colin
/
mlir-hlo
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
mlir-hlo/BUILD

1431 lines
40 KiB
Python
Raw Normal View History

[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
load("@llvm-project//mlir:tblgen.bzl", "gentbl_cc_library", "gentbl_filegroup", "td_library")
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
package(
default_visibility = ["//visibility:public"],
Remove comments for license declarations. These can get out of date/stale. PiperOrigin-RevId: 365936646
2021-03-31 08:54:53 +08:00
licenses = ["notice"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
exports_files([
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops.td",
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_ops.td",
])
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
td_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "hlo_ops_td_files",
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
srcs = glob(["include/mlir-hlo/Dialect/mhlo/IR/*.td"]) + [
# TODO(gcmn): These should be encapsulate in a td_library.
PR #48667: [mlir-hlo] Added RegionBranchOpInterfaces to lmhlo operations. Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/48667 Added RegionBranchOpInterfaces to lmhlo operations that use regions. This is needed, since the bufferization features in MLIR have to reason about the control flow within these operations. Copybara import of the project: -- 572fd7d850a46630b812da84e9094280f89f259e by Julian Gross <julian.gross@dfki.de>: Added RegionBranchOpInterfaces to lmhlo operations. PiperOrigin-RevId: 372070825
2021-05-05 15:26:46 +08:00
"@llvm-project//mlir:include/mlir/Interfaces/ControlFlowInterfaces.td",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:include/mlir/Interfaces/CopyOpInterface.td",
"@llvm-project//mlir:include/mlir/Interfaces/InferTypeOpInterface.td",
"@llvm-project//mlir:include/mlir/Interfaces/LoopLikeInterface.td",
"@llvm-project//mlir:include/mlir/Interfaces/ViewLikeInterface.td",
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
"@llvm-project//mlir:include/mlir/Dialect/Shape/IR/ShapeBase.td",
"@llvm-project//mlir:include/mlir/Dialect/Shape/IR/ShapeOps.td",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
includes = ["include"],
deps = [
PR #48667: [mlir-hlo] Added RegionBranchOpInterfaces to lmhlo operations. Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/48667 Added RegionBranchOpInterfaces to lmhlo operations that use regions. This is needed, since the bufferization features in MLIR have to reason about the control flow within these operations. Copybara import of the project: -- 572fd7d850a46630b812da84e9094280f89f259e by Julian Gross <julian.gross@dfki.de>: Added RegionBranchOpInterfaces to lmhlo operations. PiperOrigin-RevId: 372070825
2021-05-05 15:26:46 +08:00
"@llvm-project//mlir:ControlFlowInterfacesTdFiles",
"@llvm-project//mlir:LoopLikeInterfaceTdFiles",
Integrate LLVM at llvm/llvm-project@678241795c95 Updates LLVM usage to match [678241795c95](https://github.com/llvm/llvm-project/commit/678241795c95) PiperOrigin-RevId: 363257913
2021-03-17 04:31:59 +08:00
"@llvm-project//mlir:MemRefOpsTdFiles",
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
"@llvm-project//mlir:OpBaseTdFiles",
"@llvm-project//mlir:SideEffectTdFiles",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "MhloPassIncGen",
strip_include_prefix = "include",
tbl_outs = [
(
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
[
"-gen-pass-decls",
"-name=MHLO",
],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include/mlir-hlo/Dialect/mhlo/transforms/mhlo_passes.h.inc",
),
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/transforms/mhlo_passes.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:PassBaseTdFiles",
],
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "LmhloPassIncGen",
strip_include_prefix = "include",
tbl_outs = [
(
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
[
"-gen-pass-decls",
"-name=LMHLO",
],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include/mlir-hlo/Dialect/mhlo/transforms/lmhlo_passes.h.inc",
),
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/transforms/lmhlo_passes.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:PassBaseTdFiles",
],
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "chlo_ops_inc_gen",
strip_include_prefix = "include",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-op-decls"],
"include/mlir-hlo/Dialect/mhlo/IR/chlo_ops.h.inc",
),
(
["-gen-op-defs"],
"include/mlir-hlo/Dialect/mhlo/IR/chlo_ops.cc.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/chlo_ops.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "hlo_ops_inc_gen",
strip_include_prefix = "include",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-op-decls"],
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops.h.inc",
),
(
["-gen-op-defs"],
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops.cc.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/hlo_ops.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "hlo_ops_base_inc_gen",
strip_include_prefix = "include",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-op-decls"],
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base.h.inc",
),
(
["-gen-op-defs"],
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base.cc.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "hlo_ops_base_structs_inc_gen",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-struct-attr-decls"],
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base_structs.h.inc",
),
(
["-gen-struct-attr-defs"],
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base_structs.cc.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "hlo_ops_base_enums_inc_gen",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-enum-decls"],
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base_enums.h.inc",
),
(
["-gen-enum-defs"],
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base_enums.cc.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "hlo_ops_pattern_gen",
strip_include_prefix = "lib/Dialect/mhlo/IR/",
tbl_outs = [
(
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
["-gen-rewriters"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"lib/Dialect/mhlo/IR/hlo_patterns.cc.inc",
),
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "lib/Dialect/mhlo/IR/hlo_patterns.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":hlo_ops_td_files",
"@llvm-project//mlir:StdOpsTdFiles",
Canonicalize dynamic_broadcast_in_dim to own shape with rank narrowing on the shape to a corresponding tensor.cast. PiperOrigin-RevId: 362028291
2021-03-10 21:43:10 +08:00
"@llvm-project//mlir:TensorOpsTdFiles",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
[MLIR:LHLO] Add optional call target arg mapping to LMHLO CustomCall operations. - XLA:HLO -> LMHLO conversion drops all token arguments and return values, however custom calls that users write still expect to get buffer pointers for these token types. - To be able to support this, add an optional call target argument mapping attribute to LMHLO custom calls. When this attribute is present, it indicates the number of arguments and returns that the custom call expects and also indicates which LMHLO arg() or output() maps to which arg or result number of the custom call. PiperOrigin-RevId: 358826664
2021-02-23 00:41:59 +08:00
name = "lhlo_ops_structs_inc_gen",
strip_include_prefix = "include",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-struct-attr-decls"],
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_ops_structs.h.inc",
),
(
["-gen-struct-attr-defs"],
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_ops_structs.cc.inc",
),
[MLIR:LHLO] Add optional call target arg mapping to LMHLO CustomCall operations. - XLA:HLO -> LMHLO conversion drops all token arguments and return values, however custom calls that users write still expect to get buffer pointers for these token types. - To be able to support this, add an optional call target argument mapping attribute to LMHLO custom calls. When this attribute is present, it indicates the number of arguments and returns that the custom call expects and also indicates which LMHLO arg() or output() maps to which arg or result number of the custom call. PiperOrigin-RevId: 358826664
