Merge pull request #2351 from GMNGeoffrey:main-to-google
PiperOrigin-RevId: 318861045
diff --git a/.github/workflows/publish_docs.yml b/.github/workflows/publish_docs.yml
index 16ecca0..317f4ec 100644
--- a/.github/workflows/publish_docs.yml
+++ b/.github/workflows/publish_docs.yml
@@ -17,8 +17,7 @@
on:
push:
branches:
- # This will be switched to main after shifting to a GitHub-first workflow.
- - google
+ - main
jobs:
linux:
diff --git a/.github/workflows/synchronize_submodules.yml b/.github/workflows/synchronize_submodules.yml
index cd0aed6..aba1d6c 100644
--- a/.github/workflows/synchronize_submodules.yml
+++ b/.github/workflows/synchronize_submodules.yml
@@ -19,9 +19,6 @@
on:
push:
branches:
- - master
- # This branch doesn't exist yet, but may as part of shifting to a
- # GitHub-first workflow.
- google
jobs:
diff --git a/.github/workflows/update_submodules.yml b/.github/workflows/update_tf.yml
similarity index 95%
rename from .github/workflows/update_submodules.yml
rename to .github/workflows/update_tf.yml
index 68fd9c9..057918b 100644
--- a/.github/workflows/update_submodules.yml
+++ b/.github/workflows/update_tf.yml
@@ -15,7 +15,7 @@
# Creates a PR to update the TF submodule to HEAD & copy the latest
# version of the LLVM Bazel BUILD files from TF.
-name: Update Submodules
+name: Update TensorFlow
on:
schedule:
@@ -23,13 +23,15 @@
- cron: '0 */6 * * *'
jobs:
- update:
+ update_tf:
# Don't run this in everyone's forks.
if: github.repository == 'google/iree'
runs-on: ubuntu-18.04
steps:
- name: Checking out repository
uses: actions/checkout@v2
+ with:
+ - ref: "google"
- name: Initializing submodules
run: ./scripts/git/submodule_versions.py init
- name: Updating submodules
diff --git a/CMakeLists.txt b/CMakeLists.txt
index d8cdbaa..61d405b 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -52,8 +52,8 @@
set(IREE_TARGET_BACKENDS_TO_BUILD "all"
CACHE STRING "Semicolon-separated list of target backends to build, or \"all\"." FORCE)
# LINT.ThenChange(
-# https://github.com/google/iree/tree/master/build_tools/cmake/iree_cross_compile.cmake:iree_cross_compile_options,
-# https://github.com/google/iree/tree/master/build_tools/cmake/iree_cross_compile.cmake:iree_cross_compile_invoke
+# https://github.com/google/iree/tree/main/build_tools/cmake/iree_cross_compile.cmake:iree_cross_compile_options,
+# https://github.com/google/iree/tree/main/build_tools/cmake/iree_cross_compile.cmake:iree_cross_compile_invoke
# )
if(${IREE_BUILD_SAMPLES} OR ${IREE_BUILD_EXPERIMENTAL})
diff --git a/CONTRIBUTING.md b/CONTRIBUTING.md
index 117179d..afa3e64 100644
--- a/CONTRIBUTING.md
+++ b/CONTRIBUTING.md
@@ -24,8 +24,10 @@
consistently across the project. Most of the code style is derived from the
[Google Style Guides](http://google.github.io/styleguide/) for the appropriate
language and is generally not something we accept changes on (as clang-format
-and clang-tidy handle that for us). Improvements to code structure and clarity
-are welcome but please file issues to track such work first.
+and clang-tidy handle that for us). The compiler portion of the project follows
+[MLIR style](https://mlir.llvm.org/getting_started/DeveloperGuide/#style-guide).
+Improvements to code structure and clarity are welcome but please file issues to
+track such work first.
## Code reviews
@@ -39,25 +41,26 @@
Several of our presubmit builds will only run automatically if you are a project
collaborator. Otherwise a collaborator must label the PR with "kokoro:run". If
you are sending code changes to the project, please ask to be added as a
-collaborator, so that these can run automatically.
+collaborator, so that these can run automatically. It is generally expected that
+PRs will only be merged when all presubmit checks are passing. In some cases,
+pre-existing failures may be ignored.
+
+## Merging
+
+After review and presubmit checks, PRs should be merged with a "squash and
+merge". The squashed commit summary should match the PR title and the commit
+description should match the PR body. Accordingly, please write these as you
+would a helpful commit message. Please also keep PRs small (focused on a single
+issue) to streamline review and ease later culprit-finding. It is assumed that
+the PR author will authors merge their change unless they ask someone else to
+merge it for them (e.g. because they don't have write access).
## Peculiarities
-We use a GitHub integration to import PRs into our upstream (Google internal)
-source code management. Once it is approved internally, each PR will be merged
-into the master branch as a single commit by the same tooling. The description
-will match the PR title followed by the PR description. Accordingly, please
-write these as you would a helpful commit message. Please also keep PRs small
-(focused on a single issue) to streamline review and ease later culprit-finding.
-
-As part of a migration to make the project GitHub-first, our default branch is
-currently called `google` and all PRs should be directed there. This is an
-intermediate state. See
-https://groups.google.com/d/msg/iree-discuss/F07vsG9Ah4o/uAIusKO-BQAJ
-
Our documentation on
-[repository management](https://github.com/google/iree/blob/master/docs/repository_management.md)
+[repository management](https://github.com/google/iree/blob/main/docs/repository_management.md)
has more information on some of the oddities in our repository setup and
+workflows. For the most part, these should be transparent to normal developer
workflows.
## Community Guidelines
diff --git a/README.md b/README.md
index 84eccc6..06f88b8 100644
--- a/README.md
+++ b/README.md
@@ -110,12 +110,12 @@
## Build Status
-CI System | Build System | Platform | Component | Status
-:-------: | :----------: | :------: | :-------------: | :----:
-Kokoro | Bazel | Linux | Core | [](https://storage.googleapis.com/iree-oss-build-badges/linux/bazel/core/google_result.html)
-Kokoro | Bazel | Linux | Bindings | [](https://storage.googleapis.com/iree-oss-build-badges/linux/bazel/bindings/google_result.html)
-Kokoro | Bazel | Linux | Integrations | [](https://storage.googleapis.com/iree-oss-build-badges/linux/bazel/integrations/google_result.html)
-Kokoro | CMake | Linux | Core + Bindings | [](https://storage.googleapis.com/iree-oss-build-badges/linux/cmake/google_result.html)
+CI System | Platform | Build System | Component | Status
+:-------: | :------: | :----------: | :-------------: | :----:
+Kokoro | Linux | Bazel | Core | [](https://storage.googleapis.com/iree-oss-build-badges/linux/bazel/core/main_result.html)
+Kokoro | Linux | Bazel | Bindings | [](https://storage.googleapis.com/iree-oss-build-badges/linux/bazel/bindings/main_result.html)
+Kokoro | Linux | Bazel | Integrations | [](https://storage.googleapis.com/iree-oss-build-badges/linux/bazel/integrations/main_result.html)
+Kokoro | Linux | CMake | Core + Bindings | [](https://storage.googleapis.com/iree-oss-build-badges/linux/cmake/main_result.html)
## License
diff --git a/bindings/java/com/google/iree/Module.java b/bindings/java/com/google/iree/Module.java
index 734c25a..4eee4a5 100644
--- a/bindings/java/com/google/iree/Module.java
+++ b/bindings/java/com/google/iree/Module.java
@@ -23,7 +23,7 @@
/**
* Creates a VM module from a flatbuffer. The input ByteBuffer must be direct, and acceptable
* schemas for the flatbuffer are available at
- * https://github.com/google/iree/tree/master/iree/schemas.
+ * https://github.com/google/iree/tree/main/iree/schemas.
