C++浮点数半精度与单精度的相互转换
代码
// 单精度转半精度
unsigned short cpu_float2half(float f)
{
unsigned short ret;
unsigned x = *((int*)(void*)(&f));
unsigned u = (x & 0x7fffffff), remainder, shift, lsb, lsb_s1, lsb_m1;
unsigned sign, exponent, mantissa;
// Get rid of +NaN/-NaN case first.
if (u > 0x7f800000) {
ret = 0x7fffU;
return ret;
}
sign = ((x >> 16) & 0x8000);
// Get rid of +Inf/-Inf, +0/-0.
if (u > 0x477fefff) {
ret = sign | 0x7c00U;
return ret;
}
if (u < 0x33000001) {
ret = (sign | 0x0000);
return ret;
}
exponent = ((u >> 23) & 0xff);
mantissa = (u & 0x7fffff);
if (exponent > 0x70) {
shift = 13;
exponent -= 0x70;
} else {
shift = 0x7e - exponent;
exponent = 0;
mantissa |= 0x800000;
}
lsb = (1 << shift);
lsb_s1 = (lsb >> 1);
lsb_m1 = (lsb - 1);
// Round to nearest even.
remainder = (mantissa & lsb_m1);
mantissa >>= shift;
if (remainder > lsb_s1 || (remainder == lsb_s1 && (mantissa & 0x1))) {
++mantissa;
if (!(mantissa & 0x3ff)) {
++exponent;
mantissa = 0;
}
}
ret = (sign | (exponent << 10) | mantissa);
return ret;
}
// 半精度转单精度
float cpu_half2float(unsigned short x)
{
unsigned sign = ((x >> 15) & 1);
unsigned exponent = ((x >> 10) & 0x1f);
unsigned mantissa = ((x & 0x3ff) << 13);
if (exponent == 0x1f) { /* NaN or Inf */
mantissa = (mantissa ? (sign = 0, 0x7fffff) : 0);
exponent = 0xff;
} else if (!exponent) { /* Denorm or Zero */
if (mantissa) {
unsigned int msb;
exponent = 0x71;
do {
msb = (mantissa & 0x400000);
mantissa <<= 1; /* normalize */
--exponent;
} while (!msb);
mantissa &= 0x7fffff; /* 1.mantissa is implicit */
}
} else {
exponent += 0x70;
}
int temp = ((sign << 31) | (exponent << 23) | mantissa);
return *((float*)((void*)&temp));
}
本文为博主总结文章,欢迎转载,请注明出处。