2021-02-23 00:41:59 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/lhlo_ops_structs.td",
td_includes = [
"external/mlir-hlo/include",
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
[MLIR:LHLO] Add optional call target arg mapping to LMHLO CustomCall operations. - XLA:HLO -> LMHLO conversion drops all token arguments and return values, however custom calls that users write still expect to get buffer pointers for these token types. - To be able to support this, add an optional call target argument mapping attribute to LMHLO custom calls. When this attribute is present, it indicates the number of arguments and returns that the custom call expects and also indicates which LMHLO arg() or output() maps to which arg or result number of the custom call. PiperOrigin-RevId: 358826664
2021-02-23 00:41:59 +08:00
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "lhlo_ops_inc_gen",
strip_include_prefix = "include",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-op-decls"],
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_ops.h.inc",
),
(
["-gen-op-defs"],
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_ops.cc.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/lhlo_ops.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "lhlo_gpu_ops_structs_inc_gen",
strip_include_prefix = "include",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-struct-attr-decls"],
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_structs.h.inc",
),
(
["-gen-struct-attr-defs"],
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_structs.cc.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_structs.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "lhlo_gpu_ops_enums_inc_gen",
strip_include_prefix = "include",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-enum-decls"],
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_enums.h.inc",
),
(
["-gen-enum-defs"],
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_enums.cc.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_enums.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
Added build rules for HLO and LHLO ops MD. PiperOrigin-RevId: 373373920
2021-05-12 23:22:41 +08:00
gentbl_filegroup(
name = "hlo_ops_doc_gen",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-dialect-doc"],
"g3doc/hlo_ops.md",
),
Added build rules for HLO and LHLO ops MD. PiperOrigin-RevId: 373373920
2021-05-12 23:22:41 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/hlo_ops.td",
deps = [":hlo_ops_td_files"],
)
gentbl_filegroup(
name = "lhlo_ops_doc_gen",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-dialect-doc"],
"g3doc/lhlo_ops.md",
),
Added build rules for HLO and LHLO ops MD. PiperOrigin-RevId: 373373920
2021-05-12 23:22:41 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/lhlo_ops.td",
deps = [":hlo_ops_td_files"],
)
[HLO] Add LMHLO CollectivePermute verification. - Extract verification of source target pairs attached to collective permute into a common helper function and use that to verify both MHLO and LMHLO variants. - Change MlirGpuTestBase::ParseMlirModule to allow returning back a failure, and use that to update the mlir_gpu_compile_test to check the new behavior. PiperOrigin-RevId: 362156962
2021-03-11 07:36:22 +08:00
cc_library(
name = "hlo_ops_common",
srcs = ["lib/Dialect/mhlo/IR/hlo_ops_common.cc"],
hdrs = ["include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_common.h"],
includes = ["include"],
deps = [
[HLO] Add custom print/parse for window attributes of convolutions (in LMHLO) PiperOrigin-RevId: 373807616
2021-05-15 00:46:42 +08:00
"@llvm-project//llvm:Support",
[HLO] Add LMHLO CollectivePermute verification. - Extract verification of source target pairs attached to collective permute into a common helper function and use that to verify both MHLO and LMHLO variants. - Change MlirGpuTestBase::ParseMlirModule to allow returning back a failure, and use that to update the mlir_gpu_compile_test to check the new behavior. PiperOrigin-RevId: 362156962
2021-03-11 07:36:22 +08:00
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Support",
],
)
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
cc_library(
name = "lhlo_gpu_ops_structs",
srcs = [
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_structs.cc.inc",
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_structs.h.inc",
"lib/Dialect/mhlo/IR/lhlo_gpu_ops_structs.cc",
],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_structs.h",
],
includes = ["include"],
deps = [
":lhlo_gpu_ops_structs_inc_gen",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Support",
],
)
cc_library(
name = "lhlo_gpu_ops_enums",
srcs = [
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_enums.cc.inc",
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_enums.h.inc",
"lib/Dialect/mhlo/IR/lhlo_gpu_ops_enums.cc",
],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops_enums.h",
],
includes = ["include"],
deps = [
":lhlo_gpu_ops_enums_inc_gen",
"@llvm-project//llvm:Support",
],
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "lhlo_gpu_ops_inc_gen",
strip_include_prefix = "include",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-op-decls"],
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops.h.inc",
),
(
["-gen-op-defs"],
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops.cc.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
#TODO(aminim): revisit the naming and grouping of these rules post-move.
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "canonicalize_inc_gen",
strip_include_prefix = "lib/Dialect/mhlo/IR/",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-rewriters"],
"lib/Dialect/mhlo/IR/mhlo_canonicalize.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "lib/Dialect/mhlo/IR/mhlo_canonicalize.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "infer_fusibility_op_interface_gen",
tbl_outs = [
(
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
["-gen-op-interface-decls"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include/mlir-hlo/Dialect/mhlo/IR/infer_fusibility_op_interface.h.inc",
),
(
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
["-gen-op-interface-defs"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include/mlir-hlo/Dialect/mhlo/IR/infer_fusibility_op_interface.cpp.inc",
),
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/infer_fusibility_op_interface.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
cc_library(
name = "infer_fusibility_op_interface",
srcs = [
"lib/Dialect/mhlo/IR/infer_fusibility_op_interface.cc",
],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/IR/infer_fusibility_op_interface.h",
"include/mlir-hlo/Dialect/mhlo/IR/infer_fusibility_op_interface.h.inc",
],
includes = ["include"],
deps = [
":infer_fusibility_op_interface_gen",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Support",
],
alwayslink = 1,
)
cc_library(
name = "hlo_ops_base_structs",
srcs = [
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base_structs.h.inc",
"lib/Dialect/mhlo/IR/hlo_ops_base_structs.cc",
],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base_structs.h",
],
includes = ["include"],
deps = [
":hlo_ops_base_structs_inc_gen",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Support",
],
)
cc_library(
name = "hlo_ops_base_enums",
srcs = [
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base_enums.h.inc",
"lib/Dialect/mhlo/IR/hlo_ops_base_enums.cc",
],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base_enums.h",
],
includes = ["include"],
deps = [
":hlo_ops_base_enums_inc_gen",
"@llvm-project//llvm:Support",
],
)
cc_library(
name = "convert_op_folder",
srcs = ["lib/utils/convert_op_folder.cc"],
hdrs = ["include/mlir-hlo/utils/convert_op_folder.h"],
includes = ["include"],
