Customized precision

With the software mhodlr, you can easily customized your preferred precisions for the HODLR format conversion. The precision is defined by class @precision, you can simply use a few letters to define the precision, or specify the detail of the precision used.

We give a few examples to illustrate its usage.

One can simply specify the common floating point format by entering a given name or abbreviation (for detail, see the API reference as below)

u = precision('s');

>> ans =
precision with properties:
          t: 24
       emax: 127
      round: 1
  subnormal: 1
     explim: 1
       prob: 0.5000
       flip: 0
   randfunc: @(n)rand(n,1)
          u: 5.9605e-08

To specify the tuple (t, emax), one can enter the first input as an array like

precision([6, 12], 2, 1, 1, 0.5, 1)

Then, the result is:

>> ans =
precision with properties:

          t: 6
       emax: 12
      round: 2
  subnormal: 1
     explim: 1
       prob: 1
       flip: 0.5000
   randfunc: @(n)rand(n,1)
          u: 0.0156

Or you can leave the emax empty to use default value for it, then it can be:

precision(6, 2, 1, 1, 0.5, 1)

the result is similar:

precision with properties:
>> ans =
          t: 6
       emax: 15
      round: 2
  subnormal: 1
     explim: 1
       prob: 1
       flip: 0.5000
   randfunc: @(n)rand(n,1)
          u: 0.0156