[SOLVED] VHDL if condition '=' not suggested , why ?

[verylsi]

Hi,

I have been recently told not to use "=" under if condition rather use ">=" or "<=" .

for eg.

if count = 5 then 
   a <= b;
end if;

Better approach is

if count >= 5 then 
  a <= b;
end if;

I want to know the reason for it, if it is correct.
what is the difference in terms of hardware realization in FPGA ?

 

[root2hell]

= is an assignment operator and we are assumed to compare in a if statement.
use == instead of =.
count =5 // means you are assigning count a value i.e 5
count==5 // means compiler checks that whether count equals to 5 or not.

 

[verylsi]

I was asking in VHDL, I think you are mentioning about verilog

 

[K-J]

It is not a hard and fast rule, it depends. Let's say that 'a' is a three bit number. Then to evaluate the condition 'a=5' would be 'a=101'. To evaluate 'a>=5' would be a=5, 6 or 7 which would be 'a=101', 'a=110', 'a=111' which would reduce to 'a=101' or 'a=11-'. It is likely that no matter what technology is used to implement the logic, roughly the same amount of resources would be used. If anything, in this instance, using >= could use more.

Now change it slightly. Instead of comparing to 5, what if it was a compare with 4. Now you would have 'a=4' would be 'a=100'. To evaluate 'a>=4' would be a=4, 5, 6 or 7 which would be 'a=100', 'a=101', 'a=110', 'a=111' which would reduce to 'a=1--' or simply a(2). In this case, using >= 4 rather than =4 would result in less logic or routing resources since to evaluate the expression using = requires all three bits, but only one bit for >=. So, depending on the exact comparison, there could be a resource advantage.

Another reason for using >= versus = would be to cover 'impossible' conditions. Let's says that normally 'a' can only range from 0 to 5 and 'should' never make it to 6 or 7. But what if it does? If you used >= then those 'impossible' conditions are covered, if you use = then it is not, how does the system recover from that condition?

Kevin Jennings

 

[ads-ee]

Taking KJ's explanation further, suppose the count cycles through 0-5 and stops at 5, the a output is feed back to the count logic to set count back to 0 (perhaps based on the arrival of b). Now suppose some alpha particle causes the the 0 in the 101 to flip state before b arrives and now b is never assigned to a as 5 /= 7. Using count >=5 would protect the design from this lockup condition.

Regards