*/
public Module(ByteBuffer flatbufferData) throws Exception {
nativeAddress = nativeNew();
diff --git a/build_tools/bazel_to_cmake/bazel_to_cmake_targets.py b/build_tools/bazel_to_cmake/bazel_to_cmake_targets.py
index edf6b10..00ee03c 100644
--- a/build_tools/bazel_to_cmake/bazel_to_cmake_targets.py
+++ b/build_tools/bazel_to_cmake/bazel_to_cmake_targets.py
@@ -39,6 +39,7 @@
"@llvm-project//mlir:LLVMTransforms": ["MLIRStandardToLLVM"],
"@llvm-project//mlir:SCFToGPUPass": ["MLIRSCFToGPU"],
"@llvm-project//mlir:SCFDialect": ["MLIRSCF"],
+ "@llvm-project//mlir:ShapeTransforms": ["MLIRShapeOpsTransforms"],
"@llvm-project//mlir:SideEffects": ["MLIRSideEffectInterfaces"],
"@llvm-project//mlir:SPIRVDialect": ["MLIRSPIRV"],
"@llvm-project//mlir:SPIRVLowering": ["MLIRSPIRV", "MLIRSPIRVTransforms"],
diff --git a/build_tools/cmake/iree_cross_compile.cmake b/build_tools/cmake/iree_cross_compile.cmake
index 2568abd..a2f792d 100644
--- a/build_tools/cmake/iree_cross_compile.cmake
+++ b/build_tools/cmake/iree_cross_compile.cmake
@@ -92,8 +92,8 @@
string(REPLACE ";" "$<SEMICOLON>" _CONFIG_HAL_DRIVERS_TO_BUILD "${IREE_HAL_DRIVERS_TO_BUILD}")
string(REPLACE ";" "$<SEMICOLON>" _CONFIG_TARGET_BACKENDS_TO_BUILD "${IREE_TARGET_BACKENDS_TO_BUILD}")
# LINT.ThenChange(
- # https://github.com/google/iree/tree/master/CMakeLists.txt:iree_options,
- # https://github.com/google/iree/tree/master/build_tools/cmake/iree_cross_compile.cmake:iree_cross_compile_invoke
+ # https://github.com/google/iree/tree/main/CMakeLists.txt:iree_options,
+ # https://github.com/google/iree/tree/main/build_tools/cmake/iree_cross_compile.cmake:iree_cross_compile_invoke
# )
message(STATUS "C compiler for ${CONFIG_NAME} build: ${_CONFIG_C_COMPILER}")
@@ -117,8 +117,8 @@
-DIREE_BUILD_PYTHON_BINDINGS=${_CONFIG_BUILD_PYTHON_BINDINGS}
-DIREE_BUILD_EXPERIMENTAL=${_CONFIG_BUILD_EXPERIMENTAL}
# LINT.ThenChange(
- # https://github.com/google/iree/tree/master/CMakeLists.txt:iree_options,
- # https://github.com/google/iree/tree/master/build_tools/cmake/iree_cross_compile.cmake:iree_cross_compile_options,
+ # https://github.com/google/iree/tree/main/CMakeLists.txt:iree_options,
+ # https://github.com/google/iree/tree/main/build_tools/cmake/iree_cross_compile.cmake:iree_cross_compile_options,
# )
-DIREE_HAL_DRIVERS_TO_BUILD="${_CONFIG_HAL_DRIVERS_TO_BUILD}"
-DIREE_TARGET_BACKENDS_TO_BUILD="${_CONFIG_TARGET_BACKENDS_TO_BUILD}"
diff --git a/colab/README.md b/colab/README.md
index a1a35bd..d35df2a 100644
--- a/colab/README.md
+++ b/colab/README.md
@@ -12,20 +12,20 @@
Constructs a TF module for performing image edge detection and runs it using
IREE
-[](https://colab.research.google.com/github/google/iree/blob/master/colab/edge_detection.ipynb)
+[](https://colab.research.google.com/github/google/iree/blob/main/colab/edge_detection.ipynb)
### [low_level_invoke_function\.ipynb](low_level_invoke_function.ipynb)
Shows off some concepts of the low level IREE python bindings
-[](https://colab.research.google.com/github/google/iree/blob/master/colab/low_level_invoke_function.ipynb)
+[](https://colab.research.google.com/github/google/iree/blob/main/colab/low_level_invoke_function.ipynb)
### [mnist_tensorflow\.ipynb](mnist_tensorflow.ipynb)
Trains a TensorFlow 2.0 model for recognizing handwritten digits and runs it
using IREE
-[](https://colab.research.google.com/github/google/iree/blob/master/colab/mnist_tensorflow.ipynb)
+[](https://colab.research.google.com/github/google/iree/blob/main/colab/mnist_tensorflow.ipynb)
### [resnet\.ipynb](resnet.ipynb)
@@ -33,13 +33,13 @@
[ResNet50](https://www.tensorflow.org/api_docs/python/tf/keras/applications/ResNet50)
model and runs it using IREE
-[](https://colab.research.google.com/github/google/iree/blob/master/colab/resnet.ipynb)
+[](https://colab.research.google.com/github/google/iree/blob/main/colab/resnet.ipynb)
### [simple_tensorflow_module_import\.ipynb](simple_tensorflow_module_import.ipynb)
Defines a simple TF module, saves it and loads it in IREE
-[](https://colab.research.google.com/github/google/iree/blob/master/colab/simple_tensorflow_module_import.ipynb)
+[](https://colab.research.google.com/github/google/iree/blob/main/colab/simple_tensorflow_module_import.ipynb)
## Working with GitHub
diff --git a/colab/mnist_tensorflow.ipynb b/colab/mnist_tensorflow.ipynb
index df7c055..734e338 100644
--- a/colab/mnist_tensorflow.ipynb
+++ b/colab/mnist_tensorflow.ipynb
@@ -51,7 +51,7 @@
"# See the License for the specific language governing permissions and\n",
"# limitations under the License."
],
- "execution_count": 0,
+ "execution_count": 1,
"outputs": []
},
{
@@ -69,7 +69,7 @@
"\n",
"## Running Locally\n",
"\n",
- "* Refer to [iree/docs/using_colab.md](https://github.com/google/iree/blob/master/docs/using_colab.md) for general information\n",
+ "* Refer to [iree/docs/using_colab.md](https://github.com/google/iree/blob/main/docs/using_colab.md) for general information\n",
"* Ensure that you have a recent version of TensorFlow 2.0 [installed on your system](https://www.tensorflow.org/install)\n",
"* Enable IREE/TF integration by adding to your user.bazelrc: `build --define=iree_tensorflow=true`\n",
"* Start colab by running `python colab/start_colab_kernel.py` (see that file for additional instructions)\n",
@@ -95,7 +95,7 @@
"base_uri": "https://localhost:8080/",
"height": 51
},
- "outputId": "d471e3cd-ac1d-42f0-c344-4490a51f3a54"
+ "outputId": "fe0d703a-2ad7-4d14-9aef-c69b4c342a16"
},
"source": [
"import os\n",
@@ -119,8 +119,8 @@
{
"output_type": "stream",
"text": [
- "TensorFlow version: 2.1.0-dev20191126\n",
- "Numpy version: 1.17.4\n"
+ "TensorFlow version: 2.5.0-dev20200626\n",
+ "Numpy version: 1.18.4\n"
],
"name": "stdout"
}
@@ -132,11 +132,11 @@
"id": "43BH_9YcsGs8",
"colab_type": "code",
"cellView": "form",
- "outputId": "fc788105-8739-45d6-8df2-e245d5ff6cb1",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 34
- }
+ },
+ "outputId": "46a560cb-5947-4cfa-f71e-9065bf2c07ab"
},
"source": [
"#@title Notebook settings { run: \"auto\" }\n",
@@ -170,7 +170,7 @@
{
"output_type": "stream",
"text": [
- "Using IREE compiler backend 'vulkan-spirv' and runtime driver 'vulkan'\n"
+ "Using IREE compiler backend 'vulkan-spirv' and runtime driver 'vulkan'\r\n"
],
"name": "stdout"
}
@@ -194,11 +194,11 @@
"id": "GXZIrReTbTHN",
"colab_type": "code",
"cellView": "form",
- "outputId": "5faf02eb-aab4-4e4c-9df9-e7665ce5a810",
"colab": {
"base_uri": "https://localhost:8080/",
- "height": 506
- }
+ "height": 486
+ },
+ "outputId": "9c01fab1-f8cb-4a63-fff1-6b82a9b6b49d"
},
"source": [
"#@title Load MNIST dataset, setup training and evaluation\n",
@@ -258,8 +258,8 @@
{
"output_type": "stream",
"text": [
- "Loaded MNIST dataset!\n",
- "INFO:tensorflow:ParameterServerStrategy (CentralStorageStrategy if you are using a single machine) with compute_devices = ('/device:CPU:0',), variable_device = '/device:CPU:0'\n",
+ "Loaded MNIST dataset!\r\n",
+ "INFO:tensorflow:ParameterServerStrategy (CentralStorageStrategy if you are using a single machine) with compute_devices = ['/job:localhost/replica:0/task:0/device:CPU:0'], variable_device = '/job:localhost/replica:0/task:0/device:CPU:0'\n",
"Configured data for training and evaluation!\n",
" sample shape: (28, 28, 1)\n",
" training samples: 60000\n",
@@ -278,7 +278,7 @@
"text/plain": [
"<Figure size 432x288 with 1 Axes>"
],
- "image/png": "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\n"
+ "image/png": "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\n"
