint32_t op2, bool signal_nans, FPSCR &fpscr)
2450     fp32_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
2459         if (op1 == op2 || (!mnt1 && !mnt2)) {
2477 fplibCompare(uint64_t op1, uint64_t op2, bool signal_nans, FPSCR &fpscr)
2484     fp64_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
2493         if (op1 == op2 || (!mnt1 && !mnt2)) {
2864 fplibMulAdd(uint16_t addend, uint16_t op1, uint16_t op2, FPSCR &fpscr)
2867     uint16_t result = fp16_muladd(addend, op1, op2, 0, modeConv(fpscr), &flags);
2874 fplibMulAdd(uint32_t addend, uint32_t op1, uint32_t op2, FPSCR &fpscr)
2877     uint32_t result = fp32_muladd(addend, op1, op2, 0, modeConv(fpscr), &flags);
2884 fplibMulAdd(uint64_t addend, uint64_t op1, uint64_t op2, FPSCR &fpscr)
2887     uint64_t result = fp64_muladd(addend, op1, op2, 0, modeConv(fpscr), &flags);
2894 fplibDiv(uint16_t op1, uint16_t op2, FPSCR &fpscr)
2897     uint16_t result = fp16_div(op1, op2, modeConv(fpscr), &flags);
2904 fplibDiv(uint32_t op1, uint32_t op2, FPSCR &fpscr)
2907     uint32_t result = fp32_div(op1, op2, modeConv(fpscr), &flags);
2914 fplibDiv(uint64_t op1, uint64_t op2, FPSCR &fpscr)
2917     uint64_t result = fp64_div(op1, op2, modeConv(fpscr), &flags);
3131 fp16_minmaxnum(uint16_t *op1, uint16_t *op2, int sgn)
3134     if (!((uint16_t)~(*op1 << 1) >> FP16_MANT_BITS) &&
3136         *op1 = fp16_infinity(sgn);
3138         (uint16_t)~(*op1 << 1) >> FP16_MANT_BITS)
3143 fp32_minmaxnum(uint32_t *op1, uint32_t *op2, int sgn)
3146     if (!((uint32_t)~(*op1 << 1) >> FP32_MANT_BITS) &&
3148         *op1 = fp32_infinity(sgn);
3150         (uint32_t)~(*op1 << 1) >> FP32_MANT_BITS)
3155 fp64_minmaxnum(uint64_t *op1, uint64_t *op2, int sgn)
3158     if (!((uint64_t)~(*op1 << 1) >> FP64_MANT_BITS) &&
3160         *op1 = fp64_infinity(sgn);
3162         (uint64_t)~(*op1 << 1) >> FP64_MANT_BITS)
3168 fplibMax(uint16_t op1, uint16_t op2, FPSCR &fpscr)
3175     fp16_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3178     if ((x = fp16_process_NaNs(op1, op2, mode, &flags))) {
3181         result = ((sgn1 != sgn2 ? sgn2 : sgn1 ^ (op1 > op2)) ?
3191 fplibMax(uint32_t op1, uint32_t op2, FPSCR &fpscr)
3198     fp32_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3201     if ((x = fp32_process_NaNs(op1, op2, mode, &flags))) {
3204         result = ((sgn1 != sgn2 ? sgn2 : sgn1 ^ (op1 > op2)) ?
3214 fplibMax(uint64_t op1, uint64_t op2, FPSCR &fpscr)
3221     fp64_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3224     if ((x = fp64_process_NaNs(op1, op2, mode, &flags))) {
3227         result = ((sgn1 != sgn2 ? sgn2 : sgn1 ^ (op1 > op2)) ?
3237 fplibMaxNum(uint16_t op1, uint16_t op2, FPSCR &fpscr)
3239     fp16_minmaxnum(&op1, &op2, 1);
3240     return fplibMax<uint16_t>(op1, op2, fpscr);
3245 fplibMaxNum(uint32_t op1, uint32_t op2, FPSCR &fpscr)
3247     fp32_minmaxnum(&op1, &op2, 1);
3248     return fplibMax<uint32_t>(op1, op2, fpscr);
3253 fplibMaxNum(uint64_t op1, uint64_t op2, FPSCR &fpscr)
3255     fp64_minmaxnum(&op1, &op2, 1);
3256     return fplibMax<uint64_t>(op1, op2, fpscr);
3261 fplibMin(uint16_t op1, uint16_t op2, FPSCR &fpscr)
3268     fp16_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3271     if ((x = fp16_process_NaNs(op1, op2, mode, &flags))) {
3274         result = ((sgn1 != sgn2 ? sgn1 : sgn1 ^ (op1 < op2)) ?
