Texas Instruments TMS320C3x 사용자 설명서

다운로드
페이지 757
Floating-Point Conversion (IEEE Std. 754)
 
5-16
Case 1 maps the IEEE positive NaNs and positive infinity to the single-preci-
sion 2s-complement most positive number. Overflow is also signaled to allow
you to check for these special cases.
Case 2 maps the IEEE negative NaNs and negative infinity to the single-
precision 2s-complement most negative number. Overflow is also signaled
to allow you to check for these special cases.
Case 3 maps the IEEE positive normalized numbers to the identical value in
the 2s-complement positive number.
Case 4 maps the IEEE negative normalized numbers with a nonzero fraction
to the identical value in the 2s-complement negative number.
Case 5 maps the IEEE negative normalized numbers with a 0 fraction to the
identical value in the 2s-complement negative number.
Case 6 maps the IEEE positive and negative denormalized numbers and positive
and negative 0s to a 2s-complement 0.
Based on these definitions of the formats, two versions of the conversion routines
were developed. One version handles the complete definition of the formats. The
other ignores some of the special cases (typically the ones that are rarely used),
but it has the benefit of executing faster than the complete conversion. For this
discussion, the two versions are referred to as the complete version and the fast
version, respectively.