| 15 |
|
| |
| 16 |
|
.B \-l |
| 17 |
|
][ |
| 18 |
+ |
.B \-i{f|d}[N] |
| 19 |
+ |
][ |
| 20 |
+ |
.B \-o{f|d} |
| 21 |
+ |
][ |
| 22 |
|
.B \-tC |
| 23 |
|
][ |
| 24 |
|
.B \-N |
| 67 |
|
For products, the geometric mean is computed. |
| 68 |
|
(A logarithmic sum of absolute values is used to avoid overflow, and |
| 69 |
|
zero values are silently ignored.) |
| 70 |
+ |
.PP |
| 71 |
+ |
If the input data is binary, the |
| 72 |
+ |
.I \-id |
| 73 |
+ |
or |
| 74 |
+ |
.I \-if |
| 75 |
+ |
option may be given for 64-bit double or 32-bit float values, respectively. |
| 76 |
+ |
Either option may be followed immediately by an optional |
| 77 |
+ |
count, which defaults to 1, indicating the number of double or float |
| 78 |
+ |
binary values to read per record on the input file. |
| 79 |
+ |
(There can be no space between the option and this count.)\0 |
| 80 |
+ |
Similarly, the |
| 81 |
+ |
.I \-od |
| 82 |
+ |
and |
| 83 |
+ |
.I \-of |
| 84 |
+ |
options specify binary double or float output, respectively. |
| 85 |
+ |
These options do not need a count, as this will be determined by the |
| 86 |
+ |
number of input channels. |
| 87 |
|
.PP |
| 88 |
|
A count can be given as the number of lines to read before |
| 89 |
|
computing a result. |