},
"metadata": {
"tags": [],
@@ -289,7 +289,7 @@
"output_type": "stream",
"text": [
"\n",
- "Ground truth labels: [0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]\n"
+ "Ground truth labels: [0. 0. 1. 0. 0. 0. 0. 0. 0. 0.]\r\n"
],
"name": "stdout"
}
@@ -318,7 +318,7 @@
" model.add(tf.keras.layers.Dense(10, activation='softmax'))\n",
" return model"
],
- "execution_count": 0,
+ "execution_count": 5,
"outputs": []
},
{
@@ -327,11 +327,11 @@
"id": "7Gdxh7qWcPSO",
"colab_type": "code",
"cellView": "form",
- "outputId": "9886da4f-1916-4540-9f53-cc82f8ab113d",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 374
- }
+ },
+ "outputId": "50b0aede-9a8f-4ce5-b340-783be5fbfc06"
},
"source": [
"#@title Train the Keras model\n",
@@ -364,7 +364,7 @@
" training_loss.update_state(loss)\n",
" training_accuracy.update_state(labels, logits)\n",
"\n",
- " strategy.experimental_run_v2(step_fn, args=(next(iterator),))\n",
+ " strategy.run(step_fn, args=(next(iterator),))\n",
"\n",
" @tf.function\n",
" def test_step(iterator):\n",
@@ -378,7 +378,7 @@
" test_loss.update_state(loss)\n",
" test_accuracy.update_state(labels, logits)\n",
"\n",
- " strategy.experimental_run_v2(step_fn, args=(next(iterator),))\n",
+ " strategy.run(step_fn, args=(next(iterator),))\n",
"\n",
" for epoch in range(0, num_epochs):\n",
" tf.print(\"Running epoch #%s\" % (epoch + 1))\n",
@@ -413,24 +413,24 @@
"text": [
"Constructed Keras MNIST model, training...\n",
"Running epoch #1\n",
- " Training loss: 0.733241, accuracy: 81.463333\n",
- " Test loss : 0.388786, accuracy: 89.590004\n",
+ " Training loss: 0.732439, accuracy: 81.403336\n",
+ " Test loss : 0.390855, accuracy: 89.490005\n",
"Running epoch #2\n",
- " Training loss: 0.362705, accuracy: 89.995003\n",
- " Test loss : 0.313008, accuracy: 91.320000\n",
+ " Training loss: 0.365308, accuracy: 89.811668\n",
+ " Test loss : 0.315630, accuracy: 91.119995\n",
"Running epoch #3\n",
- " Training loss: 0.308992, accuracy: 91.294998\n",
- " Test loss : 0.279223, accuracy: 92.150002\n",
+ " Training loss: 0.312111, accuracy: 91.129997\n",
+ " Test loss : 0.281829, accuracy: 92.040001\n",
"Running epoch #4\n",
- " Training loss: 0.278085, accuracy: 92.198334\n",
- " Test loss : 0.256378, accuracy: 92.629997\n",
+ " Training loss: 0.281028, accuracy: 92.038330\n",
+ " Test loss : 0.258432, accuracy: 92.629997\n",
"Running epoch #5\n",
- " Training loss: 0.255360, accuracy: 92.818329\n",
- " Test loss : 0.238477, accuracy: 93.099998\n",
+ " Training loss: 0.257909, accuracy: 92.753334\n",
+ " Test loss : 0.240058, accuracy: 93.229996\n",
"Completed training!\n",
"\n",
"Sample prediction:\n",
- "[1.24276197 0.00645936839 94.5438538 1.92194772 5.14127169e-06 1.25558567 0.77046895 1.55102152e-05 0.25887084 2.61111491e-05]\n",
+ "[0.243134052 0.00337268948 95.5214081 0.925373673 2.25061958e-05 0.992091119 2.20864391 3.87712953e-06 0.105901182 4.44369543e-05]\n",
"\n"
],
"name": "stdout"
@@ -442,11 +442,11 @@
"metadata": {
"id": "DmespEaFcSEL",
"colab_type": "code",
- "outputId": "34ecdcca-5ceb-43f0-c793-e209e25dafce",
"colab": {
"base_uri": "https://localhost:8080/",
- "height": 122
- }
+ "height": 153
+ },
+ "outputId": "3c8579db-7b3c-4164-ff91-394346595107"
},
"source": [
"#@title Export the trained model as a SavedModel, with IREE-compatible settings\n",
@@ -472,11 +472,14 @@
{
"output_type": "stream",
"text": [
- "Exporting SavedModel to /tmp/mnist.sm\n",
- "WARNING:tensorflow:From /usr/local/lib/python3.7/dist-packages/tensorflow_core/python/ops/resource_variable_ops.py:1788: calling BaseResourceVariable.__init__ (from tensorflow.python.ops.resource_variable_ops) with constraint is deprecated and will be removed in a future version.\n",
+ "Exporting SavedModel to /tmp/mnist.sm\r\n",
+ "WARNING:tensorflow:From c:\\users\\scott\\scoop\\apps\\python\\current\\lib\\site-packages\\tensorflow\\python\\training\\tracking\\tracking.py:111: Model.state_updates (from tensorflow.python.keras.engine.training) is deprecated and will be removed in a future version.\n",
"Instructions for updating:\n",
- "If using Keras pass *_constraint arguments to layers.\n",
- "INFO:tensorflow:Assets written to: /tmp/mnist.sm/assets\n"
+ "This property should not be used in TensorFlow 2.0, as updates are applied automatically.\n",
+ "WARNING:tensorflow:From c:\\users\\scott\\scoop\\apps\\python\\current\\lib\\site-packages\\tensorflow\\python\\training\\tracking\\tracking.py:111: Layer.updates (from tensorflow.python.keras.engine.base_layer) is deprecated and will be removed in a future version.\n",
+ "Instructions for updating:\n",
+ "This property should not be used in TensorFlow 2.0, as updates are applied automatically.\n",
+ "INFO:tensorflow:Assets written to: /tmp/mnist.sm\\assets\n"
],
"name": "stdout"
}
@@ -497,11 +500,11 @@
"metadata": {
"id": "rqwIx4j4gS1a",
"colab_type": "code",
- "outputId": "28c7516c-903c-4579-c4bc-8299712a9edd",
"colab": {
"base_uri": "https://localhost:8080/",
- "height": 751
- }
+ "height": 836
+ },
+ "outputId": "24cded90-c436-47ce-b4f4-7a5da46ea38a"
},
"source": [
"#@title Load the SavedModel into IREE's compiler as MLIR xla_hlo\n",
@@ -524,16 +527,16 @@
"Imported MLIR:\n",
" \n",
"\n",
- "module attributes {tf.versions = {bad_consumers = [], min_consumer = 12 : i32, producer = 175 : i32}} {\n",
- " flow.variable @\"__iree_flow___sm_node15__model.layer-2.kernel\" opaque<\"\", \"0xDEADBEEF\"> : tensor<784x128xf32>\n",
- " flow.variable @\"__iree_flow___sm_node16__model.layer-2.bias\" opaque<\"\", \"0xDEADBEEF\"> : tensor<128xf32>\n",
- " flow.variable @\"__iree_flow___sm_node21__model.layer-3.kernel\" opaque<\"\", \"0xDEADBEEF\"> : tensor<128x10xf32>\n",
- " flow.variable @\"__iree_flow___sm_node22__model.layer-3.bias\" dense<[-0.0719004869, 0.1290171, 0.0102811698, -0.106104381, 0.0260288324, 0.166622087, -8.5693755E-4, 0.070880115, -0.222566754, -0.00140074058]> : tensor<10xf32>\n",
- " func @predict(%arg0: tensor<1x28x28x1xf32>) -> tensor<1x10xf32> attributes {iree.module.export, iree.reflection = {abi = \"sip\", abiv = 1 : i32, sip = \"I8!S5!k0_0R3!_0\"}, tf._input_shapes = [\"tfshape$dim { size: 1 } dim { size: 28 } dim { size: 28 } dim { size: 1 }\", \"tfshape$unknown_rank: true\", \"tfshape$unknown_rank: true\", \"tfshape$unknown_rank: true\", \"tfshape$unknown_rank: true\"], tf.signature.is_stateful} {\n",
- " %0 = flow.variable.address @\"__iree_flow___sm_node15__model.layer-2.kernel\" : !iree.ptr<tensor<784x128xf32>>\n",
- " %1 = flow.variable.address @\"__iree_flow___sm_node16__model.layer-2.bias\" : !iree.ptr<tensor<128xf32>>\n",
- " %2 = flow.variable.address @\"__iree_flow___sm_node21__model.layer-3.kernel\" : !iree.ptr<tensor<128x10xf32>>\n",
- " %3 = flow.variable.address @\"__iree_flow___sm_node22__model.layer-3.bias\" : !iree.ptr<tensor<10xf32>>\n",
+ "module attributes {tf.versions = {bad_consumers = [], min_consumer = 12 : i32, producer = 443 : i32}} {\n",
+ " flow.variable @\"__iree_flow___sm_node14__model.layer-1.kernel\" opaque<\"\", \"0xDEADBEEF\"> : tensor<784x128xf32> attributes {sym_visibility = \"private\"}\n",
+ " flow.variable @\"__iree_flow___sm_node15__model.layer-1.bias\" opaque<\"\", \"0xDEADBEEF\"> : tensor<128xf32> attributes {sym_visibility = \"private\"}\n",
+ " flow.variable @\"__iree_flow___sm_node20__model.layer-2.kernel\" opaque<\"\", \"0xDEADBEEF\"> : tensor<128x10xf32> attributes {sym_visibility = \"private\"}\n",
