Non-symmetric Rounding to Negative - 7.2 English

In a modification of the preceding technique, a binary value corresponding to 0.499... is added to the accumulator result and the m LSBs are removed; this is equivalent to the MATLAB software function ceil(x-0.5) . The resource usage advantage is the same, but the bias in this case is towards the negative by 2 ^-(m+1) .

Figure 3-58: Non-symmetric Rounding (a) to Positive (b) to Negative

X-Ref Target - Figure 3-58