This PR
- Fix numWarps>1 hang issue
- add existing test cases in test_gemm.py to CI, and add a common flag
`valid_on_Volta` to determine whether the test case should be activated
on Volta or just skip.
- Currently, the column-major cases are disabled.
- Add test_core.py and other tests to Volta CI
- the `test_printf.py` failed.
Most notably, this PR:
- changes the traits (and assembly format) of addptr so it can handle offsets that have arbitrary integer width.
- adds support for `cat`
TODO:
- Add more cases
- Currently, we just set vec to 4 to make the basic cases pass
Issue:
- the vec in shared layout is different compared to master branch
- when vec=1, it encounters CUDA misalignment error, it doesn't work in
master branch as well
- when setting vec to the value identical to master branch, the MMA
works
This change attaches builtin-ness as an explicit attribute, rather than
a module prefix expectation. This permits us to source those builtins
from multiple sub-modules (useful when some builtins are part of the
true cyclic implementation core, and some are just useful library
additions); but also prevents accidental inclusion of non-builtins that
happen to be in the right library.
Once the flag exists, and the compiler is using `is_builtin()` for
decision making; the existence of the current `@extern` interface
becomes isomorphic to `@builtin`; and the interface can be unified.
Leaving `@extern` a thin-wrapper, and encouraging continued use of it,
establishes future-proofing towards adding additional extern tracing,
metric hooks, or scanning in the future.
* Add `triton.impl` package to hold the core, order dependent impl
details.
* Extract `@builtin` and unify `@extern`; add `is_builtin()`
* Add sense bit so that `@builtin` detection is less fragile.
* Modify the compiler to use `is_builtin()`
Changes to make decorated API methods no longer type-opaque.
```
$ echo 'import triton; reveal_type(triton.language.max)' | mypy /dev/stdin
/dev/stdin:1: note: Revealed type is "def (input: Any, axis: Any, _builder: Any =) -> Any"
Success: no issues found in 1 source file
```
Expand `from .foo import *` to full listings, and `__all__` sections.
This reifies the module export listings, which is useful for code
importing this module; without this, clients will need special `mypy`
control pragmas for this library.
This removes a number of `# flake8` control pragmas.
Verified with `flake8`
1. Add missing barriers and revert the previous temporary solution
2. Extract the `run` method from membar analysis because the membar
analysis should have two phases, including construction, which doesn't
modify any IR, and modification, which adds barrier IRs. Hope this could
make the use of membar clear.
1. Improve pipline's comment
2. Decompose insert_slice_async when load vector size is not supported
3. Add a test that could fail our gemm code
Copy my comments here:
There's a knob that may cause performance regression when decomposition
has been performed. We should remove this knob once we have thorough
analysis on async wait. Currently, we decompose `insert_slice_async`
into `load` and `insert_slice` without knowing which `async_wait` is
responsible for the `insert_slice_async`. To guarantee correctness, we
blindly set the `async_wait` to wait for all async ops if any `insert_slice_async` has been decomposed.
There are two options to improve this:
1. We can perform a dataflow analysis to find the `async_wait` that is
responsible for the `insert_slice_async` in the backend.
4. We can modify the pipeline to perform the decomposition before the
`async_wait` is inserted. However, it is also risky because we don't
know the correct vectorized shape yet in the pipeline pass. Making the
pipeline pass aware of the vectorization could introduce additional
dependencies on the AxisInfoAnalysis and the Coalesce analysis.
This PR
- apply minimal modification to decouple the Dot helper related code
from TritonGPUToLLVM.cpp to a separate local header file to make it
easier to share some data structure for Dot
- add some patch necessary for transA and transB
- add some patch necessary for MMA v1 execution in backend
`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.
Try to add proper side effects for triton operations.
The CSE pass could fail, hang, or output incorrect IRs for unknown
reasons, if side effects are not defined properly.
For instance, suppose we have two shared memory tensors:
```
%a = triton_gpu.alloc_tensor shape0, share_encoding0
%b = triton_gpu.alloc_tensor shape0, share_encoding0
```
The CSE pass will consider `%a` and `%b` are the same thing and
eliminate one of them, resulting in mysterious outcomes.
Cross operation barriers are taken care of by the Membar pass.
Explicit barriers are only required if there's any synchronization
necessary within each operation.
