Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Fixes for Training 4 for 2025.1 #95

Open
wants to merge 4 commits into
base: main
Choose a base branch
from
Open

Fixes for Training 4 for 2025.1 #95

wants to merge 4 commits into from

Conversation

siupakmok
Copy link
Collaborator

@siupakmok siupakmok commented Mar 28, 2025

Made the following changes:

  • Included setup and teardown time in main.non-blocking.cpp
  • Added -O3 in Makefile.user so that the default behaviour is similar to what is shown in Training 4 doc
  • Updated the doc so that users will get the same results as the doc

Results from non-blocking with accelerators:
image
which is around 72 ms as stated in the document.

To have a more accurate comparison, the time spent on setting up and tearing down the accelerators should be included.
Training 4 document used -O3 number for a fair comparison between hardware and software.

However, it was omitted by mistake in the source files.
This fixes the optimization level so that users will see similar results to the Training doc.
Instruction on CPU Usage were a bit unclear.
The picture shows 11% usage but if were to follow the instructions,
it would have been 40%.

11% usage is only achieved after setting SOC_POLL_DELAY to 1000 ms.

Also updated:
- the source file for main.non_blocking.cpp to include virtual address and line number.
- minor formatting in Figure 6-23
@@ -2578,6 +2593,8 @@ By combining the 2 functions into one, we achieved the following:
SRCS = main_variations/main.fifo.cpp
```

Remove the `set_parameter SOC_POLL_DELAY 1000` line in `config.tcl`, which was added in the [CPU Usage](#cpu-usage-maincpu_usagecpp) section.
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@siupakmok why do we need to remove this setting ?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This setting changes the polling interval to 1 sec, but the program only runs for ~250 ms.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I see.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

SOC_POLL_DELAY is in microseconds, so that's 1ms.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

3 participants