On 6/16/06, Vinicius Santos <[EMAIL PROTECTED]> wrote:
Even though it sounds silly, I was trying to play around with
constructing a simple async memory. So I build this module:
[snip]
But I got the following output when executing a.out:
0 write=1 read=0 addr=0003 in=ff00 out=xxxx
1 write=0 read=1 addr=0003 in=ff00 out=xx00
I assume those 'x' means 'Don't Care', but why am I getting those?
It's even strange because it sets 0 accordingly.
This one stumped me for a couple minutes. It wasn't until I tried
changing around the mem module did I realize what was going happening.
All variables are initialized as X if unassigned, which means
invalid/conflict. (Don't care is in the verilog syntax, but uses a
dash '-')
I thought that meant the memory was only writing the lower 8-bits or
reading back the lower 8-bits.
But I traced the problem to this line:
wire [15:0] out = 0;
Outputs aren't usually initialized. What's happening is there are two
assignments (drivers) to the 'out' wire. It should be:
wire [15:0] out;
To pseudo-code what was happening in the testbench:
assign out = 0
assign out = memory_out;
If done physically, two wires are connected to each bit the 'out'
wire. A set of ground wires and the set from the memory.
When you run the testbench, you get the following:
0 write=1 read=0 addr=0003 in=ff00 out=xxxx
1 write=0 read=1 addr=0003 in=ff00 out=xx00
At time 0, out is XXXX because it's unassigned. That's expected.
At time 1, the memory module tries to assign output to 'hff00. but
the '= 0' assignment tries to force it to 'h0000.
The simulator doesn't know how to resolve the two. The top 8-bits are
assigned a 1 and a 0 at the same time so the sim assigns it an 'X'.
The bottom 8-bits are assigned a 0 and a 0, so it's easy to for the
sim to resolve: 0.
If you now go back to look at time 0, it's actually 0 and X on each
wire, which resolves to X because X and anything is X.
Hope that helps. Changing that one line should fix the output to:
0 write=1 read=0 addr=0003 in=ff00 out=xxxx
1 write=0 read=1 addr=0003 in=ff00 out=ff00
Ryan
_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)
