Skip to content

AIESW-6823, AIESW-6824: Adding cmd-chain recipe for latency and throu…#34

Merged
AShivangi merged 4 commits into
Xilinx:mainfrom
sandilya-xilinx:cmd_chain
Dec 16, 2025
Merged

AIESW-6823, AIESW-6824: Adding cmd-chain recipe for latency and throu…#34
AShivangi merged 4 commits into
Xilinx:mainfrom
sandilya-xilinx:cmd_chain

Conversation

@sandilya-xilinx
Copy link
Copy Markdown
Contributor

…ghput

Processing 1 folder(s)...
Updating xrt_smi_npu3.a from 19 files...
✓ Updated: npu3\xrt_smi_npu3.a
➕ New files added (4):
+ profile_cmd_chain_latency.json
+ profile_cmd_chain_throughput.json
+ recipe_cmd_chain_latency.json
+ recipe_cmd_chain_throughput.json

============================================================ SUMMARY:
Archives processed: 1/1
Total new files: 4
Total updated files: 0

Archive changes:
xrt_smi_npu3.a: 4 new

@aktondak
Copy link
Copy Markdown
Contributor

@sandilya-xilinx
Copy link
Copy Markdown
Contributor Author

@sandilya-xilinx Please make use of the depth and mode fields in profile.json instead of copying each kernel multiple times https://github.com/Xilinx/XRT/blob/master/src/runtime_src/core/common/runner/profile.md#:~:text=%22mode%22%3A%20mode%20%20%20%20%20%20%20%20%20%20%20//%20latency%2C%20throughput%2C%20or%20validate%0A%20%20%20%20%22depth%22%3A%20depth%20%20%20%20%20%20%20%20%20//%20clone%20the%20recipe%20runlist

Yes, sure. I am testing this and will update accordingly. Until then this is provided only for reference if anyone wants to debug.
My tests shows that if the depth is less than 6 (<=5) then its not changing and instead it is sending as individual commands. I will test further then update. Until then its labelled as Do Not Merge.

Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

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

Pull Request Overview

This PR adds command-chain profiling recipes for measuring latency and throughput on NPU3 devices. The changes introduce configuration files that define execution parameters for benchmarking command-chain operations.

  • Adds two new recipe directories for cmd_chain_latency and cmd_chain_throughput
  • Configures both recipes to run 10 iterations in non-verbose mode

Reviewed Changes

Copilot reviewed 2 out of 5 changed files in this pull request and generated no comments.

File Description
archive/npu3/cmd_chain_throughput/profile_cmd_chain_throughput.json Configuration for throughput profiling with 10 iterations
archive/npu3/cmd_chain_latency/profile_cmd_chain_latency.json Configuration for latency profiling with 10 iterations

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

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

Pull request overview

Copilot reviewed 2 out of 5 changed files in this pull request and generated no new comments.


💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

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

Pull request overview

Copilot reviewed 4 out of 5 changed files in this pull request and generated 1 comment.


💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comment on lines +16 to +54
"name": "k1",
"arguments": [],
"constants": []
},

{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
Copy link

Copilot AI Dec 16, 2025

Choose a reason for hiding this comment

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

Inconsistent indentation detected. The objects in the runs array use 12 spaces for indentation while the rest of the file uses 2 spaces. Standardize to 2 spaces for consistency with the file structure.

Suggested change
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []
},
{
"name": "k1",
"arguments": [],
"constants": []

Copilot uses AI. Check for mistakes.
…ghput

Processing 1 folder(s)...
Updating xrt_smi_npu3.a from 19 files...
✓ Updated: npu3\xrt_smi_npu3.a
  ➕ New files added (4):
    + profile_cmd_chain_latency.json
    + profile_cmd_chain_throughput.json
    + recipe_cmd_chain_latency.json
    + recipe_cmd_chain_throughput.json

============================================================
SUMMARY:
Archives processed: 1/1
Total new files: 4
Total updated files: 0

Archive changes:
  xrt_smi_npu3.a: 4 new
Copy link
Copy Markdown
Contributor

Copilot AI left a comment

Choose a reason for hiding this comment

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

Pull request overview

Copilot reviewed 4 out of 5 changed files in this pull request and generated 2 comments.


💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Comment thread archive/npu3/cmd_chain_latency/profile_cmd_chain_latency.json Outdated
Comment thread archive/npu3/cmd_chain_throughput/profile_cmd_chain_throughput.json Outdated
sandilya-xilinx and others added 2 commits December 16, 2025 23:20
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
….json

Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
@sandilya-xilinx
Copy link
Copy Markdown
Contributor Author

@aktondak @AShivangi can you review and merge

@AShivangi
Copy link
Copy Markdown
Contributor

@sandilya-xilinx,
The .a file size looks very small. Can you make sure it has all the other files that we have merged so far?

@sandilya-xilinx
Copy link
Copy Markdown
Contributor Author

@sandilya-xilinx, The .a file size looks very small. Can you make sure it has all the other files that we have merged so far?

Good catch @AShivangi, i have verified this is because i have deleted the existing archive file and generated fresh this removed memtile and noop preemption from archive, but the older archive files continue to carry the removed preemption data. I have updated the archive with incremental update

@AShivangi AShivangi merged commit f2e0557 into Xilinx:main Dec 16, 2025
5 checks passed
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.

4 participants