deps = [
"@llvm-project//mlir:IR",
],
)
cc_library(
name = "hlo",
srcs = [
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops.cc.inc",
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops.h.inc",
"lib/Dialect/mhlo/IR/chlo_ops.cc",
"lib/Dialect/mhlo/IR/hlo_ops.cc",
"lib/utils/broadcast_utils.cc",
"lib/utils/hlo_utils.cc",
],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/IR/chlo_ops.h",
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops.h",
[MLIR][HLO] Annotate `mhlo.clamp` and `mhlo.select` as element-wise broadcasting The operations allow for a limited form of broadcasting which allows some operands to be scalars. As such they are neither strictly `Elementwise`, nor `Broadcasting`. They do fulfill the requirements for `BroadcastingElementwise` though. PiperOrigin-RevId: 379719961
2021-06-16 22:58:09 +08:00
"include/mlir-hlo/Dialect/mhlo/IR/hlo_ops_base.h",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include/mlir-hlo/utils/broadcast_utils.h",
"include/mlir-hlo/utils/hlo_utils.h",
],
includes = ["include"],
deps = [
":canonicalize_inc_gen",
":chlo_ops_inc_gen",
":convert_op_folder",
":hlo_ops_base_enums",
":hlo_ops_base_inc_gen",
":hlo_ops_base_structs",
[HLO] Add LMHLO CollectivePermute verification. - Extract verification of source target pairs attached to collective permute into a common helper function and use that to verify both MHLO and LMHLO variants. - Change MlirGpuTestBase::ParseMlirModule to allow returning back a failure, and use that to update the mlir_gpu_compile_test to check the new behavior. PiperOrigin-RevId: 362156962
2021-03-11 07:36:22 +08:00
":hlo_ops_common",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":hlo_ops_inc_gen",
[HLO] Add LMHLO CollectivePermute verification. - Extract verification of source target pairs attached to collective permute into a common helper function and use that to verify both MHLO and LMHLO variants. - Change MlirGpuTestBase::ParseMlirModule to allow returning back a failure, and use that to update the mlir_gpu_compile_test to check the new behavior. PiperOrigin-RevId: 362156962
2021-03-11 07:36:22 +08:00
":hlo_ops_pattern_gen",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":infer_fusibility_op_interface",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:Analysis",
[MLIR][HLO] Add `rank_specialization_cluster` op to CHLO The operation will be used to cluster compatible operations that can be rank- specialized collectively. PiperOrigin-RevId: 373128557
2021-05-11 20:17:01 +08:00
"@llvm-project//mlir:ControlFlowInterfaces",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:IR",
"@llvm-project//mlir:InferTypeOpInterface",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:Shape",
"@llvm-project//mlir:SideEffects",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
Integrate LLVM at llvm/llvm-project@c3acda0798f9 Updates LLVM usage to match [c3acda0798f9](https://github.com/llvm/llvm-project/commit/c3acda0798f9) PiperOrigin-RevId: 348896724
2020-12-24 15:53:08 +08:00
"@llvm-project//mlir:TensorDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:TransformUtils",
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
cc_library(
name = "lhlo",
srcs = [
"lib/Dialect/mhlo/IR/lhlo_ops.cc",
[MLIR:LHLO] Add optional call target arg mapping to LMHLO CustomCall operations. - XLA:HLO -> LMHLO conversion drops all token arguments and return values, however custom calls that users write still expect to get buffer pointers for these token types. - To be able to support this, add an optional call target argument mapping attribute to LMHLO custom calls. When this attribute is present, it indicates the number of arguments and returns that the custom call expects and also indicates which LMHLO arg() or output() maps to which arg or result number of the custom call. PiperOrigin-RevId: 358826664
2021-02-23 00:41:59 +08:00
"lib/Dialect/mhlo/IR/lhlo_ops_structs.cc",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_ops.h",
Remove unnecessary duplicated source from "lhlo" (NFC) PiperOrigin-RevId: 378291564
2021-06-09 09:06:26 +08:00
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_ops_structs.h",
[XLA:GPU] Add AllReduce{Start,Done} to MLIR LHLO dialect. PiperOrigin-RevId: 379455720
2021-06-15 18:54:24 +08:00
"include/mlir-hlo/utils/lhlo_utils.h",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
includes = ["include"],
deps = [
":hlo_ops_base_enums",
":hlo_ops_base_inc_gen",
":hlo_ops_base_structs",
[HLO] Add LMHLO CollectivePermute verification. - Extract verification of source target pairs attached to collective permute into a common helper function and use that to verify both MHLO and LMHLO variants. - Change MlirGpuTestBase::ParseMlirModule to allow returning back a failure, and use that to update the mlir_gpu_compile_test to check the new behavior. PiperOrigin-RevId: 362156962
2021-03-11 07:36:22 +08:00
":hlo_ops_common",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":lhlo_ops_inc_gen",
[MLIR] Add dependency from "lhlo_ops_structs_inc_gen" to "lhlo" target. PiperOrigin-RevId: 359488020
2021-02-25 19:50:11 +08:00
":lhlo_ops_structs_inc_gen",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//llvm:Support",
"@llvm-project//mlir:Analysis",
PR #48667: [mlir-hlo] Added RegionBranchOpInterfaces to lmhlo operations. Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/48667 Added RegionBranchOpInterfaces to lmhlo operations that use regions. This is needed, since the bufferization features in MLIR have to reason about the control flow within these operations. Copybara import of the project: -- 572fd7d850a46630b812da84e9094280f89f259e by Julian Gross <julian.gross@dfki.de>: Added RegionBranchOpInterfaces to lmhlo operations. PiperOrigin-RevId: 372070825
2021-05-05 15:26:46 +08:00
"@llvm-project//mlir:ControlFlowInterfaces",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:CopyOpInterface",
"@llvm-project//mlir:IR",
PR #48667: [mlir-hlo] Added RegionBranchOpInterfaces to lmhlo operations. Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/48667 Added RegionBranchOpInterfaces to lmhlo operations that use regions. This is needed, since the bufferization features in MLIR have to reason about the control flow within these operations. Copybara import of the project: -- 572fd7d850a46630b812da84e9094280f89f259e by Julian Gross <julian.gross@dfki.de>: Added RegionBranchOpInterfaces to lmhlo operations. PiperOrigin-RevId: 372070825
2021-05-05 15:26:46 +08:00
"@llvm-project//mlir:LoopLikeInterface",
Integrate LLVM at llvm/llvm-project@678241795c95 Updates LLVM usage to match [678241795c95](https://github.com/llvm/llvm-project/commit/678241795c95) PiperOrigin-RevId: 363257913
2021-03-17 04:31:59 +08:00
"@llvm-project//mlir:MemRefDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SideEffects",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
"@llvm-project//mlir:TransformUtils",
"@llvm-project//mlir:Transforms",
"@llvm-project//mlir:ViewLikeInterface",
],
alwayslink = 1,
)
cc_library(
name = "lhlo_gpu",
srcs = [
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops.cc.inc",
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops.h.inc",
"lib/Dialect/mhlo/IR/lhlo_gpu_ops.cc",
],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/IR/lhlo_gpu_ops.h",
],
includes = ["include"],
deps = [
":hlo",
":hlo_ops_base_enums",
":hlo_ops_base_structs",
[HLO] Adopt custom syntax for convolution dims and window attributes for LMHLO_GPU PiperOrigin-RevId: 374889917
2021-05-21 00:40:28 +08:00
":hlo_ops_common",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":infer_fusibility_op_interface",
[XLA:GPU] Add AllReduce{Start,Done} to MLIR LHLO dialect. PiperOrigin-RevId: 379455720
2021-06-15 18:54:24 +08:00
":lhlo",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":lhlo_gpu_ops_enums",
":lhlo_gpu_ops_inc_gen",
":lhlo_gpu_ops_structs",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:Analysis",