3284 fplibMin(uint32_t op1, uint32_t op2, FPSCR &fpscr)
3291     fp32_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3294     if ((x = fp32_process_NaNs(op1, op2, mode, &flags))) {
3297         result = ((sgn1 != sgn2 ? sgn1 : sgn1 ^ (op1 < op2)) ?
3307 fplibMin(uint64_t op1, uint64_t op2, FPSCR &fpscr)
3314     fp64_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3317     if ((x = fp64_process_NaNs(op1, op2, mode, &flags))) {
3320         result = ((sgn1 != sgn2 ? sgn1 : sgn1 ^ (op1 < op2)) ?
3330 fplibMinNum(uint16_t op1, uint16_t op2, FPSCR &fpscr)
3332     fp16_minmaxnum(&op1, &op2, 0);
3333     return fplibMin<uint16_t>(op1, op2, fpscr);
3338 fplibMinNum(uint32_t op1, uint32_t op2, FPSCR &fpscr)
3340     fp32_minmaxnum(&op1, &op2, 0);
3341     return fplibMin<uint32_t>(op1, op2, fpscr);
3346 fplibMinNum(uint64_t op1, uint64_t op2, FPSCR &fpscr)
3348     fp64_minmaxnum(&op1, &op2, 0);
3349     return fplibMin<uint64_t>(op1, op2, fpscr);
3354 fplibMul(uint16_t op1, uint16_t op2, FPSCR &fpscr)
3357     uint16_t result = fp16_mul(op1, op2, modeConv(fpscr), &flags);
3364 fplibMul(uint32_t op1, uint32_t op2, FPSCR &fpscr)
3367     uint32_t result = fp32_mul(op1, op2, modeConv(fpscr), &flags);
3374 fplibMul(uint64_t op1, uint64_t op2, FPSCR &fpscr)
3377     uint64_t result = fp64_mul(op1, op2, modeConv(fpscr), &flags);
3384 fplibMulX(uint16_t op1, uint16_t op2, FPSCR &fpscr)
3391     fp16_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3394     result = fp16_process_NaNs(op1, op2, mode, &flags);
3404             result = fp16_mul(op1, op2, mode, &flags);
3415 fplibMulX(uint32_t op1, uint32_t op2, FPSCR &fpscr)
3422     fp32_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3425     result = fp32_process_NaNs(op1, op2, mode, &flags);
3435             result = fp32_mul(op1, op2, mode, &flags);
3446 fplibMulX(uint64_t op1, uint64_t op2, FPSCR &fpscr)
3453     fp64_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3456     result = fp64_process_NaNs(op1, op2, mode, &flags);
3466             result = fp64_mul(op1, op2, mode, &flags);
3622 fplibRSqrtStepFused(uint16_t op1, uint16_t op2, FPSCR &fpscr)
3629     op1 = fplibNeg<uint16_t>(op1);
3630     fp16_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3633     result = fp16_process_NaNs(op1, op2, mode, &flags);
3641             result = fp16_muladd(fp16_FPThree(0), op1, op2, -1, mode, &flags);
3652 fplibRSqrtStepFused(uint32_t op1, uint32_t op2, FPSCR &fpscr)
3659     op1 = fplibNeg<uint32_t>(op1);
3660     fp32_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3663     result = fp32_process_NaNs(op1, op2, mode, &flags);
3671             result = fp32_muladd(fp32_FPThree(0), op1, op2, -1, mode, &flags);
3682 fplibRSqrtStepFused(uint64_t op1, uint64_t op2, FPSCR &fpscr)
3689     op1 = fplibNeg<uint64_t>(op1);
3690     fp64_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3693     result = fp64_process_NaNs(op1, op2, mode, &flags);
3701             result = fp64_muladd(fp64_FPThree(0), op1, op2, -1, mode, &flags);
