[SOLVED] is it better to use full case than the casex statement ?

I am Designing a Instruction decoder 12 bit to 15 bit ..using case statement total 4096 statements .Earlier I used casex so that only few statements need to written .Will it be good to do with case

earlier the code was like this

casex(op)

35 cases

default case


now code changed to

case(op)
all 4096 cases


///Will there be any priority issues related to this approach

you can use default statement for the bit patterns you don't care.

lostinxlation said:

you can use default statement for the bit patterns you don't care.

Code:

case(in)
   12'h000: out = <some value1>
   12'h001: out = <some value2>
   12'h002: out = <some value3>
   default   : out = < I don't care in this case so let's put whatever value here>
endcase

Click to expand...

particular case
is

always @(inst) begin
casex (inst) // synopsys parallel_case
// *** Byte-Oriented File Register Operations
//
// A A ALU W F Z C B T O
// L L O W W W W D R P
// U U P E E E E P I T
// A B O S
// L
12'b0000_0000_0000: decodes = 15'b00_00_0000_0_0_0_0_0_0_0; // NOP
// 12'b0000_001X_XXXX: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF this expression has real 32 cases ....which expands as follows
/*12'b000000100000: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(0)
12'b000000100001: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(1)
12'b000000100010: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(2)
12'b000000100011: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(3)
12'b000000100100: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(4)
12'b000000100101: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(5)
12'b000000100110: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(6)
12'b000000100111: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(7)
12'b000000101000: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(8)
12'b000000101001: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(9)
12'b000000101010: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(10)
12'b000000101011: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(11)
12'b000000101100: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(12)
12'b000000101101: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(13)
12'b000000101110: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(14)
12'b000000101111: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(15)
12'b000000110000: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(16)
12'b000000110001: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(17)
12'b000000110010: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(18)
12'b000000110011: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(19)
12'b000000110100: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(20)
12'b000000110101: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(21)
12'b000000110110: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(22)
12'b000000110111: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(23)
12'b000000111000: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(24)
12'b000000111001: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(25)
12'b000000111010: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(26)
12'b000000111011: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(27)
12'b000000111100: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(28)
12'b000000111101: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(29)
12'b000000111110: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(30)
12'b000000111111: decodes = 15'b00_00_0010_0_1_0_0_0_0_0; // MOVWF(31)

so ....what will be a better strategy ? fully specifying the cases or just go with casex ....I have gone through sunburst design paper cliff cummings and it seems the casex has got some trouble ...synthesis stage

blooz said:

so ....what will be a better strategy ? fully specifying the cases or just go with casex ....I have gone through sunburst design paper cliff cummings and it seems the casex has got some trouble ...synthesis stage

Click to expand...

Hi blooz,

What I got from that paper is (short version):
- no full_case parallel_case directives, unless you really need it.
- no casex. ever.
- for case statements with don't-cares, use a casez statement, with a ? for the don't-cares.
- always include a default statement, even when all cases have been specified. This goes for both case and casez.

I've been sticking with that plan, and so far everything works out fine during simulation & synthesis. Below is an example of working code with a casez statement that uses ? for the don't-cares.


So to answer your question... you probably want to use casez with ? for the don't-cares, similar to the above module.

so the change needed is casex() to casez and replacing 'x' with '?'.

---------- Post added at 13:05 ---------- Previous post was at 12:58 ----------

so the change needed is casex() to casez and replacing 'x' with '?'.

blooz said:

so the change needed is casex() to casez and replacing 'x' with '?.

Click to expand...

Yes. And making sure that you have a sensible default statement.

Oh yeah, and it also can't hurt to check the synthesis report to see what it actually synthesizes based on your case statement...

As long as the synthesizer can recognize the casex statement, it is fine to use. Casex will be expended to normal cases in your any way.