PR #48667: [mlir-hlo] Added RegionBranchOpInterfaces to lmhlo operations. Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/48667 Added RegionBranchOpInterfaces to lmhlo operations that use regions. This is needed, since the bufferization features in MLIR have to reason about the control flow within these operations. Copybara import of the project: -- 572fd7d850a46630b812da84e9094280f89f259e by Julian Gross <julian.gross@dfki.de>: Added RegionBranchOpInterfaces to lmhlo operations. PiperOrigin-RevId: 372070825
2021-05-05 15:26:46 +08:00
"@llvm-project//mlir:ControlFlowInterfaces",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:CopyOpInterface",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:InferTypeOpInterface",
PR #48667: [mlir-hlo] Added RegionBranchOpInterfaces to lmhlo operations. Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/48667 Added RegionBranchOpInterfaces to lmhlo operations that use regions. This is needed, since the bufferization features in MLIR have to reason about the control flow within these operations. Copybara import of the project: -- 572fd7d850a46630b812da84e9094280f89f259e by Julian Gross <julian.gross@dfki.de>: Added RegionBranchOpInterfaces to lmhlo operations. PiperOrigin-RevId: 372070825
2021-05-05 15:26:46 +08:00
"@llvm-project//mlir:LoopLikeInterface",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SideEffects",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
"@llvm-project//mlir:TransformUtils",
"@llvm-project//mlir:Transforms",
"@llvm-project//mlir:ViewLikeInterface",
],
alwayslink = 1,
)
PR #50191: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/50191 DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. Copybara import of the project: -- c4f20a89aed71181e75bcc5265723b88bde23240 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. -- 1991d4f80ab6087943956e1c0fec4940a22ab08d by Wenyi Zhao <reyizero@gmail.com>: fix PiperOrigin-RevId: 379317586
2021-06-15 02:26:41 +08:00
gentbl_cc_library(
name = "DiscRalPassIncGen",
strip_include_prefix = "include",
tbl_outs = [
(
[
"-gen-pass-decls",
"-name=RAL",
],
"include/mlir-hlo/Dialect/mhlo/transforms/disc_ral_passes.h.inc",
),
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/transforms/disc_ral_passes.td",
td_includes = [
"external/mlir-hlo/include",
"include",
],
deps = [
"@llvm-project//mlir:PassBaseTdFiles",
],
)
gentbl_cc_library(
name = "disc_ral_ops_inc_gen",
strip_include_prefix = "include",
tbl_outs = [
(
["-gen-op-decls"],
"include/mlir-hlo/Dialect/mhlo/IR/disc_ral_ops.h.inc",
),
(
["-gen-op-defs"],
"include/mlir-hlo/Dialect/mhlo/IR/disc_ral_ops.cc.inc",
),
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "include/mlir-hlo/Dialect/mhlo/IR/disc_ral_ops.td",
td_includes = [
"external/mlir-hlo/include",
"include",
],
deps = [":hlo_ops_td_files"],
)
cc_library(
name = "disc_ral",
srcs = [
"include/mlir-hlo/Dialect/mhlo/IR/disc_ral_ops.cc.inc",
"include/mlir-hlo/Dialect/mhlo/IR/disc_ral_ops.h.inc",
"lib/Dialect/mhlo/IR/disc_ral_ops.cc",
],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/IR/disc_ral_ops.h",
],
includes = ["include"],
deps = [
":disc_ral_ops_inc_gen",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:Analysis",
"@llvm-project//mlir:ControlFlowInterfaces",
"@llvm-project//mlir:CopyOpInterface",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:InferTypeOpInterface",
"@llvm-project//mlir:LoopLikeInterface",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SideEffects",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
"@llvm-project//mlir:TransformUtils",
"@llvm-project//mlir:Transforms",
"@llvm-project//mlir:ViewLikeInterface",
],
alwayslink = 1,
)
cc_library(
name = "ral_inject_execution_context",
srcs = ["lib/Dialect/mhlo/transforms/ral_inject_execution_context.cc"],
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/passes.h"],
deps = [
":disc_ral",
":pass_details",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:Shape",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:TensorDialect",
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
cc_library(
name = "hlo_dialect_registration",
srcs = ["lib/Dialect/mhlo/IR/init.cc"],
hdrs = ["include/mlir-hlo/Dialect/mhlo/IR/register.h"],
deps = [
PR #50191: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/50191 DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. Copybara import of the project: -- c4f20a89aed71181e75bcc5265723b88bde23240 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. -- 1991d4f80ab6087943956e1c0fec4940a22ab08d by Wenyi Zhao <reyizero@gmail.com>: fix PiperOrigin-RevId: 379317586
2021-06-15 02:26:41 +08:00
":disc_ral",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":hlo",
":lhlo",
":lhlo_gpu",
"@llvm-project//mlir:IR",
],
)
cc_library(
name = "sink_constants_to_control_flow",
srcs = [
"lib/Dialect/mhlo/transforms/sink_constants_to_control_flow.cc",
],
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/passes.h"],
deps = [
":hlo",
":pass_details",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
cc_library(
name = "mhlo_control_flow_to_scf",
srcs = ["lib/Dialect/mhlo/transforms/mhlo_control_flow_to_scf.cc"],
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/passes.h"],
deps = [
":hlo",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
Integrate LLVM at llvm/llvm-project@0cf7e4b252fe Updates LLVM usage to match [0cf7e4b252fe](https://github.com/llvm/llvm-project/commit/0cf7e4b252fe) PiperOrigin-RevId: 347948887
2020-12-17 12:29:15 +08:00
"@llvm-project//mlir:TensorDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
)
cc_library(
name = "map_lmhlo_to_scalar_op",
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/map_lmhlo_to_scalar_op.h"],
deps = [
":hlo",
":lhlo",
":map_hlo_to_lhlo_op",
"@llvm-project//llvm:Support",
[MLIR] Migrate TF from STD complex ops to ComplexDialect. PiperOrigin-RevId: 352966408
2021-01-21 17:21:23 +08:00
"@llvm-project//mlir:ComplexDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:IR",
Integrate LLVM at llvm/llvm-project@2bfe27da171e Updates LLVM usage to match [2bfe27da171e](https://github.com/llvm/llvm-project/commit/2bfe27da171e) PiperOrigin-RevId: 357196336
2021-02-13 00:30:51 +08:00
"@llvm-project//mlir:MathDialect",
Define lowering of [l]mhlo.pow. For floating point operations, this uses std.pow. For integer operations, this lowers to a loop. This adds a dependency on scf. PiperOrigin-RevId: 348537232
2020-12-22 07:26:38 +08:00
"@llvm-project//mlir:SCFDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:StandardOps",
],
)
cc_library(
name = "map_chlo_to_hlo_op",
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/map_chlo_to_hlo_op.h"],
deps = [
":hlo",
"@llvm-project//mlir:IR",
],
)
cc_library(
name = "map_hlo_to_lhlo_op",
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/map_hlo_to_lhlo_op.h"],
deps = [
":hlo",
":lhlo",
],
)
cc_library(
name = "lhlo_legalize_to_affine",
srcs = ["lib/Dialect/mhlo/transforms/lhlo_legalize_to_affine.cc"],
deps = [
":hlo",
":lhlo",
":map_lmhlo_to_scalar_op",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:Affine",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:TransformUtils",
],
alwayslink = 1,
)
cc_library(
name = "lhlo_legalize_to_parallel_loops",
srcs = ["lib/Dialect/mhlo/transforms/lhlo_legalize_to_parallel_loops.cc"],
deps = [
":lhlo",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:LinalgOps",
Integrate LLVM at llvm/llvm-project@678241795c95 Updates LLVM usage to match [678241795c95](https://github.com/llvm/llvm-project/commit/678241795c95) PiperOrigin-RevId: 363257913
2021-03-17 04:31:59 +08:00
"@llvm-project//mlir:MemRefDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
cc_library(
name = "legalize_to_linalg",
srcs = ["lib/Dialect/mhlo/transforms/legalize_to_linalg.cc"],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/passes.h",
"include/mlir-hlo/Dialect/mhlo/transforms/rewriters.h",
],
deps = [
":hlo",
":lhlo",
":map_lmhlo_to_scalar_op",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:Affine",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:LinalgOps",