+ " flow.variable @\"__iree_flow___sm_node21__model.layer-2.bias\" dense<[-0.114143081, 0.0953421518, 4.84912744E-5, -0.0384164825, 0.0063888072, 0.218958765, 0.0256200824, 0.0551806651, -0.22108613, -0.0278935507]> : tensor<10xf32> attributes {sym_visibility = \"private\"}\n",
+ " func @predict(%arg0: tensor<1x28x28x1xf32> {tf._user_specified_name = \"x\"}) -> tensor<1x10xf32> attributes {iree.module.export, iree.reflection = {abi = \"sip\", abiv = 1 : i32, sip = \"I8!S5!k0_0R3!_0\"}, tf._input_shapes = [#tf.shape<1x28x28x1>, #tf.shape<*>, #tf.shape<*>, #tf.shape<*>, #tf.shape<*>], tf.signature.is_stateful} {\n",
+ " %0 = flow.variable.address @\"__iree_flow___sm_node14__model.layer-1.kernel\" : !iree.ptr<tensor<784x128xf32>>\n",
+ " %1 = flow.variable.address @\"__iree_flow___sm_node15__model.layer-1.bias\" : !iree.ptr<tensor<128xf32>>\n",
+ " %2 = flow.variable.address @\"__iree_flow___sm_node20__model.layer-2.kernel\" : !iree.ptr<tensor<128x10xf32>>\n",
+ " %3 = flow.variable.address @\"__iree_flow___sm_node21__model.layer-2.bias\" : !iree.ptr<tensor<10xf32>>\n",
" %4 = xla_hlo.constant dense<0xFF800000> : tensor<f32>\n",
" %5 = xla_hlo.constant dense<0.000000e+00> : tensor<f32>\n",
" %6 = flow.variable.load.indirect %3 : !iree.ptr<tensor<10xf32>> -> tensor<10xf32>\n",
@@ -542,27 +545,32 @@
" %9 = flow.variable.load.indirect %0 : !iree.ptr<tensor<784x128xf32>> -> tensor<784x128xf32>\n",
" %10 = \"xla_hlo.reshape\"(%arg0) : (tensor<1x28x28x1xf32>) -> tensor<1x784xf32>\n",
" %11 = \"xla_hlo.dot\"(%10, %9) : (tensor<1x784xf32>, tensor<784x128xf32>) -> tensor<1x128xf32>\n",
- " %12 = \"xla_hlo.add\"(%11, %8) {broadcast_dimensions = dense<1> : tensor<1xi64>} : (tensor<1x128xf32>, tensor<128xf32>) -> tensor<1x128xf32>\n",
- " %13 = \"xla_hlo.maximum\"(%12, %5) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<1x128xf32>, tensor<f32>) -> tensor<1x128xf32>\n",
- " %14 = \"xla_hlo.dot\"(%13, %7) : (tensor<1x128xf32>, tensor<128x10xf32>) -> tensor<1x10xf32>\n",
- " %15 = \"xla_hlo.add\"(%14, %6) {broadcast_dimensions = dense<1> : tensor<1xi64>} : (tensor<1x10xf32>, tensor<10xf32>) -> tensor<1x10xf32>\n",
- " %16 = \"xla_hlo.reduce\"(%15, %4) ( {\n",
- " ^bb0(%arg1: tensor<f32>, %arg2: tensor<f32>):\t// no predecessors\n",
- " %21 = xla_hlo.maximum %arg1, %arg2 : tensor<f32>\n",
- " \"xla_hlo.return\"(%21) : (tensor<f32>) -> ()\n",
+ " %12 = \"xla_hlo.broadcast_in_dim\"(%8) {broadcast_dimensions = dense<1> : tensor<1xi64>} : (tensor<128xf32>) -> tensor<1x128xf32>\n",
+ " %13 = xla_hlo.add %11, %12 : tensor<1x128xf32>\n",
+ " %14 = \"xla_hlo.broadcast_in_dim\"(%5) {broadcast_dimensions = dense<[]> : tensor<0xi64>} : (tensor<f32>) -> tensor<1x128xf32>\n",
+ " %15 = xla_hlo.maximum %14, %13 : tensor<1x128xf32>\n",
+ " %16 = \"xla_hlo.dot\"(%15, %7) : (tensor<1x128xf32>, tensor<128x10xf32>) -> tensor<1x10xf32>\n",
+ " %17 = \"xla_hlo.broadcast_in_dim\"(%6) {broadcast_dimensions = dense<1> : tensor<1xi64>} : (tensor<10xf32>) -> tensor<1x10xf32>\n",
+ " %18 = xla_hlo.add %16, %17 : tensor<1x10xf32>\n",
+ " %19 = \"xla_hlo.reduce\"(%18, %4) ( {\n",
+ " ^bb0(%arg1: tensor<f32>, %arg2: tensor<f32>): // no predecessors\n",
+ " %26 = xla_hlo.maximum %arg1, %arg2 : tensor<f32>\n",
+ " \"xla_hlo.return\"(%26) : (tensor<f32>) -> ()\n",
" }) {dimensions = dense<1> : tensor<1xi64>} : (tensor<1x10xf32>, tensor<f32>) -> tensor<1xf32>\n",
- " %17 = \"xla_hlo.subtract\"(%15, %16) {broadcast_dimensions = dense<0> : tensor<1xi64>} : (tensor<1x10xf32>, tensor<1xf32>) -> tensor<1x10xf32>\n",
- " %18 = \"xla_hlo.exponential\"(%17) : (tensor<1x10xf32>) -> tensor<1x10xf32>\n",
- " %19 = \"xla_hlo.reduce\"(%18, %5) ( {\n",
- " ^bb0(%arg1: tensor<f32>, %arg2: tensor<f32>):\t// no predecessors\n",
- " %21 = xla_hlo.add %arg1, %arg2 : tensor<f32>\n",
- " \"xla_hlo.return\"(%21) : (tensor<f32>) -> ()\n",
+ " %20 = \"xla_hlo.broadcast_in_dim\"(%19) {broadcast_dimensions = dense<0> : tensor<1xi64>} : (tensor<1xf32>) -> tensor<1x10xf32>\n",
+ " %21 = xla_hlo.subtract %18, %20 : tensor<1x10xf32>\n",
+ " %22 = \"xla_hlo.exponential\"(%21) : (tensor<1x10xf32>) -> tensor<1x10xf32>\n",
+ " %23 = \"xla_hlo.reduce\"(%22, %5) ( {\n",
+ " ^bb0(%arg1: tensor<f32>, %arg2: tensor<f32>): // no predecessors\n",
+ " %26 = xla_hlo.add %arg1, %arg2 : tensor<f32>\n",
+ " \"xla_hlo.return\"(%26) : (tensor<f32>) -> ()\n",
" }) {dimensions = dense<1> : tensor<1xi64>} : (tensor<1x10xf32>, tensor<f32>) -> tensor<1xf32>\n",
- " %20 = \"xla_hlo.divide\"(%18, %19) {broadcast_dimensions = dense<0> : tensor<1xi64>} : (tensor<1x10xf32>, tensor<1xf32>) -> tensor<1x10xf32>\n",
- " return %20 : tensor<1x10xf32>\n",
+ " %24 = \"xla_hlo.broadcast_in_dim\"(%23) {broadcast_dimensions = dense<0> : tensor<1xi64>} : (tensor<1xf32>) -> tensor<1x10xf32>\n",
+ " %25 = xla_hlo.divide %22, %24 : tensor<1x10xf32>\n",
+ " return %25 : tensor<1x10xf32>\n",
" }\n",
- "}\n",
- "Wrote MLIR to path '/usr/local/google/home/scotttodd/saved_models/mnist.mlir'\n"
+ "}\r\n",
+ "Wrote MLIR to path 'C:\\Users\\Scott\\saved_models\\mnist.mlir'\n"
],
"name": "stdout"
}
@@ -573,11 +581,11 @@
"metadata": {
"id": "IDHI7h3khJr9",
"colab_type": "code",
- "outputId": "f2abfd0c-5ada-4533-ab2f-a47f77ccd260",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 51
- }
+ },
+ "outputId": "b8958b7f-c7bb-4fbd-b800-e58c46134086"
},
"source": [
"#@title Compile the xla_hlo MLIR and prepare a context to execute it\n",
@@ -596,8 +604,8 @@
{
"output_type": "stream",
"text": [
- "Created IREE driver vulkan: <pyiree.rt.binding.HalDriver object at 0x7f58bc3d27f0>\n",
- "SystemContext driver=<pyiree.rt.binding.HalDriver object at 0x7f58bc3d27f0>\n"
+ "Created IREE driver vulkan: <pyiree.rt.binding.HalDriver object at 0x000001DC44C47370>\n",
+ "SystemContext driver=<pyiree.rt.binding.HalDriver object at 0x000001DC44C47370>\n"
],
"name": "stderr"
}
@@ -608,11 +616,11 @@
"metadata": {
"id": "SKflpnLtkLYE",
"colab_type": "code",
- "outputId": "59c52840-5cc6-4038-c330-707bd9313f4b",
"colab": {
"base_uri": "https://localhost:8080/",
"height": 102
- }
+ },
+ "outputId": "337ddf79-6746-4685-89f5-65787280c0af"
},
"source": [
"#@title Execute the compiled module and compare the results with TensorFlow\n",
@@ -632,10 +640,10 @@
"output_type": "stream",
"text": [
"IREE prediction ('vulkan-spirv' backend, 'vulkan' driver):\n",
- "[1.24276221 0.00645937538 94.5438538 1.92194688 5.14127305e-06 1.25558555 0.770468712 1.55102171e-05 0.25887078 2.61111945e-05]\n",
+ "[0.243133873 0.00337268622 95.5214233 0.92537272 2.25061631e-05 0.992090821 2.20864058 3.87712225e-06 0.105901062 4.44369434e-05]\n",
"\n",
"TensorFlow prediction:\n",
- "[1.24276197 0.00645936839 94.5438538 1.92194772 5.14127169e-06 1.25558567 0.77046895 1.55102152e-05 0.25887084 2.61111491e-05]\n"
+ "[0.243134052 0.00337268948 95.5214081 0.925373673 2.25061958e-05 0.992091119 2.20864391 3.87712953e-06 0.105901182 4.44369543e-05]\n"
],
"name": "stdout"
}
diff --git a/docs/GetStarted/getting_started_linux_bazel.md b/docs/GetStarted/getting_started_linux_bazel.md
index 738a365..8d4a9dc 100644
--- a/docs/GetStarted/getting_started_linux_bazel.md
+++ b/docs/GetStarted/getting_started_linux_bazel.md
@@ -20,7 +20,7 @@
### Install Bazel
Install Bazel version > 2.0.0 (see
-[`.bazelversion`](https://github.com/google/iree/blob/master/.bazelversion) for
+[`.bazelversion`](https://github.com/google/iree/blob/main/.bazelversion) for
the specific version IREE uses) by following the
[official docs](https://docs.bazel.build/versions/master/install.html).