3898 fplibRecipStepFused(uint16_t op1, uint16_t op2, FPSCR &fpscr)
3905     op1 = fplibNeg<uint16_t>(op1);
3906     fp16_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3909     result = fp16_process_NaNs(op1, op2, mode, &flags);
3917             result = fp16_muladd(fp16_FPTwo(0), op1, op2, 0, mode, &flags);
3928 fplibRecipStepFused(uint32_t op1, uint32_t op2, FPSCR &fpscr)
3935     op1 = fplibNeg<uint32_t>(op1);
3936     fp32_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3939     result = fp32_process_NaNs(op1, op2, mode, &flags);
3947             result = fp32_muladd(fp32_FPTwo(0), op1, op2, 0, mode, &flags);
3958 fplibRecipStepFused(uint64_t op1, uint64_t op2, FPSCR &fpscr)
3965     op1 = fplibNeg<uint64_t>(op1);
3966     fp64_unpack(&sgn1, &exp1, &mnt1, op1, mode, &flags);
3969     result = fp64_process_NaNs(op1, op2, mode, &flags);
3977             result = fp64_muladd(fp64_FPTwo(0), op1, op2, 0, mode, &flags);
4264 fplibScale(uint16_t op1, uint16_t op2, FPSCR &fpscr)
4267     uint16_t result = fp16_scale(op1, (int16_t)op2, modeConv(fpscr), &flags);
4274 fplibScale(uint32_t op1, uint32_t op2, FPSCR &fpscr)
4277     uint32_t result = fp32_scale(op1, (int32_t)op2, modeConv(fpscr), &flags);
4284 fplibScale(uint64_t op1, uint64_t op2, FPSCR &fpscr)
4287     uint64_t result = fp64_scale(op1, (int64_t)op2, modeConv(fpscr), &flags);
4324 fplibSub(uint16_t op1, uint16_t op2, FPSCR &fpscr)
4327     uint16_t result = fp16_add(op1, op2, 1, modeConv(fpscr), &flags);
4334 fplibSub(uint32_t op1, uint32_t op2, FPSCR &fpscr)
4337     uint32_t result = fp32_add(op1, op2, 1, modeConv(fpscr), &flags);
4344 fplibSub(uint64_t op1, uint64_t op2, FPSCR &fpscr)
4347     uint64_t result = fp64_add(op1, op2, 1, modeConv(fpscr), &flags);
4354 fplibTrigMulAdd(uint8_t coeff_index, uint16_t op1, uint16_t op2, FPSCR &fpscr)
4380         fp16_muladd(coeff[op2 >> (FP16_BITS - 1)][coeff_index], op1,
4388 fplibTrigMulAdd(uint8_t coeff_index, uint32_t op1, uint32_t op2, FPSCR &fpscr)
4414         fp32_muladd(coeff[op2 >> (FP32_BITS - 1)][coeff_index], op1,
4422 fplibTrigMulAdd(uint8_t coeff_index, uint64_t op1, uint64_t op2, FPSCR &fpscr)
4448         fp64_muladd(coeff[op2 >> (FP64_BITS - 1)][coeff_index], op1,
4456 fplibTrigSMul(uint16_t op1, uint16_t op2, FPSCR &fpscr)
4463     uint16_t result = fp16_mul(op1, op1, mode, &flags);
4476 fplibTrigSMul(uint32_t op1, uint32_t op2, FPSCR &fpscr)
4483     uint32_t result = fp32_mul(op1, op1, mode, &flags);
4495 fplibTrigSMul(uint64_t op1, uint64_t op2, FPSCR &fpscr)
4502     uint64_t result = fp64_mul(op1, op1, mode, &flags);
4514 fplibTrigSSel(uint16_t op1, uint16_t op2, FPSCR &fpscr)
4519         op1 = fpOne;
4520     return op1 ^ ((op2 >> 1) << (FP16_BITS - 1));
4525 fplibTrigSSel(uint32_t op1, uint32_t op2, FPSCR &fpscr)
4530         op1 = fpOne;
4531     return op1 ^ ((op2 >> 1) << (FP32_BITS - 1));
4536 fplibTrigSSel(uint64_t op1, uint64_t op2, FPSCR &fpscr)
4541         op1 = fpOne;
4542     return op1 ^ ((op2 >> 1) << (FP64_BITS - 1));