Integrate LLVM at llvm/llvm-project@2bfe27da171e Updates LLVM usage to match [2bfe27da171e](https://github.com/llvm/llvm-project/commit/2bfe27da171e) PiperOrigin-RevId: 357196336
2021-02-13 00:30:51 +08:00
"@llvm-project//mlir:MathDialect",
mlir-hlo-opt: set preloadDialectsInContext to false. This requires specifying dependent dialects in several passes. PiperOrigin-RevId: 365758084
2021-03-30 16:06:12 +08:00
"@llvm-project//mlir:MemRefDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Pass",
Define lowering of [l]mhlo.pow. For floating point operations, this uses std.pow. For integer operations, this lowers to a loop. This adds a dependency on scf. PiperOrigin-RevId: 348537232
2020-12-22 07:26:38 +08:00
"@llvm-project//mlir:SCFDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:StandardOps",
Lower integer matmuls to linalg PiperOrigin-RevId: 359306495
2021-02-25 01:43:54 +08:00
"@llvm-project//mlir:Support",
[KERNEL_GEN] Add a pattern for hlo.dyn_broadcast->linalg to enable is_inf kernel. PiperOrigin-RevId: 351179620
2021-01-12 02:11:39 +08:00
"@llvm-project//mlir:TensorDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
cc_library(
name = "transform_unranked_hlo",
srcs = ["lib/Dialect/mhlo/transforms/transform_unranked_hlo.cc"],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/passes.h",
"include/mlir-hlo/Dialect/mhlo/transforms/rewriters.h",
],
deps = [
":hlo",
":map_chlo_to_hlo_op",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:Shape",
"@llvm-project//mlir:StandardOps",
Integrate LLVM at llvm/llvm-project@c3acda0798f9 Updates LLVM usage to match [c3acda0798f9](https://github.com/llvm/llvm-project/commit/c3acda0798f9) PiperOrigin-RevId: 348896724
2020-12-24 15:53:08 +08:00
"@llvm-project//mlir:TensorDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
[MHLO] Add pass to move up dynamic broadcasts for fusion For now, the pass only reifies the required shape computations. Moving broadcasts will follow to allow for fusion across them. PiperOrigin-RevId: 362033715
2021-03-10 22:20:43 +08:00
cc_library(
[MLIR][HLO] Rename `move-up-dynamic-broadcasts-for-fusion` to `broadcast-propagation` PiperOrigin-RevId: 378102608
2021-06-08 16:50:02 +08:00
name = "broadcast_propagation",
srcs = ["lib/Dialect/mhlo/transforms/broadcast_propagation.cc"],
[MHLO] Add pass to move up dynamic broadcasts for fusion For now, the pass only reifies the required shape computations. Moving broadcasts will follow to allow for fusion across them. PiperOrigin-RevId: 362033715
2021-03-10 22:20:43 +08:00
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/passes.h",
"include/mlir-hlo/Dialect/mhlo/transforms/rewriters.h",
],
deps = [
":hlo",
":map_chlo_to_hlo_op",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:InferTypeOpInterface",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:Shape",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:TensorDialect",
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
[MLIR][HLO] Add `rank-specialization-cluster` pass Add a pass to cluster unranked C/HLO operations in one `chlo.rank_specialization_cluster` op. The C/HLO operations are moved to the body of the operation. Later passes can use this to rank-specialize all these operations together. PiperOrigin-RevId: 373336725
2021-05-12 18:45:09 +08:00
cc_library(
name = "rank_specialization",
srcs = ["lib/Dialect/mhlo/transforms/rank_specialization.cc"],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/passes.h",
"include/mlir-hlo/Dialect/mhlo/transforms/rewriters.h",
],
deps = [
":hlo",
[MLIR][KernelGen] Make maximum supported rank in rank specialization configurable The maximum supported target rank of 5 is sufficient for all operations but `select`. Make the maximum target rank configurable in the rank specialization. This reduces the number of generated kernels for operations that don't require it. PiperOrigin-RevId: 376822496
2021-06-01 21:53:30 +08:00
":pass_details",
[MLIR][HLO] Add `rank-specialization-cluster` pass Add a pass to cluster unranked C/HLO operations in one `chlo.rank_specialization_cluster` op. The C/HLO operations are moved to the body of the operation. Later passes can use this to rank-specialize all these operations together. PiperOrigin-RevId: 373336725
2021-05-12 18:45:09 +08:00
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:InferTypeOpInterface",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:Shape",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:TensorDialect",
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
cc_library(
name = "lhlo_legalize_to_gpu",
srcs = ["lib/Dialect/mhlo/transforms/lhlo_legalize_to_gpu.cc"],
deps = [
":hlo",
":lhlo",
":map_lmhlo_to_scalar_op",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:Affine",
"@llvm-project//mlir:GPUDialect",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:LinalgOps",
mlir-hlo-opt: set preloadDialectsInContext to false. This requires specifying dependent dialects in several passes. PiperOrigin-RevId: 365758084
2021-03-30 16:06:12 +08:00
"@llvm-project//mlir:MemRefDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
cc_library(
name = "lhlo_fuse_linalg",
srcs = ["lib/Dialect/mhlo/transforms/lhlo_fuse_linalg.cc"],
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/passes.h"],
deps = [
":lhlo",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:Affine",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:LinalgTransforms",
Integrate LLVM at llvm/llvm-project@678241795c95 Updates LLVM usage to match [678241795c95](https://github.com/llvm/llvm-project/commit/678241795c95) PiperOrigin-RevId: 363257913
2021-03-17 04:31:59 +08:00
"@llvm-project//mlir:MemRefDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
Integrate LLVM at llvm/llvm-project@c3acda0798f9 Updates LLVM usage to match [c3acda0798f9](https://github.com/llvm/llvm-project/commit/c3acda0798f9) PiperOrigin-RevId: 348896724
2020-12-24 15:53:08 +08:00
"@llvm-project//mlir:TensorDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:TransformUtils",
"@llvm-project//mlir:ViewLikeInterface",
],
alwayslink = 1,
)
cc_library(
name = "hlo_legalize_to_lhlo",
srcs = ["lib/Dialect/mhlo/transforms/hlo_legalize_to_lhlo.cc"],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/passes.h",
"include/mlir-hlo/Dialect/mhlo/transforms/rewriters.h",
],
deps = [
":hlo",
":lhlo",
":map_hlo_to_lhlo_op",
PR #49970: [MLIR][DISC] bufferize DynamicReshape and DynamicBroadcastInDim Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/49970 1, add hlo-to-lhlo support for DynamicReshape and DynamicBroadcastInDim 2, add a flag `convert-to-lmhlo-only` to seperate following two case: - hlo-to-lhlo only. Simply lowers all mhlo ops to their lmhlo counterparts, do not apply any optimization (e.g. elide any buffer copy). Buffer optimization is not easy in dynamic shape world especially when involving control flow, thus we leave this to another dedicated pass. - hlo-to-lhlo-or-memref-directly. Lowers some metadata-only mhlo ops (e.g. reshape) to memref dialect directly and Lowers others to their lmhlo counterparts. Copybara import of the project: -- 562bd65a368f6194405c4ae6900e3b4388a5ec03 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] bufferize DynamicReshape and DynamicBroadcastInDim 1, add hlo-to-lhlo support for DynamicReshape and DynamicBroadcastInDim 2, add a flag `convert-to-lmhlo-only` to seperate following two case: - hlo-to-lhlo only. Simply lowers all mhlo ops to their lmhlo counterparts, do not apply any optimization (e.g. elide any buffer copy). Buffer optimization is not easy in dynamic shape world especially when involving control flow, thus we leave this to another dedicated pass. - hlo-to-lhlo-or-memref-directly. Lowers some metadata-only mhlo ops (e.g. reshape) to memref dialect directly and Lowers others to their lmhlo counterparts. PiperOrigin-RevId: 377603395
2021-06-05 06:35:08 +08:00
":pass_details",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
