EmbDev.net

Forum: FPGA, VHDL & Verilog Use Xilinx Microblaze performance monitoring engine from AXI4Lite


Author: Giacomo Valente (giav)
Posted on:
Attached files:

Rate this post
0 useful
not useful
Hi,

I would like to use the MicroBlaze performance monitoring engine to 
collect some information on my application code running on the processor 
itself. I would need these information at application level. I have a 
problem to access these values.

In order to do these things, I have created a project with MicroBlaze 
and extended debug function, with 5 event counters and 1 latency 
counter.

Then I have inserted a MDM in my design and created an AXI4lite slave 
port to access from MicroBlaze itself. In this mode I should be able to 
access to internal performance monitoring engine by using DBG_CTRL and 
DBG_DATA registers indicated to page 12 of pg115 (MDM product guide), 
with values indicated on page 92 of MB reference guide (version 
2015.04).

Suppose I need to measure number of any valid executed instruction, I 
have wrote the code attached to this post.

The problem happens when I perform the read from *dbg_data: the read 
action on this variable does not give any response.

The register pointed by variable is R/W.

Do you have any suggestion about this wrong behaviour?


Thank you.

Author: Duke Scarring (Guest)
Posted on:

Rate this post
0 useful
not useful
Try to define your *dbg_data as volatile.

Duke

Author: Giacomo Valente (giav)
Posted on:
Attached files:

Rate this post
0 useful
not useful
Hi,
thank you for the suggestion. I have tried and the behaviour is the 
same.
The problem is the AXI engine under the AXI4slave peripheral (in this 
case, the microblaze debug module): it does not answer to a request.
In the previous post, I have indicated that the problem is the read 
action from *dbg_data. By making a simulation, it appears that the 
problem also happens during an AXI write on *dbg_data memory area.
More specifically, considering attached file, and focusing on MDM 
section, it can be noted that a write on address 41400010 (the content 
of dbg_ctrl variable) happens correctly, because there is a READY 
response from slave to a VALID transition.
Then, there is a write on address 41400014 (the content of dbg_data 
variable): there is no READY response.
The above first two writes are related to the first lines of the code:
    *dbg_ctrl=0x4a404;
    *dbg_data=1;

This causes a block in the application execution, because following AXI 
guidelines, if a master issues a request to write and raises up the 
VALID signal, it must wait the READY response from slave. So without 
READY answering, the master is blocked.

Do you have other suggestion?

Thank you,
Giacomo

Author: masterLuke (Guest)
Posted on:

Rate this post
0 useful
not useful
In case someone is interested in a solution, just like me looking 
through some old posts trying to fix my bugs...

you need to unlock the register access by writing the "magic value" into 
the lock register (adress: 0x41400018). according to the manual (section 
debug register access sequence of MDM manual)this has to be the value 
0xebab.
then the access to the dbg_data-register should be granted.

Reply

Entering an e-mail address is optional. If you want to receive reply notifications by e-mail, please log in.

Rules — please read before posting

  • Post long source code as attachment, not in the text
  • Posting advertisements is forbidden.

Formatting options

  • [c]C code[/c]
  • [avrasm]AVR assembler code[/avrasm]
  • [vhdl]VHDL code[/vhdl]
  • [code]code in other languages, ASCII drawings[/code]
  • [math]formula (LaTeX syntax)[/math]




Bild automatisch verkleinern, falls nötig
Note: the original post is older than 6 months. Please don't ask any new questions in this thread, but start a new one.