* Several NFC changes to signatures/includes.

* Deletes prior code generator from previous attempt (moved some of it into this one).
* Renames...

Register the following for the multiply op:
- tcf.mul
- tcp.mul
- TCP->TCP lowering
- Shape tran...

* Two op interfaces, one for querying instance metadata and one for getting static data needed t...

Fix linker error:

lib/Python/libNPCOMPPythonCommon.a(MlirInit.cpp.o): in function `mlir::npcomp...

* Still some more work to do on the Transforms tree to bring it in line with the others (will do...

* For tests with large constants, decreases issues with lit pipelines.
* Bumps llvm-project to p...

It now subsumes the one we had.

Date: Mon Oct 19 15:59:03 2020 -0700

* Enables the conv2d fwd test and ResA (which are both small).
* Deletes resnet18 and vgg, which...

* Adds Basicpy List, Tuple, Dict types and plumbs through C API.
* Started debugging the issues ...

The add-path command is deprecated and should be replaced as stated in
https://github.blog/chang...

* Now gets far enough to capture batch_norm.
* Has some issues still with in-place ops.
* Can ma...

Otherwise, on my machine MLIR somehow gets configured with Python 2,
which is not supported.

Th...

npcomp-capi-ir-test was missing, causing the tests to fail if it wasn't
built. Just build everyt...

The time has come for BypassShapes/LowerShapedResultsToMemref to go away :(
For the reference ba...

Now that it has grown source/target materialization capabilities
(spelled with ops tensor_load/t...

Also, remove the PYTHONPATH stuff for now, I think it is out of date and
I haven't been able to ...

Date: Wed Oct 14 11:26:22 2020 -0700

* Sadly, non-trivial ones fail.
* Bugs filed and marked XFAIL.

* Adds a trampoline/loader 'torch_mlir' module.
* Plumbs through the MLIR python Context and Mod...

* Also adds two lit tests to verify that all of our extensions load without fireworks, which is ...

Date: Mon Oct 12 11:26:50 2020 -0700

Now the reference backend is cleanly accepts "TCP"+scalar ops.

We introduce tcf-refback-lowerin...

Missed this one in the previous change.

* Need to have a dag of shared library deps in order to interop across python extensions (as pre...

I now realize that VerboseCamelCase is not the best choice for dialect
directory/file names and ...

Other than the dialect definitions (which will live in standard Dialect/
subdirectory), the goal...

* Forks a subset of my shell functions into docker_shell_funcs.sh, specifically needed to create...

I missed this in the previous commits.

This is the first in a patch series that is refactoring the
constellation of things variously ca...

NPCOMPFoo to NPCOMPFooDialect for consistency with others.

* Resolves various TODOs that required an LLVM change/bump.
* Bumps LLVM to 4aa217160e5f06a96c6e...

* Minor change of MLIRStandardOps -> MLIRStandard

* Had to stop short of modifying the function return signature because of a missing C-API upstre...

This unblocks backend progress while the PyTorch frontend work is coming
online. Hopefully we ca...

I was seeing some miscompiles due to the uninitialized data read here
before. Interestingly, thi...

* Adds at::Tensor -> MlirValue tracking.
* Adds conversions for tensor and scalar types to MLIR ...

* Uses the MLIR-C API since that will save us a lot of grief down the road (i.e. will give PyTor...

* Does not bump mhlo as an upstream integrate on that project has not taken place and there is n...

* Uses the new dispatcher API.
* Just prints to the console for the moment when an op is capture...

Now that we upstreamed our pass, we can remove it.
The final pass that landed upstream doesn't d...

Date: Fri Sep 18 13:55:52 2020 -0700

- Update to linalg syntax
- New generated builders are b...

This is fairly mechanical.

* Adds a trivial missing test for _torch_mlir.c10.get_registered_ops()
* Disables the regression...

Co-authored-by: Bryce Arden <[email protected]>

* Exposes the op registry via a get_registered_ops method.
* Moves the aten dialect generation ...

This cleans up the lowering pipeline to easily allow extending to
multiple binary ops. It looks ...

This also required adding a lowering for ForOp in our tensor->memref
conversion.

* Still need to add a systematic mechanism for discovering gradient ops.
* Work needed on the va...

It was previously going through this awkward route that prematurely
created linalg.generic ops, ...

- new build directory layout
- build NPCOMPNativePyExt, now that lit tests use it

I'm pretty happy with how this turned out. It looks pretty much like it
should -- one change at ...

I was seeing some of the error messages come out with some garbage at
the end. This fixes it.

Date: Thu Sep 17 18:16:41 2020 -0700

* This gets the pytorch frontend broadly working and what is left appears to be legitimate failu...

* Includes pybind11 directly (for some reason using the pytorch helper header for this depends o...

This now gets the overall "RefE2E" compilation stack to a point that I'm
fairly happy with. We ...

* Add a new python script to auto-generate ATen op ODS definitions.

* There is still some wor...

* Bump LLVM to llvm/llvm-project@7d1ed69
* Bump MLIR-HLO to tensorflow/mlir-hlo@1880f87
* Adop...

That commit is a Thu Sep 10 22:04:58 2020 -0700

That change is required for a PR that I'm going...

* mlir-hlo: 062a3ac4a0671d15b5199ed2cd3a9ce02a5bf077

Fixes:

* numInputs() just returns an int ...

* llvm-project: b5924a8e27536d19dd5c4d302db29fb6163d5faa
* mhlo: 848ca244d20f045b7921da55a98a04d...

* Previous instructions were referring to the option to use an external llvm-project checkout wi...

* This extracts metadata from python invocations (nearly) sufficient to generate ODS and a Torch...

* Make code that depends on the legacy "type dispatch" mechanism optional.

* This code is fai...

This patch adds a pytorch interface to npcomp. This interface is modeled
after pytorch_xla and...

* Was causing build break on GCC9.

* Adds/updates readmes with some notes about code organization and direction.
* Meant to prepar...

* I expect that mlir-hlo will be a non-optional dependency of the project, so adding as a sub-mo...

This patch adds a dialect intended to be used as a frontend dialect
to facilitate lowering from...

The RuntimeShlib.so now lives in /lib.

This debug rule only works with add_mlir_library.