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

[Mem Modeling] Todos to get memory modeling automatic generation #72

Open
weiya711 opened this issue Jan 13, 2023 · 1 comment
Open
Assignees

Comments

@weiya711
Copy link
Owner

Things needed for memory modeling automatic generation of tests
branch name: tiling_loop_generation
file name: scripts/test_generating_code.py

TODO: @Ritvik1sharma needs to split this into test_generating_code_mem.py and test_generating_code_ideal.py. Also test_generating_code_mem.py should call test_generating_code_ideal.py to generate the inner-most loop.

Needed from pre-tiling scripts

  1. Generate format arrays (seg/crd for sparse, dim for dense) for EACH level of the memory hierarchy. Must follow the same naming convention as @Ritvik1sharma's convention
  2. Change naming convention nested file structure for pre-tiled tiles. I.e. i00_k00_i0_k0 --> i00_k00/i0_k0
  3. Make sure to give crd array for tile file names that goes from ref --> crd tuple

Compiler

  1. Name of expressions/test
  2. Input tensor list and result tensor at the top of the SAM graph (where formats are given). I.e. in_tensors="B(i,k),C(k,j)" and out_tensors="X(i,j)"

For next steps/future work with different tiling index reorderings:
Either need 1) a database of all "input iteration" permutations for orders 1, 2, 3, ...
or 2) Need the compiler to generate the "input iteration" graph for each level that's a different index ordering

@weiya711 weiya711 changed the title [Mem Modeling] [Mem Modeling] Todos to get memory modeling automatic generation Jan 13, 2023
@weiya711
Copy link
Owner Author

Naming convention is Memory level 2: "00", Memory level 1: "0", Memory level 0: ""

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

No branches or pull requests

2 participants