@@ -112,7 +112,7 @@
```
Translate a
-[MLIR file](https://github.com/google/iree/blob/master/iree/tools/test/simple.mlir)
+[MLIR file](https://github.com/google/iree/blob/main/iree/tools/test/simple.mlir)
and execute a function in the compiled module:
```shell
diff --git a/docs/GetStarted/getting_started_linux_cmake.md b/docs/GetStarted/getting_started_linux_cmake.md
index ed7856e..70da146 100644
--- a/docs/GetStarted/getting_started_linux_cmake.md
+++ b/docs/GetStarted/getting_started_linux_cmake.md
@@ -77,7 +77,7 @@
> Tip:<br>
> The root
-> [CMakeLists.txt](https://github.com/google/iree/blob/master/CMakeLists.txt)
+> [CMakeLists.txt](https://github.com/google/iree/blob/main/CMakeLists.txt)
> file has options for configuring which parts of the project to enable.<br>
> These are further documented in [CMake Options and Variables](cmake_options_and_variables.md).
@@ -99,7 +99,7 @@
```
Translate a
-[MLIR file](https://github.com/google/iree/blob/master/iree/tools/test/simple.mlir)
+[MLIR file](https://github.com/google/iree/blob/main/iree/tools/test/simple.mlir)
and execute a function in the compiled module:
```shell
diff --git a/docs/GetStarted/getting_started_linux_vulkan.md b/docs/GetStarted/getting_started_linux_vulkan.md
index 9ee8c61..5a0fd7f 100644
--- a/docs/GetStarted/getting_started_linux_vulkan.md
+++ b/docs/GetStarted/getting_started_linux_vulkan.md
@@ -39,7 +39,7 @@
### Quick Start
The
-[dynamic_symbols_test](https://github.com/google/iree/blob/master/iree/hal/vulkan/dynamic_symbols_test.cc)
+[dynamic_symbols_test](https://github.com/google/iree/blob/main/iree/hal/vulkan/dynamic_symbols_test.cc)
checks if the Vulkan loader and a valid ICD are accessible.
Run the test:
@@ -58,7 +58,7 @@
HAL, which includes checking for supported layers and extensions.
Run the
-[device creation test](https://github.com/google/iree/blob/master/iree/hal/cts/device_creation_test.cc):
+[device creation test](https://github.com/google/iree/blob/main/iree/hal/cts/device_creation_test.cc):
```shell
# -- CMake --
@@ -79,7 +79,7 @@
SwiftShader's Vulkan ICD (Installable Client Driver).
IREE has a
-[helper script](https://github.com/google/iree/blob/master/build_tools/third_party/swiftshader/build_vk_swiftshader.sh)
+[helper script](https://github.com/google/iree/blob/main/build_tools/third_party/swiftshader/build_vk_swiftshader.sh)
for building SwiftShader from source using CMake:
```shell
diff --git a/docs/GetStarted/getting_started_macos_bazel.md b/docs/GetStarted/getting_started_macos_bazel.md
index a169f8d..2b285eb 100644
--- a/docs/GetStarted/getting_started_macos_bazel.md
+++ b/docs/GetStarted/getting_started_macos_bazel.md
@@ -115,7 +115,7 @@
```
Translate a
-[MLIR file](https://github.com/google/iree/blob/master/iree/tools/test/simple.mlir)
+[MLIR file](https://github.com/google/iree/blob/main/iree/tools/test/simple.mlir)
and execute a function in the compiled module:
```shell
diff --git a/docs/GetStarted/getting_started_macos_cmake.md b/docs/GetStarted/getting_started_macos_cmake.md
index 59d22e8..51ef0ab 100644
--- a/docs/GetStarted/getting_started_macos_cmake.md
+++ b/docs/GetStarted/getting_started_macos_cmake.md
@@ -77,7 +77,7 @@
> Tip:<br>
> The root
-> [CMakeLists.txt](https://github.com/google/iree/blob/master/CMakeLists.txt)
+> [CMakeLists.txt](https://github.com/google/iree/blob/main/CMakeLists.txt)
> file has options for configuring which parts of the project to enable.<br>
> These are further documented in [CMake Options and Variables](cmake_options_and_variables.md).
@@ -99,7 +99,7 @@
```
Translate a
-[MLIR file](https://github.com/google/iree/blob/master/iree/tools/test/simple.mlir)
+[MLIR file](https://github.com/google/iree/blob/main/iree/tools/test/simple.mlir)
and execute a function in the compiled module:
```shell
diff --git a/docs/GetStarted/getting_started_python.md b/docs/GetStarted/getting_started_python.md
index 627dd3d..3aec4e2 100644
--- a/docs/GetStarted/getting_started_python.md
+++ b/docs/GetStarted/getting_started_python.md
@@ -61,9 +61,8 @@
## Using Colab
See
-[start_colab_kernel.py](https://github.com/google/iree/blob/master/colab/start_colab_kernel.py)
-and
-[Using Colab](https://github.com/google/iree/blob/master/docs/using_colab.md)
+[start_colab_kernel.py](https://github.com/google/iree/blob/main/colab/start_colab_kernel.py)
+and [Using Colab](https://github.com/google/iree/blob/main/docs/using_colab.md)
for setup instructions, then take a look through the
-[Colab directory](https://github.com/google/iree/tree/master/colab) for some
+[Colab directory](https://github.com/google/iree/tree/main/colab) for some
sample notebooks.
diff --git a/docs/GetStarted/getting_started_windows_bazel.md b/docs/GetStarted/getting_started_windows_bazel.md
index 3eb660d..d3f01c3 100644
--- a/docs/GetStarted/getting_started_windows_bazel.md
+++ b/docs/GetStarted/getting_started_windows_bazel.md
@@ -25,7 +25,7 @@
### Install Bazel
Install Bazel version > 2.0.0 (see
-[`.bazelversion`](https://github.com/google/iree/blob/master/.bazelversion) for
+[`.bazelversion`](https://github.com/google/iree/blob/main/.bazelversion) for
the specific version IREE uses) by following the
[official docs](https://docs.bazel.build/versions/master/install-windows.html).
@@ -108,7 +108,7 @@
```
Translate a
-[MLIR file](https://github.com/google/iree/blob/master/iree/tools/test/simple.mlir)
+[MLIR file](https://github.com/google/iree/blob/main/iree/tools/test/simple.mlir)
and execute a function in the compiled module:
```powershell
diff --git a/docs/GetStarted/getting_started_windows_cmake.md b/docs/GetStarted/getting_started_windows_cmake.md
index 77d4fe4..dcab418 100644
--- a/docs/GetStarted/getting_started_windows_cmake.md
+++ b/docs/GetStarted/getting_started_windows_cmake.md
@@ -75,7 +75,7 @@
> Tip:<br>
> The root
-> [CMakeLists.txt](https://github.com/google/iree/blob/master/CMakeLists.txt)
+> [CMakeLists.txt](https://github.com/google/iree/blob/main/CMakeLists.txt)
> file has options for configuring which parts of the project to enable.<br>
> These are further documented in [CMake Options and Variables](cmake_options_and_variables.md).
@@ -97,7 +97,7 @@
```
Translate a
-[MLIR file](https://github.com/google/iree/blob/master/iree/tools/test/simple.mlir)
+[MLIR file](https://github.com/google/iree/blob/main/iree/tools/test/simple.mlir)
and execute a function in the compiled module:
```powershell
diff --git a/docs/GetStarted/getting_started_windows_vulkan.md b/docs/GetStarted/getting_started_windows_vulkan.md
index 9ceee2a..4e6c7b9 100644
--- a/docs/GetStarted/getting_started_windows_vulkan.md
+++ b/docs/GetStarted/getting_started_windows_vulkan.md
@@ -39,7 +39,7 @@
### Quick Start
The
-[dynamic_symbols_test](https://github.com/google/iree/blob/master/iree/hal/vulkan/dynamic_symbols_test.cc)
+[dynamic_symbols_test](https://github.com/google/iree/blob/main/iree/hal/vulkan/dynamic_symbols_test.cc)
checks if the Vulkan loader and a valid ICD are accessible.
