Commit Graph

23 Commits

Author SHA1 Message Date
Gheorghe-Teodor Bercea b28c6906b4
Fix building ONNF with latest LLVM/MLIR (#89)
* Fix build and link errors.

* Fix end to end tests.

* Fix indentation.

* Fix type conversion.

* Use newest LLVM version.

* Use newest LLVM version.
2020-02-19 18:15:02 -05:00
Tung D. Le 2b56c09454
Rewrite ReduceL1, ReduceL2, ReduceLogSum, ReduceLogSumExp, ReduceSumSquare in the ONNX dialect (#38)
* Rewrite ReduceSumSquare

* Edit gen_doc.py

* Revise the code

* Do shape inference after canonicalization so that there is no need to implement shape inference of rewritten ops

* Rewrite ReduceL2

* Add onnx_rewrite.cpp for all rewriting for ONNX ops

* Rewrite ReduceL1, ReduceLogSum, ReduceLogSumExp

* Edit comments

* Change the use of -> to .

* Checkout gen_doc.py from the master branch

* Use emplace_back instead of push_back

* Revise the code

* Edit comments

Co-authored-by: Tian Jin <tjingrant@gmail.com>
2020-01-31 19:00:39 +08:00
Tian Jin 8665ecd998
Enable e2e tests (#29)
* Sync with latest MLIR.

* Enable ONNX backend tests as a means to test ONNF lowering end-to-end.

* Install ONNX using quiet mode.

* Remove debug comments.

* Install ONNX from third_party/onnx.

* Check python version and fix pip command for installing ONNX.

* Using --user install option to prevent permission denied.

* Remove unused imports.

* Try using stock ONNX pip package as there are more tests in them.

* Pip got stuck building wheels, try sudo.

* Use verbose install to debug.

* Invalidate cache to build LLVM tools.

* Fix mlir installation script location.

* Debug to locate ONNF.

* Sanity check.

* Check out ONNF code first.

* Use verbose LIT output.

* 1. Update documentation to always use verbose LIT.
2. Update krnl ops to reflect new affine map attribute syntax.

* See if conda exists

* Install ONNX by manually cloning the repo.

* Install cmake first.

* Using sudo priviledge when installing.

* Limit build parallelism.

* Limit parallelism.

* Larger memory.

* Install onnx package with pip.

* Build MLIR tools.

* Invalidate cache.

* Compile model.so with -fPIC.

* Remove module dump to get concise debug output.

* Print command before executing.

* Use quiet install mode to reduce logging.

* Use -relocation-model=pic to generate position independent code.

* 1. Remove MAKEFLAGS because now buildbot has enough memory.
2. Run DocCheck as a last step.

* 1. Add verbose mode for backtend test.

* When dumping to LLVM bitcode, do not dump module IR, but print a message indicating that bitcode has been written to disk.

* Do not pass MakeFlags to CMake.

* Add more explaination for posible reasons of failing to identify tests.
2020-01-20 12:30:08 -05:00
Tian Jin 238c937f1b rewrite cli description 2019-12-23 01:14:35 -05:00
Tian Jin da4527c961 flatten src directory structure 2019-12-23 00:13:52 -05:00
Tian Jin 82d513096e a commandline interface for onnf 2019-12-22 23:52:49 -05:00
Tian Jin 911cc2ad92
Merge pull request #1 from doru1004/remove-boost
clean up, remove dependency for boost
2019-12-22 23:15:46 -05:00
Tian Jin 685bf23b40 Enable ONNX Backend Test (#1)
* wip, commit before merging with upstream

* organize API, return wrapped output

* enable onnx backend test

* undo unintentional commit

* fix krnl ops tablegen

* format krnl ops

* reorder fillDynMemRefWithMemRef to be after fillPtrToMemRefWithDynMemRef, better comments

* more onnx backend tests

* ensure that test names refer to existing tests

* improve code readability by shortening type names

* nit

* restore unintentional changes

* more nits

* fix ; -> :

* split runtime implementation into header and body file, add support for data types

* comment on the onnx backend test

* make the comments read better

* do not dump when lowering
2019-12-22 23:14:57 -05:00
Tian Jin 2cb054324d clean up, remove dependency for boost 2019-12-22 20:49:29 -05:00
Tian Jin 5573cb39fe clean up, remove dependency for boost 2019-12-22 20:33:33 -05:00
Tian Jin a6a40cf989 Format Key Files using LLVM Style (#403)
* format using llvm style

* merge and format
2019-12-21 02:11:49 -05:00
GHEORGHE-TEOD BERCEA e81a7654f9 [MLIR] Add support for reshape (#390)
* Add reshape op handling.

* Lower reshape to KRNL dialect.

* Add comments.

* Propagate reshape to KRNL IR.

* Lower KRNL reshape to affine and standard ops level dialects.

* Add lowering of reshape operation to Krnl and LLVM Dialects.

* Add test for LLVM IR dialect output for reshape.

* Fix rebase.

* Fix test variable.

