Updates various guides

- Updates `RELEASE.md` to mention things to be careful of when creating a new release.

- Updates `debugging.md` to be more concise.

Author: pranavm-nvidia

tripy/RELEASE.md

@@ -5,15 +5,21 @@ This document explains how to release a new version of Tripy.
 1. Update version numbers in [`pyproject.toml`](./pyproject.toml) and
     [`__init__.py`](./tripy/__init__.py) (make sure they match!).
 
+    Often, updates to Tripy will also require updates to dependencies,
+    like MLIR-TRT, so make sure to update those version numbers as well.
+
 2. Add a new entry to [`packages.html`](./docs/packages.html).
     This ensures that we will be able to `pip install` Tripy.
 
 3. If there were any other functional changes since the most recent
     L1, make sure to run L1 testing locally.
 
-4. Create a PR with the above two changes.
+4. Create a PR with the above changes.
+
+5. Once the PR created in (4) is merged, **WAIT FOR THE POST-MERGE PIPELINES TO COMPLETE**.
+    This is a very important step as otherwise the release pipeline could fail.
 
-5. Once the PR created in (4) is merged, create a new tag with:
+    Once the post-merge pipelines have succeeded, create a new tag with:
     ```bash
     git tag tripy-vX.Y.Z
     ```

tripy/docs/post0_developer_guides/architecture.md

@@ -3,7 +3,6 @@
 This document explains the overall architecture of Tripy.
 
 
-
 ## Overview
 
 The main technical requirement of Tripy is twofold:

tripy/docs/post0_developer_guides/debugging.md

@@ -1,29 +1,37 @@
-# Debugging MLIR-TensorRT backend
+# Debugging MLIR-TensorRT
 
-1. Install new python bindings for compiler and runtime. Assuming `tripy/mlir-tensorrt` directory exists. No need to update `LD_LIBRARY_PATH`.
+While developing Tripy features, you may need to debug MLIR-TRT code.
+This guide outlines some methods of doing so.
 
-<!-- Tripy: DOC: NO_EVAL Start -->
-	```bash
-	python3 -m pip install --force-reinstall mlir-tensorrt/build/wheels/trt100/**/*.whl
-	```
-<!-- Tripy: DOC: NO_EVAL End -->
 
-2. Set environment flags for debugging:
+## Environment Variables
+
+We include some environment variables to enable extra debugging information from MLIR-TRT:
+
+- `export TRIPY_MLIR_DEBUG_ENABLED=1` will enable debug prints in MLIR-TRT and dump all intermediate IRs to a directory.
+- `export TRIPY_MLIR_DEBUG_PATH=<mlir-debug-path>` sets the directory for IR dumps. The default path is `mlir-dumps`.
+- `export TRIPY_TRT_DEBUG_ENABLED=1` will dump TensorRT engines and their layer information.
+- `export TRIPY_TRT_DEBUG_PATH=<trt-debug-path>` sets the directory for TensorRT dumps. Default path is `tensorrt-dumps`.
 
-- `export TRIPY_MLIR_DEBUG_ENABLED=1` to enable MLIR-TRT debugging. It will enable debugging prints in MLIR-TRT as well as dump all intermediate IRs after each pass.
-- `export TRIPY_MLIR_DEBUG_PATH=<mlir-debug-path>` to set debug path for MLIR-TRT dumps. Default path is `mlir-dumps` under the repo directory. This will create one or more folders named like `module_ins_t1_outs_t2_1`.
-- `export TRIPY_TRT_DEBUG_ENABLED=1` to enable TensorRT debugging. It will dump TensorRT engines and their layer information (if there are any TensorRT built during compilation).
-- `export TRIPY_TRT_DEBUG_PATH=<trt-debug-path>` to set debug path for TensorRT dumps. Default path is `tensorrt-dumps` under the repo directory.
 
+## Using A Debugger
 
-3. Use LLDB for debugging MLIR-TensorRT backend.
-In order to use `lldb` in tripy container, launch the container with extra security options:
+For more involved bugs, it may be helpful to step into MLIR-TRT code.
+To do so, you will need a debug build of MLIR-TRT;
+see [CONTRIBUTING.md](source:/CONTRIBUTING.md)
+for details on using custom builds of MLIR-TRT.
+
+Once you've installed the debug build in the container, you should be able to use `gdb` as normal.
+
+Alternatively, you can use [LLDB](https://lldb.llvm.org/) if you launch the container with extra security options:
 
 <!-- Tripy: DOC: NO_EVAL Start -->
 ```bash
 docker run --gpus all --cap-add=SYS_PTRACE \
-	--security-opt seccomp=unconfined --security-opt apparmor=unconfined \
-	-p 8080:8080 -v $(pwd):/tripy/ -it --rm tripy:latest
+    --security-opt seccomp=unconfined --security-opt apparmor=unconfined \
+    -p 8080:8080 -v $(pwd):/tripy/ -it --rm tripy:latest
 ```
 <!-- Tripy: DOC: NO_EVAL End -->
-See https://forums.swift.org/t/debugging-using-lldb/18046 for more details.
+
+See [this post](https://forums.swift.org/t/debugging-using-lldb/18046) for details on
+why these security options are required.