mlir-hlo-opt: set preloadDialectsInContext to false. This requires specifying dependent dialects in several passes. PiperOrigin-RevId: 365758084
2021-03-30 16:06:12 +08:00
"@llvm-project//mlir:MemRefDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:Shape",
"@llvm-project//mlir:ShapeTransforms",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:StandardOpsTransforms",
"@llvm-project//mlir:Support",
Integrate LLVM at llvm/llvm-project@0cf7e4b252fe Updates LLVM usage to match [0cf7e4b252fe](https://github.com/llvm/llvm-project/commit/0cf7e4b252fe) PiperOrigin-RevId: 347948887
2020-12-17 12:29:15 +08:00
"@llvm-project//mlir:TensorDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
cc_library(
name = "cycle_detector",
srcs = ["lib/utils/cycle_detector.cc"],
hdrs = ["include/mlir-hlo/utils/cycle_detector.h"],
includes = ["include"],
deps = [
"@llvm-project//llvm:Support",
],
alwayslink = 1,
)
cc_library(
name = "mhlo_fusion",
srcs = ["lib/Dialect/mhlo/transforms/mhlo_fusion.cc"],
deps = [
":cycle_detector",
":hlo",
"@llvm-project//llvm:Core",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
"@llvm-project//mlir:TransformUtils",
],
alwayslink = 1,
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "legalize_to_standard_inc_gen",
strip_include_prefix = "lib/Dialect/mhlo/transforms/",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-rewriters"],
"lib/Dialect/mhlo/transforms/generated_legalize_to_standard.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "lib/Dialect/mhlo/transforms/legalize_to_standard_patterns.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":hlo_ops_td_files",
"@llvm-project//mlir:StdOpsTdFiles",
],
)
cc_library(
name = "legalize_control_flow",
srcs = ["lib/Dialect/mhlo/transforms/legalize_control_flow.cc"],
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/passes.h"],
deps = [
":hlo",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
Integrate LLVM at llvm/llvm-project@0cf7e4b252fe Updates LLVM usage to match [0cf7e4b252fe](https://github.com/llvm/llvm-project/commit/0cf7e4b252fe) PiperOrigin-RevId: 347948887
2020-12-17 12:29:15 +08:00
"@llvm-project//mlir:TensorDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
alwayslink = 1,
)
cc_library(
name = "legalize_to_standard",
srcs = ["lib/Dialect/mhlo/transforms/legalize_to_standard.cc"],
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/passes.h"],
deps = [
":hlo",
":legalize_to_standard_inc_gen",
":legalize_trigonometric_to_approximation",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
"@llvm-project//mlir:TransformUtils",
],
alwayslink = 1,
)
cc_library(
name = "legalize_gather_to_torch_index_select",
srcs = ["lib/Dialect/mhlo/transforms/legalize_gather_to_torch_index_select.cc"],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/passes.h",
"include/mlir-hlo/Dialect/mhlo/transforms/rewriters.h",
],
deps = [
":hlo",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
cc_library(
name = "legalize_trigonometric_to_approximation",
srcs = ["lib/Dialect/mhlo/transforms/legalize_trigonometric_to_approximation.cc"],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/passes.h",
"include/mlir-hlo/Dialect/mhlo/transforms/rewriters.h",
],
includes = ["include"],
deps = [
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
Integrate LLVM at llvm/llvm-project@2bfe27da171e Updates LLVM usage to match [2bfe27da171e](https://github.com/llvm/llvm-project/commit/2bfe27da171e) PiperOrigin-RevId: 357196336
2021-02-13 00:30:51 +08:00
"@llvm-project//mlir:MathDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "lower_complex_inc_gen",
strip_include_prefix = "lib/Dialect/mhlo/transforms/",
tbl_outs = [
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
(
["-gen-rewriters"],
"lib/Dialect/mhlo/transforms/generated_lower_complex.inc",
),
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "lib/Dialect/mhlo/transforms/lower_complex_patterns.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":hlo_ops_td_files",
"@llvm-project//mlir:StdOpsTdFiles",
],
)
cc_library(
#TODO(aminim): find a better name here?
name = "mhlo_to_mhlo_lowering_patterns",
srcs = [
"lib/Dialect/mhlo/transforms/lower_complex.cc",
"lib/Dialect/mhlo/transforms/lower_general_dot.cc",
"lib/Dialect/mhlo/transforms/optimize_mhlo.cc",
],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/passes.h",
"include/mlir-hlo/Dialect/mhlo/transforms/rewriters.h",
],
deps = [
":hlo",
":lower_complex_inc_gen",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:Analysis",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
cc_library(
name = "materialize_broadcasts",
srcs = [
"lib/Dialect/mhlo/transforms/materialize_broadcasts.cc",
],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/passes.h",
"include/mlir-hlo/Dialect/mhlo/transforms/rewriters.h",
],
deps = [
":hlo",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Transforms",
],
)
cc_library(
name = "unfuse_batch_norm",
srcs = ["lib/Dialect/mhlo/transforms/unfuse_batch_norm.cc"],
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/passes.h",
],
deps = [
":hlo",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
Integrate LLVM at llvm/llvm-project@678241795c95 Updates LLVM usage to match [678241795c95](https://github.com/llvm/llvm-project/commit/678241795c95) PiperOrigin-RevId: 363257913
2021-03-17 04:31:59 +08:00
"@llvm-project//mlir:MemRefDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:StandardOps",
Integrate LLVM at llvm/llvm-project@96ef4f307df2 Updates LLVM usage to match [96ef4f307df2](https://github.com/llvm/llvm-project/commit/96ef4f307df2) PiperOrigin-RevId: 352786460
2021-01-20 23:08:32 +08:00
"@llvm-project//mlir:TensorDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Transforms",
],
)
PR #49598: [MLIR][DISC] legalize tensor_load inserted during hlo-to-lhlo conversion Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/49598 This PR implements logic for lowering memref.tensor_load ops that are inserted during `mhlo-legalize-to-lmhlo` Copybara import of the project: -- 80eb377af4e02182e1aecc943a41ca5d7d1c2100 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] legalize tensor_load inserted during hlo-to-lhlo conversion This PR implements logic for lowering memref.tensor_load ops that are inserted during `mhlo-legalize-to-lmhlo`. -- ac452fe3dcd591211cd5c59be9189fe2f7153b41 by Wenyi Zhao <reyizero@gmail.com>: minor fix -- 6b36017f8632a06adbc3e05a62975fa641d0260f by Wenyi Zhao <reyizero@gmail.com>: minor refine -- 846005cc76d0033112e47825c2e9a97790b6925f by Wenyi Zhao <reyizero@gmail.com>: minor fix -- f6a4becaa287d5ca323b2d152a4d0ae053730fd9 by Wenyi Zhao <reyizero@gmail.com>: fix -- 5555749f60f7fce8f57962860ef65efccf0362ba by Wenyi Zhao <reyizero@gmail.com>: fix -- 8873b9b6d9315c1199ca9f7c133ecf377ecd2fa6 by Wenyi Zhao <reyizero@gmail.com>: fix PiperOrigin-RevId: 376942547
2021-06-02 07:27:07 +08:00
cc_library(
name = "legalize_tensor_load_op",
srcs = ["lib/Dialect/mhlo/transforms/legalize_tensor_load_op.cc"],
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/passes.h"],
deps = [
":lhlo",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:MemRefDialect",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:Shape",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
"@llvm-project//mlir:TensorDialect",
"@llvm-project//mlir:Transforms",
],
)
PR #50020: [MLIR][DISC] support fusion on buffer Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/50020 This pass implements the logic to group kLoop/kInput fusion patterns on buffer level. The reason for this is that we can avoid a lot of headaches to handle `shape-only` consumers specially (e.g. memref.dim, shape.shapeOf) since shapes are already resolved in buffer world. It may be better to move this pass to tensor level after more shape inference/constraint infras are ready on mhlo level. Copybara import of the project: -- e31f8344b59aa9860097197585215ea1689b8ff4 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] support fusion on buffer This pass implements the logic to group kLoop/kInput fusion patterns on buffer level. The reason for this is that we can avoid a lot of headaches to handle `shape-only` consumers specially (e.g. memref.dim, shape.shapeOf) since shapes are already resolved in buffer world. It may be better to move this pass to tensor level after more shape inference/constraint infras are ready on mhlo level. -- 35f2eb2791241b0ab5db1ddcaf1b4006278ddccf by Wenyi Zhao <reyizero@gmail.com>: fix -- 923c8d61f7fe00a2a0df22d5be396508f0667964 by Wenyi Zhao <reyizero@gmail.com>: fix sanity check failure PiperOrigin-RevId: 379743424