Run the test:
@@ -58,7 +58,7 @@
HAL, which includes checking for supported layers and extensions.
Run the
-[device creation test](https://github.com/google/iree/blob/master/iree/hal/cts/device_creation_test.cc):
+[device creation test](https://github.com/google/iree/blob/main/iree/hal/cts/device_creation_test.cc):
```powershell
# -- CMake --
diff --git a/docs/developer_overview.md b/docs/developer_overview.md
index 7ab969b..f23cd89 100644
--- a/docs/developer_overview.md
+++ b/docs/developer_overview.md
@@ -5,48 +5,48 @@
## Project Code Layout
-[iree/](https://github.com/google/iree/blob/master/iree/)
+[iree/](https://github.com/google/iree/blob/main/iree/)
* Core IREE project
-[integrations/](https://github.com/google/iree/blob/master/integrations/)
+[integrations/](https://github.com/google/iree/blob/main/integrations/)
* Integrations between IREE and other frameworks, such as TensorFlow
-[bindings/](https://github.com/google/iree/blob/master/bindings/)
+[bindings/](https://github.com/google/iree/blob/main/bindings/)
* Language and platform bindings, such as Python
-[colab/](https://github.com/google/iree/blob/master/colab/)
+[colab/](https://github.com/google/iree/blob/main/colab/)
* Colab notebooks for interactively using IREE's Python bindings
## IREE Code Layout
-[iree/base/](https://github.com/google/iree/blob/master/iree/base/)
+[iree/base/](https://github.com/google/iree/blob/main/iree/base/)
* Common types and utilities used throughout IREE
-[iree/compiler/](https://github.com/google/iree/blob/master/iree/compiler/)
+[iree/compiler/](https://github.com/google/iree/blob/main/iree/compiler/)
* IREE's MLIR dialects, LLVM compiler passes, module translation code, etc.
Code here should not depend on anything in the runtime
-[iree/hal/](https://github.com/google/iree/blob/master/iree/hal/)
+[iree/hal/](https://github.com/google/iree/blob/main/iree/hal/)
* **H**ardware **A**bstraction **L**ayer for IREE's runtime, with
implementations for hardware and software backends
-[iree/schemas/](https://github.com/google/iree/blob/master/iree/schemas/)
+[iree/schemas/](https://github.com/google/iree/blob/main/iree/schemas/)
* Shared data storage format definitions, primarily using
[FlatBuffers](https://google.github.io/flatbuffers/)
-[iree/tools/](https://github.com/google/iree/blob/master/iree/tools/)
+[iree/tools/](https://github.com/google/iree/blob/main/iree/tools/)
* Assorted tools used to optimize, translate, and evaluate IREE
-[iree/vm/](https://github.com/google/iree/blob/master/iree/vm/)
+[iree/vm/](https://github.com/google/iree/blob/main/iree/vm/)
* Bytecode **V**irtual **M**achine used to work with IREE modules and invoke
IREE functions
@@ -77,7 +77,7 @@
`FileCheck` should be used to test the generated output.
Here's an example of a small compiler pass running on a
-[test file](https://github.com/google/iree/blob/master/iree/compiler/Dialect/IREE/Transforms/test/drop_compiler_hints.mlir):
+[test file](https://github.com/google/iree/blob/main/iree/compiler/Dialect/IREE/Transforms/test/drop_compiler_hints.mlir):
```shell
$ bazel run iree/tools:iree-opt -- \
@@ -89,7 +89,7 @@
For a more complex example, here's how to run IREE's complete transformation
pipeline targeting the VMLA backend on the
-[fullyconnected.mlir](https://github.com/google/iree/blob/master/iree/test/e2e/models/fullyconnected.mlir)
+[fullyconnected.mlir](https://github.com/google/iree/blob/main/iree/test/e2e/models/fullyconnected.mlir)
model file:
```shell
@@ -100,7 +100,7 @@
```
Custom passes may also be layered on top of `iree-opt`, see
-[iree/samples/custom_modules/dialect](https://github.com/google/iree/blob/master/iree/samples/custom_modules/dialect)
+[iree/samples/custom_modules/dialect](https://github.com/google/iree/blob/main/iree/samples/custom_modules/dialect)
for a sample.
### iree-translate
@@ -123,7 +123,7 @@
```
Custom translations may also be layered on top of `iree-translate`, see
-[iree/samples/custom_modules/dialect](https://github.com/google/iree/blob/master/iree/samples/custom_modules/dialect)
+[iree/samples/custom_modules/dialect](https://github.com/google/iree/blob/main/iree/samples/custom_modules/dialect)
for a sample.
### iree-run-module
@@ -150,7 +150,7 @@
and executes it as a series of
[googletest](https://github.com/google/googletest) tests. This is the test
runner for the IREE
-[check framework](https://github.com/google/iree/tree/master/docs/testing_guide.md#end-to-end-tests).
+[check framework](https://github.com/google/iree/tree/main/docs/testing_guide.md#end-to-end-tests).
```shell
$ bazel run iree/tools:iree-translate -- \
@@ -176,7 +176,7 @@
function as exported by default and running all of them.
For example, to execute the contents of
-[iree/tools/test/simple.mlir](https://github.com/google/iree/blob/master/iree/tools/test/simple.mlir):
+[iree/tools/test/simple.mlir](https://github.com/google/iree/blob/main/iree/tools/test/simple.mlir):
```shell
$ bazel run iree/tools:iree-run-mlir -- \
@@ -210,7 +210,7 @@
### Useful Vulkan driver flags
For IREE's Vulkan runtime driver, there are a few useful
-[flags](https://github.com/google/iree/blob/master/iree/hal/vulkan/vulkan_driver.cc):
+[flags](https://github.com/google/iree/blob/main/iree/hal/vulkan/vulkan_driver.cc):
#### `--vulkan_renderdoc`
diff --git a/docs/repository_management.md b/docs/repository_management.md
index 1d6a342..df769f7 100644
--- a/docs/repository_management.md
+++ b/docs/repository_management.md
@@ -7,12 +7,14 @@
transparency. If any of these things are particularly troublesome or painful for
your workflow, please reach out to us so we can prioritize a fix.
-NOTE: We are currently in the process of migrating our repository to be
-GitHub-first and hide the merging complexity in a separate `google` feature
-branch so that standard development workflows don't have to bear the cost for
-every contribution. During this part of the migration period, please direct PRs
-to the `google` branch (which will be marked as the default branch). See
-https://groups.google.com/d/msg/iree-discuss/F07vsG9Ah4o/uAIusKO-BQAJ.
+## Branches
+
+The default branch is called `main`. PRs should be sent there. We also have a
+`google` branch that is synced from the Google internal source repository. This
+branch is merged from and to `main` frequently to upstream any changes coming
+from `google` and as part of updating the internal source repository. For the
+most part this integration with `google` should have minimal effect on normal
+developer workflows.
## Dependencies
@@ -31,7 +33,7 @@
of truth for the correct submodule state and is what is used by the CI. When
updating the submodule state from Google's source repository, only this file is
updated and another
-[GitHub Actions workflow](https://github.com/google/iree/blob/master/.github/workflows/synchronize_submodules.yml)
+[GitHub Actions workflow](https://github.com/google/iree/blob/main/.github/workflows/synchronize_submodules.yml)
immediately commits a submodule update on top of that.
Shortcut commands (read below for full documentation):
@@ -46,6 +48,11 @@
### The special relationship with LLVM and TensorFlow
+TL;DR: Dependency management with TensorFlow and LLVM are managed as part of the
+integration with Google's internal source repository and updates will be picked
+up by the `main` branch at an approximately daily cadence when merging in the
+`google` branch.
+
Currently, the two most challenging projects to manage as dependencies are
TensorFlow and LLVM. Both are typically pinned to specific versions that are
integrated in the Google source repository at a cadence up to many times per
@@ -67,15 +74,14 @@
source of truth prior to accepting code that depends on it.
LLVM is integrated into Google's source repository at a cadence up to several
-times a day. When this necessitates updating IREE to changing LLVM and MLIR
-APIs, these updates are performed atomically as part of the integration.
-Previously this did not trigger a change to the LLVM submodule version in open
-source, which would result in breakages due to version skew. We are
-transitioning to make each integration of LLVM update IREE's submodule in OSS.
-This means that API updates will have some chance of not breaking the build.
-However, it can introduce breakages as well, especially in the Bazel build where
-we still have to wait to update the BUILD files where necessary. We are
-continuing to work on improving this situation.
+times a day. The LLVM submodule version used by IREE is also updated as part of
+this integration. When this necessitates updating IREE to changing LLVM and MLIR
+APIs, these updates are performed atomically as part of the integration. However
+the changes performed here only guarantee that the internal build is not broken,
+which means that such integrations may very well break the OSS build on the
+`google` branch (in particular, it does not update teh LLVM BUILD files). The
+IREE build cop will fix the breakage on the `google` branch. We are continuing
+to work on improving this situation.
