|
|
|
@ -42,56 +42,16 @@ float ShortToFloat32(uint16_t src_value) {
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
uint16_t Float32ToShort(float src_value) {
|
|
|
|
|
float *psrcValue = NULL;
|
|
|
|
|
psrcValue = &src_value;
|
|
|
|
|
unsigned int srcValueBit = (unsigned int)(*psrcValue);
|
|
|
|
|
unsigned int sign = srcValueBit >> (FP32_BIT_SIZE - 1);
|
|
|
|
|
unsigned int mantissa = srcValueBit & 0x007FFFFF;
|
|
|
|
|
float32_bits src_value_bits;
|
|
|
|
|
src_value_bits.f = src_value;
|
|
|
|
|
uint16_t res = 0;
|
|
|
|
|
// mantissa
|
|
|
|
|
res += (src_value_bits.u >> 13);
|
|
|
|
|
// exponent
|
|
|
|
|
int exp = ((srcValueBit & 0x7F800000) >> FP32_SIGNIFICAND) + FP16_EXPONENT_BIAS - FP32_EXPONENT_BIAS;
|
|
|
|
|
uint16_t short_res;
|
|
|
|
|
if (exp > 0 && exp < FP16_EXPONENT_MAX) {
|
|
|
|
|
// use rte rounding mode, round the significand, combine sign, exponent and significand into a short.
|
|
|
|
|
short_res = (sign << (FP16_BIT_SIZE - 1)) | (exp << FP16_SIGNIFICAND) |
|
|
|
|
|
((mantissa + 0x00001000) >> (FP32_SIGNIFICAND - FP16_SIGNIFICAND));
|
|
|
|
|
} else if (srcValueBit == 0) {
|
|
|
|
|
short_res = 0;
|
|
|
|
|
} else {
|
|
|
|
|
if (exp <= 0) {
|
|
|
|
|
short_res = 0;
|
|
|
|
|
if (exp >= FP16_EXPONENT_MIN) {
|
|
|
|
|
mantissa = (mantissa | 0x00800000) >> (1 - exp);
|
|
|
|
|
if ((mantissa & 0x00001000) > 0) {
|
|
|
|
|
mantissa = mantissa + 0x00002000;
|
|
|
|
|
}
|
|
|
|
|
short_res = (sign << FP16_EXPONENT_BIAS) | (mantissa >> (FP32_SIGNIFICAND - FP16_SIGNIFICAND));
|
|
|
|
|
}
|
|
|
|
|
} else if (exp == (FP32_EXPONENT_MAX - FP32_EXPONENT_BIAS + FP16_EXPONENT_BIAS)) {
|
|
|
|
|
if (mantissa == 0) {
|
|
|
|
|
// input float is infinity, return infinity half
|
|
|
|
|
short_res = (sign << FP16_EXPONENT_BIAS) | 0x7C00;
|
|
|
|
|
} else {
|
|
|
|
|
// input float is NaN, return half NaN
|
|
|
|
|
short_res = (sign << FP16_EXPONENT_BIAS) | 0x7C00 | (mantissa >> (FP32_SIGNIFICAND - FP16_SIGNIFICAND));
|
|
|
|
|
}
|
|
|
|
|
} else {
|
|
|
|
|
// exp > 0, normalized single, round to nearest
|
|
|
|
|
if ((mantissa & 0x00001000) > 0) {
|
|
|
|
|
mantissa = mantissa + 0x00002000;
|
|
|
|
|
if ((mantissa & 0x00800000) > 0) {
|
|
|
|
|
mantissa = 0;
|
|
|
|
|
exp = exp + 1;
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
if (exp > FP16_EXPONENT_MAX) {
|
|
|
|
|
// exponent overflow - return infinity half
|
|
|
|
|
short_res = (sign << FP16_EXPONENT_BIAS) | 0x7C00;
|
|
|
|
|
} else {
|
|
|
|
|
// combine sign, exp and mantissa into normalized half
|
|
|
|
|
short_res = (sign << FP16_EXPONENT_BIAS) | (exp << FP16_SIGNIFICAND) |
|
|
|
|
|
(mantissa >> (FP32_SIGNIFICAND - FP16_SIGNIFICAND));
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
}
|
|
|
|
|
return short_res;
|
|
|
|
|
res += (src_value_bits.u >> 13) & 0x3fc00;
|
|
|
|
|
res -= (127 - 15) << 13;
|
|
|
|
|
|
|
|
|
|
// sign
|
|
|
|
|
res |= (src_value_bits.u & 0x400000000) >> 16;
|
|
|
|
|
return res;
|
|
|
|
|
}
|
|
|
|
|