Skip to content

Triton compiler#135

Open
uzairnawaz wants to merge 25 commits intomainfrom
triton-compiler
Open

Triton compiler#135
uzairnawaz wants to merge 25 commits intomainfrom
triton-compiler

Conversation

@uzairnawaz
Copy link

@uzairnawaz uzairnawaz commented Sep 11, 2025

Adds compiler integration

  • gen_kernel_module.py:build generates the source for a kernel module that can perform a forward pass on a nn
  • compile_test.py is a script to verify that torch models give the same output as their compiled kernel modules

Uzair Nawaz and others added 12 commits March 31, 2025 10:48
@uzairnawaz uzairnawaz requested a review from a team as a code owner September 11, 2025 01:01
Uzair Nawaz added 13 commits September 11, 2025 17:56
linter fix
faa7bc1
seperate model input vs output size
c7ed9b7
increased primal precision
9af291a
cleaned up scripts; directly store output size as constant in generat…
abb15fe 
…ed module
begin adding support for addmm external kernel (not working yet)
385d570
debugging; minimum failing addmm kernel call
3843e67
debugging; minimum failing addmm kernel call
f4bd8d4
explicitly tell kbuild to allow sse so that it passes float args thro…
53691a6 
…ugh regs instead of the stack
successful with larger neural nets
6e23354
created python module, model_compile returns list of generated files
309f344
module_test.py works as part of module; user_kernel_comparison benchm…
0d57105 
…ark written
dont need sudo to build user space module for benchmark
731d5d2
improved benchmarks; fixed memory bug in generated code
4de03d7
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

No reviews

Assignees

No one assigned

Labels

None yet

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

1 participant

@uzairnawaz