* Emit errors during reshape shape inference. Address other reviewer comments.
2019-12-21 02:06:14 -05:00
Tian Jin b2a1103915 [MLIR] Refactor Krnl Dialect and Krnl Dialect Lowering (#375)
* Store bounds as affine map attributes & check in test cases with generic printer

* Upgrading MLIR

MLIR is outdated on Buildbot, rebuilding a newer version.

* work with new version of mlir

* check-in parser tests

* custom printer

* nit

* bug fix

* enable custom asm printer test

* enable custom asm printer test

* more consistent variable naming

* test max/min

* variable naming scheme change to MLIR style

* can lower krnl to llvm

* kernel -> llvm

* comments

* bug fix

* try fixing ci

* fix ci

* deactivate model test

* fix lit test

* nit

* fix z buildbot
2019-12-21 01:34:14 -05:00
GHEORGHE-TEOD BERCEA b02652dd76 [MLIR] Lowering of frontend dialect to KRNL dialect (#382)
* Partial support for lowering operations to KRNL dialect.

* Attempt to lower to KRNL IR.

* Update file.

* Add lowering.

* Address comments. Fix alloc dynamic dimensions. Correctly link StandardOps.

* Temporarily remove deallocation of locally allocated tensors.
2019-12-21 01:11:14 -05:00
Tian Jin d01ac7732f [MLIR] compartmentalize build script (#369)
* compartmentalize build script, temporarily remove dependency of onnf_opt on helper.cpp

* fix test includes

* fix op directory include

* compiler -> op

* compiler test depends on boost system

* fix function name

* specify libcompiler dependencies

* let cmake take care of transitive dependencies

* remove unnecessary includes

* use ONNF_SRC_ROOT and ONNF_BIN_ROOT

* allow whole-archive linked libraries to be appended

* [MLIR] Support filecheck (#371)

* support lit+FileCheck

* add lit into build script

* format MLIR.cmake

* format cmake

* [MLIR] Remove input/output ops (#372)

* remove input/output ops

* get output tensor type from symbol table
2019-12-21 00:34:51 -05:00
GHEORGHE-TEOD BERCEA dc36fd416b [MLIR] Fix shape inference and enable ONNF to read in MLIR files. (#367)
* Fix inference. Enable ONNF to read in MLIR files.

* Fix input of onnx or mlir models.

* Address comments.
2019-12-21 00:12:27 -05:00
GHEORGHE-TEOD BERCEA 63596e723f [MLIR] Add optimization based on TableGen pattern (#363)
* Define pattern.

* Fix file names.

* Add canonicalizer optimization based on TableGen pattern.

* Remove custom builders.

* Enable canonicalization in ONNF and ONNF-OPT.
2019-12-21 00:04:51 -05:00
Tian Jin 780e6f0aa0 [MLIR] Krnl Dialect Definition (#357)
* a complete, roud-trippable Krnl dialect operation definition

* remove old dialect definition files, edit build files

* register dialect

* check in src for onnf_opt and dimension handler types

* re-trigger jenkins

* fix build

* clarify operation semantics

* add verifier for krnl.iterate

* refactor to make things clear

* do not hard code types

* nit and add comments

* fix rebase

* update op implementation

* fix merge

* update kernel dialect definition

* more comment on how to use the builder for krnl.iterate operation

* ammend the comment

* can parse krnl.iterate

* can parse and print if bounds are not SSA values

* address comments

* better error handling

* Update CMakeLists.txt

* update comment

* reflow comments
2019-12-20 11:37:09 -05:00
GHEORGHE-TEOD BERCEA 03be41f7df [MLIR] Add shape inference pass (#361)
* Add shape inference pass.

* Format code.

* Enable shape inference pass.

* Rename new files to use ONNF convention.

* Use include path from src folder.

* Replace guards with pragma once.

* Format variable names. Shuffle headers.

* Fix comment.

* Fix comments.
2019-12-19 20:14:07 -05:00
GHEORGHE-TEOD BERCEA 958a2fbffc ONNX Dialect creation and registration with MLIR (#358)
* Create and register ONNX Dialect with one Add operation.

* Fix file formatting.

* Change name from ONNX to SGIR.

* Use ONNX dialect. Change SGIR to frontend placeholder dialect.

* Add comments.

* Type clean-up.
2019-12-19 16:56:30 -05:00
GHEORGHE-TEOD BERCEA 45a5a8499c Fix MLIR-enabled build. (#354)
* Add new build paths. Add Ranker or Unracked tensor types.

* Remove path to LLVM root folder, it is not needed anymore.

* Code clean-up.

* Clean-up code.

* update LLVM_PROJECT_SRC

* Update p9.sh

* Update MLIR.cmake
2019-12-19 16:16:06 -05:00
Tian Jin 00aee2e0b6 Use TableGen (#347)
* use table gen

* fix name of the dialect

* add old compilation path

* add some doc

* fix bug, sgir importer imports every op twice

* knl.visit -> knl.iterate
2019-12-19 15:18:35 -05:00
Doru Bercea 63427030ca Enable building of first commit. 2019-12-19 14:24:37 -05:00