The reset value being stuck at 0 may not be a significant issue but it is one 
of two issues I am having and the easiest to debug.

The other issue is that if the addr 0 is not writeable, then the control port 
will not read/write from any of the registers. All of it simulates fine but 
after building and testing, I get ACK timeouts on the controlport bus.

Jeff

From: Hodges, Jeff via USRP-users <[email protected]>
Sent: Monday, September 20, 2021 3:09 PM
To: Wade Fife <[email protected]>; Rob Kossler <[email protected]>
Cc: [email protected]
Subject: [USRP-users] Re: rfnoc 4.0 gain block tutorial --> register problem

Rob/Wade,

Thank you for the feedback. I have been running the testbenches and they 
succeed but when I build the code the default value is still 0.

I went back and copied the two files: rfnoc_block_gain.v and 
rfnoc_block_gain_tb.sv to the OOT module, replacing the gain code I had, and 
reran the testbench and rebuilt the image and the default value of register 
address 0x0 is still 0.

Any thoughts on what else could be wrong?

Jeff

From: Wade Fife <[email protected]<mailto:[email protected]>>
Sent: Tuesday, September 14, 2021 3:25 PM
To: Rob Kossler <[email protected]<mailto:[email protected]>>
Cc: Hodges, Jeff 
<[email protected]<mailto:[email protected]>>; 
[email protected]<mailto:[email protected]>
Subject: Re: [USRP-users] Re: rfnoc 4.0 gain block tutorial --> register problem

Hi Jeff,

I agree with Rob that "=" should be "<=" but I don't think that's causing any 
harm here. I copied and pasted your code into the gain example and tested it, 
and it seems to work as I expected. So there's probably something wrong that's 
unrelated to that code snippet. Take a look at the gain example to see what's 
different in yours. I would also suggest simulating your code rather than 
resorting to chipscope. You can use the example gain testbench as a starting 
point.

https://github.com/EttusResearch/uhd/blob/master/host/examples/rfnoc-example/fpga/rfnoc_block_gain/rfnoc_block_gain.v
https://github.com/EttusResearch/uhd/blob/master/host/examples/rfnoc-example/fpga/rfnoc_block_gain/rfnoc_block_gain_tb.sv

The reset should be asserted when the graph is first initialized, so you should 
not be able to read the initial value of the register (REG_USER_DEFAULT), only 
the reset value (REG_USER_DEFAULT2). If that reset isn't happening then perhaps 
there's something wrong with the reset connection or the clock connection. 
Also, it's usually good practice to make the reset and initial value the same. 
Some FPGAs don't really support them being different, although I think this is 
supported on Xilinx 7 Series.

Thanks,

Wade

On Mon, Sep 13, 2021 at 3:03 PM Rob Kossler 
<[email protected]<mailto:[email protected]>> wrote:
Hi Jeff,
I am mostly a novice at FPGA development, but I have created a few blocks in 
rfnoc 4.0 with multiple registers and they seem to behave properly. One thing I 
noticed in your original post below is that when you assign the value 
REG_USER_DEFAULT2, you use "=" rather than "<=". It seems it should be the 
latter.  I don't really know what issues, if any, this would cause.
Rob

On Mon, Sep 13, 2021 at 1:07 PM Hodges, Jeff via USRP-users 
<[email protected]<mailto:[email protected]>> wrote:

I'm now convinced there is something majorly wrong with UHD4.0 controlport.  
Not only am I experiencing the issue I have described here but if I create an 
image with two write registers (0x0 and 0x4) it works, but if I create two 
write registers addresses of 0x4 and 0x8, it fails both read and write.

Is anyone using this UHD4.0? I never had these problems with UHD3.15.  I'm 
building for the X310.



I can try to debug with chipscope but that is a pain and would rather not do 
waste time if someone else has already identified the source of this problem.



Thanks,



Jeff

________________________________
From: Hodges, Jeff via USRP-users 
<[email protected]<mailto:[email protected]>>
Sent: Monday, September 13, 2021 11:48:06 AM
To: [email protected]<mailto:[email protected]>
Subject: [USRP-users] Re: rfnoc 4.0 gain block tutorial --> register problem


I can read and write from register addr 0x0 but the default value read from 
peek32(0x0) is 0 regardless of what I set the REG_USER_DEFAULT value to.



This only occurs on register addr 0x0.



Jeff

________________________________
From: Hodges, Jeff via USRP-users 
<[email protected]<mailto:[email protected]>>
Sent: Saturday, September 11, 2021 5:02 PM
To: [email protected]<mailto:[email protected]>
Subject: [USRP-users] rfnoc 4.0 gain block tutorial --> register problem


Following the gain block tutorial from GRCON20, when I read from the registers 
using gain_block->regs().peek32(0x0), the default value that is return is not 
the same as set during reset. The default value I read is always 0 even when my 
verilog sets it to some other value after rebuilding the image.



Has anyone seen this behavior?



Also i am having problems with other FPGA builds involving the registers to 
please tell me if something is wrong with rfnoc 4.0 controlport.

All i've done is taken the default gain block and added another default value 
after reset:




  localparam REG_USER_ADDR    = 0; // Address for example user register
  localparam REG_USER_DEFAULT = 1; // Default value for user register
  localparam REG_USER_DEFAULT2 = 2; //Default value for user register after 
reset

  reg [31:0] reg_user = REG_USER_DEFAULT;

  always @(posedge ctrlport_clk) begin
    if (ctrlport_rst) begin
      reg_user = REG_USER_DEFAULT2;
    end else begin
      // Default assignment
      m_ctrlport_resp_ack <= 0;

      // Read user register
      if (m_ctrlport_req_rd) begin // Read request
        case (m_ctrlport_req_addr)
          REG_USER_ADDR: begin
            m_ctrlport_resp_ack  <= 1;
            m_ctrlport_resp_data <= reg_user;
          end
        endcase
      end

      // Write user register
      if (m_ctrlport_req_wr) begin // Write requst
        case (m_ctrlport_req_addr)
          REG_USER_ADDR: begin
            m_ctrlport_resp_ack <= 1;
            reg_user            <= m_ctrlport_req_data[31:0];
          end
        endcase
      end
    end
  end


Jeff
_______________________________________________
USRP-users mailing list -- 
[email protected]<mailto:[email protected]>
To unsubscribe send an email to 
[email protected]<mailto:[email protected]>
_______________________________________________
USRP-users mailing list -- 
[email protected]<mailto:[email protected]>
To unsubscribe send an email to 
[email protected]<mailto:[email protected]>
_______________________________________________
USRP-users mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to