2021-06-17 00:50:41 +08:00
cc_library(
name = "fusion_utils",
srcs = ["lib/Dialect/mhlo/transforms/fusion_utils.cc"],
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/fusion_utils.h"],
deps = [
":lhlo",
"@llvm-project//llvm:Core",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Shape",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
],
alwayslink = 1,
)
cc_library(
name = "lhlo_fusion",
srcs = ["lib/Dialect/mhlo/transforms/lhlo_fusion.cc"],
deps = [
":cycle_detector",
":fusion_utils",
":lhlo",
":pass_details",
"@llvm-project//llvm:Core",
"@llvm-project//llvm:Support",
"@llvm-project//mlir:IR",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:Shape",
"@llvm-project//mlir:StandardOps",
"@llvm-project//mlir:Support",
"@llvm-project//mlir:TransformUtils",
],
alwayslink = 1,
)
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
cc_library(
name = "chlo_legalize_to_hlo",
srcs = ["lib/Dialect/mhlo/transforms/chlo_legalize_to_hlo.cc"],
hdrs = ["include/mlir-hlo/Dialect/mhlo/transforms/rewriters.h"],
deps = [
":chlo_legalize_to_hlo_inc_gen",
":hlo",
":map_chlo_to_hlo_op",
[MLIR][CHLO] Generalize lowering with upcast to n-ary operation Allows reuse for zeta lowering now and for the polygamma lowering soon. PiperOrigin-RevId: 357739910
2021-02-17 01:46:33 +08:00
"@llvm-project//llvm:Support",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:IR",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:Shape",
"@llvm-project//mlir:StandardOps",
Integrate LLVM at llvm/llvm-project@c3acda0798f9 Updates LLVM usage to match [c3acda0798f9](https://github.com/llvm/llvm-project/commit/c3acda0798f9) PiperOrigin-RevId: 348896724
2020-12-24 15:53:08 +08:00
"@llvm-project//mlir:TensorDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Transforms",
],
)
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
gentbl_cc_library(
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
name = "chlo_legalize_to_hlo_inc_gen",
strip_include_prefix = "lib/Dialect/mhlo/transforms/",
tbl_outs = [
(
[mlir] update gentbl() Bazel macro Rename `gentbl` to `gentbl_cc_library` to make it clearer which kind of rule is ultimately used. Update `gentbl_*` macros to take `tbl_outs` options as a list rather a whitespace-separated string and remove the related string handling. PiperOrigin-RevId: 373406352
2021-05-13 01:58:02 +08:00
["-gen-rewriters"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"lib/Dialect/mhlo/transforms/generated_chlo_legalize_to_hlo.inc",
),
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "lib/Dialect/mhlo/transforms/chlo_legalize_to_hlo_patterns.td",
td_includes = [
Change hardcoded external/ include to match repository name This is the preferred name for the Bazel repository when included from another project. Note: These hardcoded "external/" includes are a horrible hack that we can hopefully get rid of soon with more robust Bazel tablegen rules. PiperOrigin-RevId: 350810768
2021-01-09 04:06:39 +08:00
"external/mlir-hlo/include",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"include",
],
Adopt td_library This avoids needing to list all transitive include dependencies and tracks include directories. PiperOrigin-RevId: 360779798
2021-03-04 08:10:20 +08:00
deps = [":hlo_ops_td_files"],
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
)
cc_library(
name = "pass_details",
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/PassDetail.h",
],
visibility = [
"//visibility:private", # This target is a private detail of pass implementations
],
deps = [
PR #50191: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/50191 DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. Copybara import of the project: -- c4f20a89aed71181e75bcc5265723b88bde23240 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. -- 1991d4f80ab6087943956e1c0fec4940a22ab08d by Wenyi Zhao <reyizero@gmail.com>: fix PiperOrigin-RevId: 379317586
2021-06-15 02:26:41 +08:00
":DiscRalPassIncGen",
PR #50020: [MLIR][DISC] support fusion on buffer Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/50020 This pass implements the logic to group kLoop/kInput fusion patterns on buffer level. The reason for this is that we can avoid a lot of headaches to handle `shape-only` consumers specially (e.g. memref.dim, shape.shapeOf) since shapes are already resolved in buffer world. It may be better to move this pass to tensor level after more shape inference/constraint infras are ready on mhlo level. Copybara import of the project: -- e31f8344b59aa9860097197585215ea1689b8ff4 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] support fusion on buffer This pass implements the logic to group kLoop/kInput fusion patterns on buffer level. The reason for this is that we can avoid a lot of headaches to handle `shape-only` consumers specially (e.g. memref.dim, shape.shapeOf) since shapes are already resolved in buffer world. It may be better to move this pass to tensor level after more shape inference/constraint infras are ready on mhlo level. -- 35f2eb2791241b0ab5db1ddcaf1b4006278ddccf by Wenyi Zhao <reyizero@gmail.com>: fix -- 923c8d61f7fe00a2a0df22d5be396508f0667964 by Wenyi Zhao <reyizero@gmail.com>: fix sanity check failure PiperOrigin-RevId: 379743424
2021-06-17 00:50:41 +08:00
":LmhloPassIncGen",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":MhloPassIncGen",
"@llvm-project//mlir:Pass",
],
)
cc_library(
name = "test_passes",
srcs = [
"include/mlir-hlo/Dialect/mhlo/transforms/rewriters.h",
"lib/Dialect/mhlo/transforms/chlo_legalize_to_hlo_pass.cc",
"lib/Dialect/mhlo/transforms/materialize_broadcasts_pass.cc",
"lib/Dialect/mhlo/transforms/optimize_mhlo_pass.cc",
"lib/Dialect/mhlo/transforms/test_infer_shaped_type_pass.cc",
"lib/Dialect/mhlo/transforms/unfuse_batch_norm_pass.cc",
],
deps = [
":chlo_legalize_to_hlo", # build-cleaner: keep
":hlo",
":lhlo",
":materialize_broadcasts", # build-cleaner: keep
":pass_details",
":unfuse_batch_norm", # build-cleaner: keep
"@llvm-project//mlir:IR",
"@llvm-project//mlir:InferTypeOpInterface",
"@llvm-project//mlir:LLVMDialect",
"@llvm-project//mlir:LLVMTransforms",
mlir-hlo-opt: set preloadDialectsInContext to false. This requires specifying dependent dialects in several passes. PiperOrigin-RevId: 365758084
2021-03-30 16:06:12 +08:00
"@llvm-project//mlir:MemRefDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:SCFDialect",
"@llvm-project//mlir:Shape",
"@llvm-project//mlir:StandardOps",
Integrate LLVM at llvm/llvm-project@c3acda0798f9 Updates LLVM usage to match [c3acda0798f9](https://github.com/llvm/llvm-project/commit/c3acda0798f9) PiperOrigin-RevId: 348896724
2020-12-24 15:53:08 +08:00
"@llvm-project//mlir:TensorDialect",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:Transforms",
],
alwayslink = 1,
)
cc_library(
name = "all_passes",
hdrs = [
"include/mlir-hlo/Dialect/mhlo/transforms/register_passes.h",
],
deps = [
PR #50191: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/50191 DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. Copybara import of the project: -- c4f20a89aed71181e75bcc5265723b88bde23240 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. -- 1991d4f80ab6087943956e1c0fec4940a22ab08d by Wenyi Zhao <reyizero@gmail.com>: fix PiperOrigin-RevId: 379317586
2021-06-15 02:26:41 +08:00
":DiscRalPassIncGen",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":LmhloPassIncGen",
":MhloPassIncGen",
[MLIR][HLO] Rename `move-up-dynamic-broadcasts-for-fusion` to `broadcast-propagation` PiperOrigin-RevId: 378102608