### Tasks:
@@ -147,31 +153,3 @@
Any CI systems or other things that retrieve an arbitrary commit should invoke
this prior to running just to make sure that their git view of the submodule
state is consistent.
-
-#### Updating TensorFlow and LLVM versions
-
-WARNING: These scripts have not been updated to reflect the new tooling to
-automatically update the LLVM hash directly from Google's source repository
-rather than through TensorFlow.
-
-```shell
-# By default, updates third_party/tensorflow to the remote head and
-# third_party/llvm-project to the commit that tensorflow references.
-# Also updates the IREE mirrors of key build files. All changes will
-# be staged for commit.
-./scripts/git/update_tf_llvm_submodules.py
-
-# You can also select various other options:
-# Don't update the tensorflow commit, just bring LLVM to the correct
-# linked commit.
-./scripts/git/update_tf_llvm_submodules.py --tensorflow_commit=KEEP
-
-# Don't update the LLVM commit.
-./scripts/git/update_tf_llvm_submodules.py --llvm_commit=KEEP
-
-# Update to a specific TensorFlow commit (and corresponding LLVM).
-./scripts/git/update_tf_llvm_submodules.py --tensorflow_commit=<somehash>
-
-# Update to a specific LLVM commit.
-./scripts/git/update_tf_llvm_submodules.py --llvm_commit=<somehash>
-```
diff --git a/docs/testing_guide.md b/docs/testing_guide.md
index 8819f54..ce1ea4c 100644
--- a/docs/testing_guide.md
+++ b/docs/testing_guide.md
@@ -11,7 +11,7 @@
### Running a Test
-For the test https://github.com/google/iree/tree/master/iree/base/arena_test.cc
+For the test https://github.com/google/iree/tree/main/iree/base/arena_test.cc
With CMake, run this from the build directory:
@@ -78,7 +78,7 @@
We have created a corresponding CMake function `iree_cc_test` that mirrors the
Bazel rule's behavior. Our
-[Bazel to CMake converter](https://github.com/google/iree/tree/master/build_tools/bazel_to_cmake/bazel_to_cmake.py)
+[Bazel to CMake converter](https://github.com/google/iree/tree/main/build_tools/bazel_to_cmake/bazel_to_cmake.py)
should generally derive the `CMakeLists.txt` file from the BUILD file:
```cmake
@@ -101,7 +101,7 @@
### Running a Test
For the test
-https://github.com/google/iree/tree/master/iree/compiler/Dialect/VMLA/Conversion/HLOToVMLA/test/math_ops.mlir
+https://github.com/google/iree/tree/main/iree/compiler/Dialect/VMLA/Conversion/HLOToVMLA/test/math_ops.mlir
With CMake, run this from the build directory:
@@ -123,7 +123,7 @@
are testing. Instead of `mlir-opt`, use `iree-opt`, which registers IREE
dialects and passes and doesn't register some unnecessary core ones. Instead of
`FileCheck`, use
-[`IreeFileCheck`](https://github.com/google/iree/tree/master/iree/tools/IreeFileCheck.sh),
+[`IreeFileCheck`](https://github.com/google/iree/tree/main/iree/tools/IreeFileCheck.sh),
a shell-script wrapper around FileCheck that passes it a few
`--do-the-right-thing` flags.
@@ -150,7 +150,7 @@
```
There is a corresponding CMake function, calls to which will be generated by our
-[Bazel to CMake Converter](https://github.com/google/iree/tree/master/build_tools/bazel_to_cmake/bazel_to_cmake.py).
+[Bazel to CMake Converter](https://github.com/google/iree/tree/main/build_tools/bazel_to_cmake/bazel_to_cmake.py).
```cmake
file(GLOB _GLOB_X_MLIR LIST_DIRECTORIES false RELATIVE ${CMAKE_CURRENT_SOURCE_DIR} CONFIGURE_DEPENDS *.mlir)
@@ -184,7 +184,7 @@
### Running a Test
For the test
-https://github.com/google/iree/tree/master/iree/test/e2e/xla_ops/floor.mlir
+https://github.com/google/iree/tree/main/iree/test/e2e/xla_ops/floor.mlir
compiled for the VMLA target backend and running on the VMLA driver (here they
match exactly, but in principle there's a many-to-many mapping from backends to
drivers).
@@ -363,7 +363,7 @@
The CMake functions follow a similar pattern. The calls to them are generated in
our `CMakeLists.txt` file by
-[bazel_to_cmake](https://github.com/google/iree/tree/master/build_tools/bazel_to_cmake/bazel_to_cmake.py).
+[bazel_to_cmake](https://github.com/google/iree/tree/main/build_tools/bazel_to_cmake/bazel_to_cmake.py).
## Binding Tests
diff --git a/integrations/tensorflow/e2e/BUILD b/integrations/tensorflow/e2e/BUILD
index 3500d99..bb22ec9 100644
--- a/integrations/tensorflow/e2e/BUILD
+++ b/integrations/tensorflow/e2e/BUILD
@@ -16,7 +16,7 @@
# files. A coverage table generated from this file can be viewed here:
# https://google.github.io/iree/TensorFlowE2ECoverage
# Updates made to test suite names should also be reflected here:
-# https://github.com/google/iree/blob/master/scripts/update_e2e_coverage.py
+# https://github.com/google/iree/blob/main/scripts/update_e2e_coverage.py
load(
"//bindings/python:build_defs.oss.bzl",
@@ -73,9 +73,7 @@
"fill_test.py",
"mandelbrot_test.py",
"matrix_ops_test.py",
- "resource_ops_test.py",
"ring_buffer_test.py",
- "sliding_window_test.py",
"strings_test.py",
]
@@ -88,9 +86,7 @@
"fill_test.py",
"mandelbrot_test.py",
"matrix_ops_test.py",
- "resource_ops_test.py",
"ring_buffer_test.py",
- "sliding_window_test.py",
"strings_test.py",
"tensorlist_test.py",
]
diff --git a/integrations/tensorflow/e2e/keras/BUILD b/integrations/tensorflow/e2e/keras/BUILD
index 7b21bf8..6752669 100644
--- a/integrations/tensorflow/e2e/keras/BUILD
+++ b/integrations/tensorflow/e2e/keras/BUILD
@@ -16,7 +16,7 @@
# files. A coverage table generated from this file can be viewed here:
# https://google.github.io/iree/TensorFlowE2ECoverage
# Updates made to test suite names should also be reflected here:
-# https://github.com/google/iree/blob/master/scripts/update_e2e_coverage.py
+# https://github.com/google/iree/blob/main/scripts/update_e2e_coverage.py
load(
"//bindings/python:build_defs.oss.bzl",
diff --git a/iree/base/shaped_buffer_string_util.h b/iree/base/shaped_buffer_string_util.h
index a4a3622..a351750 100644
--- a/iree/base/shaped_buffer_string_util.h
+++ b/iree/base/shaped_buffer_string_util.h
@@ -14,7 +14,7 @@
// Utilities for parsing and printing ShapedBuffer, mostly useful for
// testing. The format is as described in
-// https://github.com/google/iree/tree/master/iree/base/buffer_string_util.h
+// https://github.com/google/iree/tree/main/iree/base/buffer_string_util.h
#ifndef IREE_BASE_SHAPED_BUFFER_STRING_UTIL_H_
#define IREE_BASE_SHAPED_BUFFER_STRING_UTIL_H_
diff --git a/iree/compiler/Dialect/Flow/Transforms/CMakeLists.txt b/iree/compiler/Dialect/Flow/Transforms/CMakeLists.txt
index 668b29a..568ec4f 100644
--- a/iree/compiler/Dialect/Flow/Transforms/CMakeLists.txt
+++ b/iree/compiler/Dialect/Flow/Transforms/CMakeLists.txt
@@ -42,6 +42,7 @@
MLIRAnalysis
MLIRIR
MLIRPass
+ MLIRShapeOpsTransforms
MLIRStandardOps
MLIRSupport
MLIRTransformUtils
diff --git a/iree/compiler/Dialect/HAL/IR/HALBase.td b/iree/compiler/Dialect/HAL/IR/HALBase.td
index dacc716..f3bb22f 100644
--- a/iree/compiler/Dialect/HAL/IR/HALBase.td
+++ b/iree/compiler/Dialect/HAL/IR/HALBase.td
@@ -225,7 +225,7 @@
}
// Values are FourCC codes.
-// See https://github.com/google/iree/tree/master/iree/hal/executable_format.h for more info.
+// See https://github.com/google/iree/tree/main/iree/hal/executable_format.h for more info.
// https://ubershmekel.github.io/fourcc-to-text/ can be used to generate the
// integer values for inclusion here.
