XST and Verilog $readmemh

In theory, XST claims to support the Verilog $readmemh to initialize
memory.  I'm
using the latest 9.x s/w verion.

I look at the .syr output file from XST, and it claims to have read
the file.  But...

If I hook a logic analyzer up to the output of the memory, it looks
like it never got
initialized.

Another problem I'm seeing is that XST appears to not like having an
address line
(@000) in the file.  Also, if the the file is too short, XST complains
and discards
the initialization.

I'll dig into this some more, but I was wondering if anyone had had
any success with this.

Thanks!

John Providenza

On Apr 11, 6:32 pm, "johnp" <johnp3+nos...@probo.com> wrote:
> In theory, XST claims to support the Verilog $readmemh to initialize
> memory.  I'm
> using the latest 9.x s/w verion.
>
> I look at the .syr output file from XST, and it claims to have read
> the file.  But...
>
> If I hook a logic analyzer up to the output of the memory, it looks
> like it never got
> initialized.
>
> Another problem I'm seeing is that XST appears to not like having an
> address line
> (@000) in the file.  Also, if the the file is too short, XST complains
> and discards
> the initialization.
>
> I'll dig into this some more, but I was wondering if anyone had had
> any success with this.
>
> Thanks!
>
> John Providenza



Here's some more info... In the .syr report file from XST, I see the
lines:

INFO:Xst:2546 - "C:/usr/JOHNP/mentor/enet/hw/pblaze_rtl/../rtl/
pblaze.v" line 84: reading initialization file "../sim/foo.memh".

but later...

WARNING:Xst:1781 - Signal <dp_mem> is used but never assigned. Tied to
default value.

This explains why I get 0's from the memory - it's been optimised out
and tied to 0.

Any ideas for simple work-arounds?  I've opened a WebCase, but my
expectations from
them are realistic, ie, very low.

John Providenza