This PR adds a BF16 data-type, along with FP32 <-> BF16 conversion instructions in the LLVM codegen. Other kinds of ops on bfloat16 are not yet supported.
Improved codegen for the Ampere GPUs.
* Make the layout pass recognize the multistage pipelined pattern.
* Now the pipeline pass can automate the multistage pipelining transformation.
* Remove extra barriers (from the prefetch pass & WAR) on Ampere.
* Update the code generator (generator.cc) to make Triton generate n-buffered shared memory loads/stores.
This PR implements a major overhaul of the frontend for Triton, and replaces Triton-C by a pure Python API in which kernels are defined as @triton.jit decorated functions. The documentation and tutorials have also been updated to accommodate these changes.
See documentations for more information on the new API
Recently there has been more and more report about installation issues:
- Installing Triton before upgrading pytorch can create some issues because Triton uses some torch headers
- llvm-10-dev not available on some platform; llvm-11-dev not available on e.g. Ubuntu.
absence of nightly builds
This PR should fix all these issues. Some CMake tricks are used to download and install llvm at build time. Triton Python bindings were modified to remove dependence on pytorch ops. Midnight CI job added to generate binary wheels for all Triton version and update them on pypi's new triton-nightly project.
This PR will also make it very easy to use LLVM forks in the future for whatever needs we have.
Before this commit, the benchmarking infrastructure used heterogeneous protocols between library (e.g., CUTLASS uses a C++ binary that reports mean TFLOPS; torch and triton use python call and report 10th, 50th and 90th quantiles). For the sake of uniformity and fair benchmark practices, this PR adds a python wrapper for auto-tuned CUTLASS matrix multiplication. Benchmarks have been rewritten to use this wrapper with `triton.testing.do_bench` rather than system calls to CUTLASS profiler. Importantly, this also ensures that all the matmuls are done on the *same* input data which should stabilize clock across providers.
This PR adds an automatic memory alignment mechanism in the Triton runtime. Specifically, the JIT compiler detects the alignment (in bytes) of each pointer argument as well as the largest power of two divisor (between 1 and 16) of each integer argument. Proper .aligned and .multipleof attributes are then added to the Triton-IR on-the-fly for all auto-tunable kernels. There is a cache that remembers all the kernels compiled for each possible configuration.
This PR also includes substantial cleaning of the Python API. This adds 2-3us overhead, mostly due to accessing integer #defines from the auto-tuned compilation options. The previous solution was slightly faster but hacky and potentially unsafe, so this is preferred for now.
* Simplified `triton.kernel` API to achieve lower latency:
> .data_ptr() must now be passed as kernel argument. No more implicit
conversion from torch.tensor
> compilation options are now constant attributes, i.e., opt.d('VAR')
becomes opt.VAR
> torch.device must now be passed explicitly to triton.kernel (no
longer inferred from torch.tensor arguments)
* C++ tests moved to `python/tests/`
* C++ tutorial created in `tutorials/`
* Python tutorial created in python/tutorials/
* Version changed to 1.0alpha
* No longer copying C++ headers into the Python package
* added python/triton/ops/ package for pre-written Triton ops
- A100 support via mma.16816
- Thread swizzling for conflict-free shared memory accesses without
padding
- Complete overhaul of the LLVM code generation in
codegen/selection/generator.cc to remove overengineering
- Added debugging capabilities in the Python binding
- Compilation error for kernels that spill
