SystemVerilog expression parsing

[xjtu.yohuang]

Hi,

The IEEE1685-2014 standard says it utilizes SystemVerilog expressions as a means to specify an equation as the value of an element.

In below example, there is a parameter my_param of 'byte' type, and its value is an expression (param_A - param_B):

<ipxact:moduleParameters>
 <ipxact:moduleParameter type="byte">
 <ipxact:name>my_param</ipxact:name>
 <ipxact:value>param_A - param_B</ipxact:value>
 </ipxact:moduleParameter>
 </ipxact:moduleParameters>

Is it valid to set 32-bit value to param_A and param_B, e.g., param_A==0xFFFFFFFF and param_B==0xFFFFFFFE, whose computed value is 1?

Should the expression parser convert param_A and param_B to byte type firstly and then do computation, or cast each operands to its int type and do int computation and convert the result to byte?

 

thanks,

Yohuang

[kock]

Hi Yohuang,

param_A and param_B should be casted to there own type and range first. Then the subtraction is applied and the result should be casted to the type and range of my_param.

Please note that there is a newer version 1685-2022. Please use this version if you can because it supersedes 1685-2014.

Best regards,
Erwin

 

 

[xjtu.yohuang]

Thanks for quick response.

What if the my_param is of type int, and the ipxact:value looks like this:

<ipxact:moduleParameters>
 <ipxact:moduleParameter type="int">
 <ipxact:name>my_param</ipxact:name>
 <ipxact:value>0x100000000 - 0xFFFFFFFF</ipxact:value>
 </ipxact:moduleParameter>
 </ipxact:moduleParameters>

Should systemVerilog parser cast up-cast both 0x100000000 (33bit) and 0xFFFFFFFF (32bit) to longint (64bit) and then cast the substraction result to int?

>> Please note that there is a newer version 1685-2022. Please use this version if you can because it supersedes 1685-2014.

Thanks for the suggestion, yes, I do have plan to move this newer version. Currently I'm doing some experiments with 1685-2014.


thanks,

Yohuang

[kock]

Hi Yohuang,

I need to read the SV on this. My understanding is that unsized constant numbers need to be represented as int (32 bit) and therefore the substraction is also represented as int (32 bits) (Section 11.6.1 and 11.6.2 in IEEE Std 1800-2012). So no up conversion to longint here.

Best regards,
Erwin