Commit Graph

7 Commits

Author SHA1 Message Date
Philippe Tillet
8edfe813a5 [FRONTEND][BACKEND] Added trans instruction; made flash attention bwd pass work (#943) 2022-12-03 09:58:24 -08:00
Keren Zhou
153aecb339 [Triton-MLIR][BACKEND] insert_slice_async on GPUs < sm80 (#908)
`insert_slice_async` is decomposed into `load + insert_slice` in the
backend.

Not sure if V100 perf can match the master branch though in this way.
Maybe the performance can be improved if instructions are arranged in
the following form:

```
%0 = load
%1 = load 
%2 = load 
...
insert_slice %0
insert_slice %1
insert_slice %2
```

Tested on A100 when manually enabling this decomposition.
Tests on V100 haven't been integrated yet, we can divide the tests into
two phases:
1. Test only load, insert_slice, and insert_slice_async, given TritonGPU
IRs in `test_backend.py`.
2. End to end gemm tests on V100.
2022-11-24 14:05:54 -08:00
Chenggang Zhao
516a241234 [Triton-MLIR] Fix some typos (#874)
Fix some typos
2022-11-13 18:15:53 -08:00
Keren Zhou
fdd59900f7 [Triton-MLIR] Replace triton.extract_slice with tensor.extract_slice and support more general tensor slicing (#837)
## Features

- Allow taking a block of tensor slice, as long as each dimension is
contiguous (unit stride).
- Fix some problems in `insert_slice_async`'s semantic.
- More general verification for ops that return shared layout encoding.

## Known Limitations

- `insert_slice_async` still uses the old semantic. May submit another
PR later to support similar semantic like `tensor.extract_slice`.
- No encoding verification for `tensor.extract_slice`.
- 3d tensor ops are broken.
- Strided accesses are not allowed.
- May cause a little performance slowdown since we are passing strides
as values but not constants (e.g., int).
It would be difficult to pass strides as attributes when we have control
flows. A block argument is possible to accept tensors with different
strides.
2022-11-06 22:59:03 -08:00
Keren Zhou
16aed94ff5 [Analysis/Allocation] Allocation passes now assumes that slices always alias (#108)
This code in this branch assumes the `src` operand in
`insert_slice_async` always aliases the result, which shouldn't hold for
generally cases but is just a workaround to make the pipeline pass work.

I'm also working on the complete analysis in another
[branch](https://github.com/openai/triton-mlir/tree/keren/analyze-slice).
2022-09-09 12:03:41 -07:00
Yan Chunwei
a9464f4993 [Backend] Vectorize Load/Store Ops (#86)
This PR does the following things:

- Code refactoring on Load and Store op codegen, rewrite with same logic
and share much code
- Support the vectorized load/store
2022-09-06 12:28:09 -07:00
Keren Zhou
02ebf24d35 Analyze shared memory alias (#81)
The purpose of this PR is analyzing shared memory aliases so that we can
fix memory allocation bugs and save memory allocations in triton code
involving complex control flows.

Changes to memory bar and allocation are on the way.

Co-authored-by: Philippe Tillet <phil@openai.com>
2022-08-29 10:43:20 -07:00