//
@@ -251,7 +251,7 @@
let convertFromStorage = "static_cast<IREE::HAL::ExecutableFormat>($_self.getInt())";
let cppNamespace = "::mlir::iree_compiler::IREE::HAL";
}
-// LINT.ThenChange(https://github.com/google/iree/tree/master/iree/hal/executable_format.h:fourcc)
+// LINT.ThenChange(https://github.com/google/iree/tree/main/iree/hal/executable_format.h:fourcc)
def HAL_ExecutableFormatArrayAttr : TypedArrayAttrBase<
HAL_ExecutableFormatAttr,
diff --git a/iree/compiler/Dialect/HAL/IR/HALTypes.cpp b/iree/compiler/Dialect/HAL/IR/HALTypes.cpp
index 7220a97..9d10e92 100644
--- a/iree/compiler/Dialect/HAL/IR/HALTypes.cpp
+++ b/iree/compiler/Dialect/HAL/IR/HALTypes.cpp
@@ -89,7 +89,7 @@
}
return llvm::None;
}
-// LINT.ThenChange(https://github.com/google/iree/tree/master/iree/hal/api.h:element_type)
+// LINT.ThenChange(https://github.com/google/iree/tree/main/iree/hal/api.h:element_type)
IntegerAttr getElementTypeAttr(Type type) {
auto elementType = getElementTypeValue(type);
diff --git a/iree/hal/api.h b/iree/hal/api.h
index 34a924f..ee11178 100644
--- a/iree/hal/api.h
+++ b/iree/hal/api.h
@@ -500,7 +500,7 @@
typedef uint32_t iree_hal_element_type_t;
// clang-format on
-// LINT.ThenChange(https://github.com/google/iree/tree/master/iree/compiler/Dialect/HAL/IR/HALTypes.h:element_type)
+// LINT.ThenChange(https://github.com/google/iree/tree/main/iree/compiler/Dialect/HAL/IR/HALTypes.h:element_type)
// A dimension within a shape.
typedef int32_t iree_hal_dim_t;
diff --git a/iree/hal/executable_format.h b/iree/hal/executable_format.h
index 013ca42..2f5a678 100644
--- a/iree/hal/executable_format.h
+++ b/iree/hal/executable_format.h
@@ -50,30 +50,30 @@
MakeExecutableFormatID("IREE");
// IREE VMLA executable in FlatBuffer format using the
-// https://github.com/google/iree/tree/master/iree/schemas/vmla_executable_def.fbs
+// https://github.com/google/iree/tree/main/iree/schemas/vmla_executable_def.fbs
// schema.
constexpr ExecutableFormat kExecutableFormatVMLA =
MakeExecutableFormatID("VMLA");
// SPIR-V executable in FlatBuffer format using the
-// https://github.com/google/iree/tree/master/iree/schemas/spirv_executable_def.fbs
+// https://github.com/google/iree/tree/main/iree/schemas/spirv_executable_def.fbs
// schema.
constexpr ExecutableFormat kExecutableFormatSpirV =
MakeExecutableFormatID("SPVE");
// LLVMIR executable in FlatBuffer format using the
-// https://github.com/google/iree/tree/master/iree/schemas/llvmir_executable_def.fbs
+// https://github.com/google/iree/tree/main/iree/schemas/llvmir_executable_def.fbs
// schema.
constexpr ExecutableFormat kExecutableFormatLLVM =
MakeExecutableFormatID("LLVM");
// Dynamic Library (dylib) executable in FlatBuffer format using the
-// https://github.com/google/iree/tree/master/iree/schemas/dylib_executable_def.fbs
+// https://github.com/google/iree/tree/main/iree/schemas/dylib_executable_def.fbs
// schema
constexpr ExecutableFormat kExecutableFormatDyLib =
MakeExecutableFormatID("DLIB");
-// LINT.ThenChange(https://github.com/google/iree/tree/master/iree/compiler/Dialect/HAL/IR/HALBase.td:executable_format)
+// LINT.ThenChange(https://github.com/google/iree/tree/main/iree/compiler/Dialect/HAL/IR/HALBase.td:executable_format)
} // namespace hal
} // namespace iree
diff --git a/iree/hal/vulkan/timepoint_util.cc b/iree/hal/vulkan/timepoint_util.cc
index c212856..98e9f96 100644
--- a/iree/hal/vulkan/timepoint_util.cc
+++ b/iree/hal/vulkan/timepoint_util.cc
@@ -74,9 +74,17 @@
<< "Fence pool out of free fences";
}
- auto* fence = free_fences_.front();
- free_fences_.pop_front();
- return add_ref(fence);
+ // To acquire from the pool, we:
+ // 1) Pop from the front of the queue (reference count of 0);
+ // 2) Release the unique_ptr, since lifetime will be managed by ref counts;
+ // 3) Return as a raw RefObject with a reference count of 1;
+ // When the reference count goes back to 0, it will be returned to the pool,
+ // wrapped with unique_ptr.
+ // When the pool is destroyed, all free fences are freed by unique_ptr
+ // automatically.
+ std::unique_ptr<TimePointFence> fence =
+ free_fences_.take(free_fences_.front());
+ return add_ref(fence.release());
}
void TimePointFencePool::ReleaseResolved(TimePointFence* fence) {
@@ -84,7 +92,7 @@
VkFence f = fence->value();
syms()->vkResetFences(*logical_device_, 1, &f);
absl::MutexLock lock(&mutex_);
- free_fences_.push_back(fence);
+ free_fences_.push_back(absl::WrapUnique(fence));
}
TimePointFencePool::TimePointFencePool(ref_ptr<VkDeviceHandle> logical_device)
@@ -110,7 +118,7 @@
VK_RETURN_IF_ERROR(syms()->vkCreateFence(*logical_device_, &create_info,
logical_device_->allocator(),
&fence));
- fences[i].reset(new TimePointFence(this, fence));
+ fences[i] = absl::make_unique<TimePointFence>(this, fence);
}
}
diff --git a/iree/hal/vulkan/timepoint_util.h b/iree/hal/vulkan/timepoint_util.h
index 7cd6370..4135326 100644
--- a/iree/hal/vulkan/timepoint_util.h
+++ b/iree/hal/vulkan/timepoint_util.h
@@ -153,7 +153,9 @@
absl::Mutex mutex_;
- IntrusiveList<TimePointFence> free_fences_ ABSL_GUARDED_BY(mutex_);
+ // Track via unique_ptr, since IntrusiveList doesn't manage memory itself.
+ IntrusiveList<std::unique_ptr<TimePointFence>> free_fences_
+ ABSL_GUARDED_BY(mutex_);
};
// A pool of `VkSemaphore`s that can be used by `EmulatedTimelineSemaphore` to
diff --git a/iree/samples/models/mnist.mlir b/iree/samples/models/mnist.mlir
index c6ae699..a78c669 100644
--- a/iree/samples/models/mnist.mlir
+++ b/iree/samples/models/mnist.mlir
@@ -1,5 +1,5 @@
// Trained MNIST model generated by
-// https://github.com/google/iree/blob/master/colab/mnist_tensorflow.ipynb.
+// https://github.com/google/iree/blob/main/colab/mnist_tensorflow.ipynb.
//
// Model structure is from tf.keras:
//
diff --git a/iree/test/e2e/regression/BUILD b/iree/test/e2e/regression/BUILD
index 2e29da4..dc9bdb1 100644
--- a/iree/test/e2e/regression/BUILD
+++ b/iree/test/e2e/regression/BUILD
@@ -14,7 +14,7 @@
# Tests for end-to-end IREE support of specific features to prevent regression.
# These should focus on support by IREE itself, not for issues with specific runner tools. Place
-# those tests in https://github.com/google/iree/tree/master/iree/tools/test/
+# those tests in https://github.com/google/iree/tree/main/iree/tools/test/
load("//iree:lit_test.bzl", "iree_lit_test_suite")
diff --git a/iree/tools/vm_util.h b/iree/tools/vm_util.h
index 64023db..c039a1d 100644
--- a/iree/tools/vm_util.h
+++ b/iree/tools/vm_util.h
@@ -46,7 +46,7 @@
// Buffers should be in the IREE standard shaped buffer format:
// [shape]xtype=[value]
// described in
-// https://github.com/google/iree/tree/master/iree/base/buffer_string_util.h
+// https://github.com/google/iree/tree/main/iree/base/buffer_string_util.h
// Uses |allocator| to allocate the buffers.
// Uses descriptors in |descs| for type information and validation.
// The returned variant list must be freed by the caller.
@@ -61,7 +61,7 @@
// Prints buffers in the IREE standard shaped buffer format:
// [shape]xtype=[value]
// described in
-// https://github.com/google/iree/tree/master/iree/base/buffer_string_util.h
+// https://github.com/google/iree/tree/main/iree/base/buffer_string_util.h
// Uses descriptors in |descs| for type information and validation.
Status PrintVariantList(absl::Span<const RawSignatureParser::Description> descs,
iree_vm_variant_list_t* variant_list,
diff --git a/scripts/update_op_coverage.py b/scripts/update_op_coverage.py
index e9e83de..b01b687 100755
--- a/scripts/update_op_coverage.py
+++ b/scripts/update_op_coverage.py
@@ -31,7 +31,7 @@
# TODO(scotttodd): LLVM AOT (dylib-llvm-aot) HAL target(s)
OP_COVERAGE_DESCRIPTION = """# HLO Op Coverage
-There are three backend [targets](https://github.com/google/iree/tree/master/iree/compiler/Dialect/HAL/Target) in IREE:
+There are three backend [targets](https://github.com/google/iree/tree/main/iree/compiler/Dialect/HAL/Target) in IREE:
- vmla
- llvm-ir