2021-06-08 16:50:02 +08:00
":broadcast_propagation",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":chlo_legalize_to_hlo",
":hlo_legalize_to_lhlo",
":legalize_control_flow",
":legalize_gather_to_torch_index_select",
PR #49598: [MLIR][DISC] legalize tensor_load inserted during hlo-to-lhlo conversion Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/49598 This PR implements logic for lowering memref.tensor_load ops that are inserted during `mhlo-legalize-to-lmhlo` Copybara import of the project: -- 80eb377af4e02182e1aecc943a41ca5d7d1c2100 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] legalize tensor_load inserted during hlo-to-lhlo conversion This PR implements logic for lowering memref.tensor_load ops that are inserted during `mhlo-legalize-to-lmhlo`. -- ac452fe3dcd591211cd5c59be9189fe2f7153b41 by Wenyi Zhao <reyizero@gmail.com>: minor fix -- 6b36017f8632a06adbc3e05a62975fa641d0260f by Wenyi Zhao <reyizero@gmail.com>: minor refine -- 846005cc76d0033112e47825c2e9a97790b6925f by Wenyi Zhao <reyizero@gmail.com>: minor fix -- f6a4becaa287d5ca323b2d152a4d0ae053730fd9 by Wenyi Zhao <reyizero@gmail.com>: fix -- 5555749f60f7fce8f57962860ef65efccf0362ba by Wenyi Zhao <reyizero@gmail.com>: fix -- 8873b9b6d9315c1199ca9f7c133ecf377ecd2fa6 by Wenyi Zhao <reyizero@gmail.com>: fix PiperOrigin-RevId: 376942547
2021-06-02 07:27:07 +08:00
":legalize_tensor_load_op",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":legalize_to_linalg",
":legalize_to_standard",
":legalize_trigonometric_to_approximation",
":lhlo",
":lhlo_fuse_linalg",
PR #50020: [MLIR][DISC] support fusion on buffer Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/50020 This pass implements the logic to group kLoop/kInput fusion patterns on buffer level. The reason for this is that we can avoid a lot of headaches to handle `shape-only` consumers specially (e.g. memref.dim, shape.shapeOf) since shapes are already resolved in buffer world. It may be better to move this pass to tensor level after more shape inference/constraint infras are ready on mhlo level. Copybara import of the project: -- e31f8344b59aa9860097197585215ea1689b8ff4 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] support fusion on buffer This pass implements the logic to group kLoop/kInput fusion patterns on buffer level. The reason for this is that we can avoid a lot of headaches to handle `shape-only` consumers specially (e.g. memref.dim, shape.shapeOf) since shapes are already resolved in buffer world. It may be better to move this pass to tensor level after more shape inference/constraint infras are ready on mhlo level. -- 35f2eb2791241b0ab5db1ddcaf1b4006278ddccf by Wenyi Zhao <reyizero@gmail.com>: fix -- 923c8d61f7fe00a2a0df22d5be396508f0667964 by Wenyi Zhao <reyizero@gmail.com>: fix sanity check failure PiperOrigin-RevId: 379743424
2021-06-17 00:50:41 +08:00
":lhlo_fusion",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":lhlo_legalize_to_affine",
":lhlo_legalize_to_gpu",
":lhlo_legalize_to_parallel_loops",
":mhlo_control_flow_to_scf",
":mhlo_fusion",
":mhlo_to_mhlo_lowering_patterns",
PR #50191: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/50191 DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. Copybara import of the project: -- c4f20a89aed71181e75bcc5265723b88bde23240 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. -- 1991d4f80ab6087943956e1c0fec4940a22ab08d by Wenyi Zhao <reyizero@gmail.com>: fix PiperOrigin-RevId: 379317586
2021-06-15 02:26:41 +08:00
":ral_inject_execution_context",
[MLIR][HLO] Add `rank-specialization-cluster` pass Add a pass to cluster unranked C/HLO operations in one `chlo.rank_specialization_cluster` op. The C/HLO operations are moved to the body of the operation. Later passes can use this to rank-specialize all these operations together. PiperOrigin-RevId: 373336725
2021-05-12 18:45:09 +08:00
":rank_specialization",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":sink_constants_to_control_flow",
":test_passes",
":transform_unranked_hlo",
"@llvm-project//mlir:Pass",
],
)
cc_binary(
name = "mlir-hlo-opt",
srcs = [
"tools/mlir-hlo-opt/mlir-hlo-opt.cpp",
],
deps = [
":all_passes",
PR #50191: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect Imported from GitHub PR https://github.com/tensorflow/tensorflow/pull/50191 DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. Copybara import of the project: -- c4f20a89aed71181e75bcc5265723b88bde23240 by Wenyi Zhao <reyizero@gmail.com>: [MLIR][DISC] Add RAL (Runtime abstraction layer) Dialect DISC is a e2e flow, including both compiler side and runtime side. For runtime side, we have different targeting environments (e.g. tensorflow, pytorch, or sometimes even a standalone binary). In order to simplify the design of the compiler side, we design a Runtime Abstraction Layer (RAL) to sperate the compiler side and runtime side. Thus the compiler side only need to target RAL itself and it is the responsibility of RAL to handle the differences between different targeting environments. One of the most important functions of RAL is to manage stateful resources. To this end, it provides a context object, and hides all stateful operations behind this context, thus the compiler side itself doesn't need to care about the resource initialization. For example, a kernel must be loaded before it can be launched on GPU. However, the loading operation should only be taken once during the whole lifetime of the context in order to achieve the best performance. Based on the initialization-free interfaces provided by RAL, compiler side can focus on its core optimization logic and lets the RAL to manage the resource status. The context mentioned above is passed as a parameter to the entry function and all RAL APIs should always use the context as their first argument. This CR also provides a pass to help to ensure this property. The pass rewrites the entry function to make sure their first argument is the context. For entry function, the pass also rewrites its inputs and outputs. To be concrete, all the original inputs and outputs of the entry function are received from and sent to RAL through a sequence of RAL API calls correspondingly. The motivation behind this is to hide the implementation details of I/Os. This design may also potentially enable partial execution of the compiled module when some of the inputs are ready. -- 1991d4f80ab6087943956e1c0fec4940a22ab08d by Wenyi Zhao <reyizero@gmail.com>: fix PiperOrigin-RevId: 379317586
2021-06-15 02:26:41 +08:00
":disc_ral",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
":hlo",
":lhlo",
":lhlo_gpu",
"@llvm-project//llvm:Support",
Switch deps AllPassesAndDialectsNoRegistration -> AllPassesAndDialects These targets are now identical as all registration is explicit. Temporarily leaving the old target as a (deprecated) alias while changes propagate. PiperOrigin-RevId: 366513211
2021-04-03 05:25:23 +08:00
"@llvm-project//mlir:AllPassesAndDialects",
Add Bazel build configuration to MLIR HLO standalone repo. Tested that this works in OSS in tree with `bazel build //...` and as a submodule in IREE. PiperOrigin-RevId: 347919106
2020-12-17 08:22:14 +08:00
"@llvm-project//mlir:IR",
"@llvm-project//mlir:MlirOptLib",
"@llvm-project//mlir:Pass",
"@llvm-project//mlir:Support",
],
)
Add mhlo python binding generator target This just invokes the generator backend & creates a filegroup. PiperOrigin-RevId: 377318653
2021-06-04 01:25:38 +08:00
# Python library.
td_library(
name = "MhloOpsPyTdFiles",
srcs = [
"@llvm-project//mlir:include/mlir/Bindings/Python/Attributes.td",
],
includes = ["include"],
deps = [
":hlo_ops_td_files",
"@llvm-project//mlir:OpBaseTdFiles",
],
)
gentbl_filegroup(
name = "MhloOpsPyGen",
tbl_outs = [
(
[
"-gen-python-op-bindings",
"-bind-dialect=mhlo",
],
"python/_mhlo_ops_gen.py",
),
],
tblgen = "@llvm-project//mlir:mlir-tblgen",
td_file = "python/MhloOps.td",
deps = [
":MhloOpsPyTdFiles",
],
)
filegroup(
name = "MhloOpsPyFiles",
srcs = [
"python/mhlo.py",
":MhloOpsPyGen",
],
)