Enable e2e lowering of neura-compiler with arch

[BenkangPeng]

Related issue: #189

This PR

• Enables end-to-end lowering pipeline in neura-compiler
• Adds support for specifying the architecture_spec via command-line flags
• The YAML parsing logic has been refactored and moved out of passes.

TODO:

• Pass the architecture into CodeGenPass.

[Copilot]

Pull request overview

This PR enables end-to-end lowering in the neura-compiler by adding support for specifying architecture specifications via command-line flags. The YAML parsing logic has been refactored and extracted from MapToAcceleratorPass into reusable utility modules (ArchParser and ParserUtils), improving code organization and maintainability.

Changes:

• Added command-line argument parsing for --architecture-spec flag in both neura-compiler and mlir-neura-opt
• Refactored YAML parsing into dedicated utility classes (ArchParser and ParserUtils) for better code reuse
• Updated the compiler pass pipeline to include MapToAcceleratorPass and GenerateCodePass for complete end-to-end lowering

Reviewed changes

Copilot reviewed 15 out of 15 changed files in this pull request and generated 16 comments.

Show a summary per file

File	Description
tools/neura-compiler/neura-compiler.cpp	Added manual command-line parsing for --architecture-spec flag and implementation of getArchitecture() function
tools/mlir-neura-opt/mlir-neura-opt.cpp	Added getArchitecture() function to provide architecture access to passes
test/visualize/test2.mlir	Updated test to use --architecture-spec flag and updated FileCheck assertions for new output format
lib/NeuraDialect/Util/ParserUtils.cpp	New file containing refactored YAML parsing utility functions
lib/NeuraDialect/Util/ArchParser.cpp	New file containing ArchParser class for parsing architecture YAML files
lib/NeuraDialect/Transforms/MapToAcceleratorPass.cpp	Removed inline YAML parsing code, now uses getArchitecture() function
lib/NeuraDialect/Transforms/GenerateCodePass.cpp	Added max_ctrl_mem_items parameter to Architecture constructor call
lib/NeuraDialect/NeuraPasses.cpp	Added MapToAcceleratorPass and GenerateCodePass to the neura-conversion pipeline
lib/NeuraDialect/CMakeLists.txt	Added new source files (ArchParser.cpp and ParserUtils.cpp) to build
lib/NeuraDialect/Architecture/Architecture.cpp	Added max_ctrl_mem_items parameter to constructor and getter method
include/NeuraDialect/Util/ParserUtils.h	New header file declaring parser utility functions
include/NeuraDialect/Util/ArchParser.h	New header file declaring ArchParser class
include/NeuraDialect/Architecture/Architecture.h	Added max_ctrl_mem_items parameter to constructor, getter method, and getArchitecture() function declaration
.github/workflows/main.yml	Optimized git clone with --depth 1 and --filter=blob:none for faster CI builds

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[tancheng]

Are we able to create e2e (e.g., create a compiler_e2e folder) test using this compiler? (instead opt invoking each pass)

[BenkangPeng]

Add it in 62daa94