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

NTT Evaulation #962

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

NTT Evaulation #962

wants to merge 4 commits into from

Conversation

sequencer
Copy link
Member

  • [eval] add build
  • [eval] add eval.ntt*

@sequencer
Copy link
Member Author

@Avimitin Please add CI to run all evaluation cases under benchmark_dlen*_vlen* for performance benchmark.

Copy link
Member Author

@sequencer sequencer left a comment

Choose a reason for hiding this comment

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

I think we should do some thing like this:

  • For ntt_*_main.c it should detect the VLEN at runtime and delegate to different algorithm.
  • For static const int arr[xxx] and its layers, a script should be added to generate those these vector(maybe from ref.ws?) rather than hardcode.
  • Neither do I like gen_data.py, it should also be rewritten into ws for simplicity.
  • How does twiddle get calculated? From gen_vector_ntt_tw.py? I think this python should be rewrite into ws as a function, and codegen corresponding headers.

Please clarify the build stage flows:

  1. Put all user inputs(maybe arr?) into a csv or json file.
  2. At the compile time, a ws should read the csv and generate data and twiddle, the output should be store as json and csv
  3. A codegen python/Scala should read output from ws and generate a corresponding C header file.
  4. The test entry should include the C header file and execute to get the result.

Some random thing I can think about:

  • How do software programmer toggle performance counters though C-API?
  • What's the SW engineer API for benchmark? I think we may need a generic post-evaluation flow for the benchmark.

Comment on lines +5 to +11
9997, 6362, 7134, 11711, 5849, 9491, 5972, 4164, 5894, 11069, 7697,
8319, 2077, 12086, 10239, 5394, 4898, 1370, 1205, 2997, 5274,
4625, 11983, 1789, 3645, 7666, 12128, 10883, 7376, 8883, 2321,
1889, 2026, 8059, 2741, 865, 1785, 9955, 2395, 9330, 11465, 7383,
9649, 11285, 3647, 578, 1158, 9936, 12019, 11114, 7894, 4832,
10148, 10363, 11388, 9122, 10758, 2642, 4171, 10586, 1194, 5280,
3055, 9220
Copy link
Member Author

Choose a reason for hiding this comment

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

Q: How to choose these numbers? Is this configurable?

Comment on lines 97 to 104
// for (int i = 0; i < n; i++) {
// printf("%d", dst[i]);
// if ((i + 1) % 8 == 0) {
// printf("\n");
// } else {
// printf(" ");
// }
// }
Copy link
Member Author

Choose a reason for hiding this comment

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

Don't include comment here, you can add #IFDEF DEBUG macro for these codes.

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