MATLAB: Do I get an error while trying to compare two numbers of FIXDT data type in the Embedded Matlab Block comparison operator in Simulink 7.3 (R2009a)

fifixdtsimulinkufixuint8

I am comparing two fixed point signals inside of an Embedded MATLAB Block, one of type fixdt(0,9,0), and one of type fixdt(0,7,0), as below:

function y = fcn(u1,u2)
%#eml
if u1>u2
y = 1;
else
      y=0;
end

However, if I change either one to data type fixdt(0,8,0), Simulink changes the data type to a uint8 (not a ufix8), and the Embedded MATLAB block cannot perform the operation. I get the following error:

 In fi > non-fi, the non-fi must be a constant.
 Function 'Embedded MATLAB Function1' (#104.33.38), line 3, column 4:
 "u1>u2"

Best Answer

This is expected behavior. To get the Embedded MATLAB block to recognize this signal as an FI object, please perform the following steps:

There is drop-down menu in the EML Ports and Data Manager that says "Treat these inherited Simulink signal types as fi objects" - select "Fixed-point & integer" for this parameter.

Similarly:

 fixdt(1,8,0) -> int8
 fixdt(0,16,0) -> uint16
 fixdt(1,16,0) -> int16
 fixdt(0,32,0) -> uint32
 fixdt(1,32,0) -> int32
 fixdt(0,64,0) -> uint64 AND
 fixdt(1,64,0) -> int64

If your wordlength is not 8,16,32 or 64 (when fraction length is 0), only then is the type treated as a non-builtin fixed-point type.

Alternately, you can use the following comparison in your EML code:

if u1 > fi(u2,0,8,0)

Related Solutions

MATLAB: Difference between int32 and fixed point

HDL Coder treats all integral types (both int and uint types) as the fixed point types that they are equivalent to. There is no need to convert integral types to fixed point.

For example, an int16 type is equivalent to fixdt(1,16,0), while an uint8 is equivalent to fixdt(0,8,0).

MATLAB: Some forms of Simulink logging write 64 bit integers as fi objects. Does that mean I need a Fixed-Point Designer license

As of R2017a,

a Fixed-Point Designer license would NOT be required.

As of R2017a,

constructing a fi object whose type is equivalent to MATLAB's 11 built-in numerictypes

double, single, logical, uint8, int8, uint16, int16, uint32, int32, uint64, int64

does not require a Fixed-Point Designer license.

For example, this code

v_u64 = fi( 123, 0, 64, 0)
v_s64 = fi( 123, 1, 64, 0)
v_u32 = fi( 123, 0, 32, 0)
v_s32 = fi( 123, 1, 32, 0)
v_u16 = fi( 123, 0, 16, 0)
v_s16 = fi( 123, 1, 16, 0)
v_u8  = fi( 123, 0,  8, 0)
v_s8  = fi( 123, 1,  8, 0)

would not require a Fixed-Point Designer license.

In R2016b and earlier,

a Fixed-Point Designer license would be required.

This change coincided with another important license change.

In R2016b and earlier,

use of 64 bit integers in Simulink requires a Fixed-Point Designer license.

As of R2017a,

usage of 64 bit integers in Simulink does not require any additional licenses.

FYI:

If you'd like to change a fi object to its built-in MATLAB eqivalent, it's easy to do.

In R2020a and later, you can use utilities castFiToMATLAB

a = fi(123,0,64,0)
a = 
   123
          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Unsigned
            WordLength: 64
        FractionLength: 0
>> b = castFiToMATLAB(a)
b =
  uint64
   123

or castFiToInt

> a = fi(123,0,64,0)
a = 
   123
          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Unsigned
            WordLength: 64
        FractionLength: 0
>> b = castFiToInt(a)
b =
  uint64
   123

In earlier releases, you can simply cast using the name of the MATLAB class (a.k.a. data type name)

for example

>> a = fi(123,0,64,0)
a = 
   123
          DataTypeMode: Fixed-point: binary point scaling
            Signedness: Unsigned
            WordLength: 64
        FractionLength: 0
>> b = uint64( a )
b =
  uint64
   123

Best Answer

Related Solutions

MATLAB: Difference between int32 and fixed point

MATLAB: Some forms of Simulink logging write 64 bit integers as fi objects. Does that mean I need a Fixed-Point Designer license

Related Question