Meet#7 jmorman

GSoC 2025 · GNU Radio 4.0 Block-Set Expansion

Week-7 check-in (mentor — Josh Morman)

Date	26 June 2025, 17:00 IST
Duration	45 min
Participants	Josh Morman · Krish Gupta

---

1 · Progress update

• Analog Blocks
  • Submitted PR gnuradio/gnuradio4#5 with complete Analog block implementations
  • Addressed initial feedback from Ralph about parameter naming conventions
  • Improved documentation for each block with usage examples
• Digital Core Components
  • Implemented key digital core components:
    • Lfsr.hpp - Linear Feedback Shift Register with configurable polynomial
    • Crc.hpp - Cyclic Redundancy Check with common polynomials
    • Scrambler.hpp / Descrambler.hpp - Bit scrambling with LFSR backing
    • First draft of Constellation.hpp template class

---

2 · Analog PR Highlights

The Analog PR includes comprehensive implementations:

• Modulators: FrequencyMod, PhaseModulator
• Demodulators: QuadratureDemod, FmDet, AmDemod
• Gain Control: Agc, Agc2

All blocks follow the agreed-upon patterns:

• Minimal state
• Clean CRTP implementation
• Extensive testing
• Proper reflection for runtime configuration

Initial feedback has been positive, with only minor style adjustments requested.

---

3 · Digital Core Implementation

LFSR Implementation

Implemented a flexible LFSR with customizable polynomials:

template <typename T = uint32_t>
class Lfsr {
private:
    T reg_{1};           // Shift register state
    T mask_{0x00000001}; // Polynomial mask
    T outXor_{0};        // Output XOR mask

public:
    GR_MAKE_REFLECTABLE(Lfsr, mask_, outXor_);
    
    void reset() { reg_ = 1; }
    
    T next() {
        T feedback = 0;
        if (reg_ & 0x1) {
            feedback = mask_;
            reg_ = ((reg_ >> 1) ^ feedback);
        } else {
            reg_ = (reg_ >> 1);
        }
        return reg_ ^ outXor_;
    }
    
    // Support for common polynomials
    static constexpr T POLY_CCITT() { return 0x00008810; }
    static constexpr T POLY_CDMA2K() { return 0x0000cc00; }
};

GR_REGISTER_BLOCK(Lfsr<uint32_t>);

This serves as the foundation for scramblers, random number generation, and more.

---

4 · Constellation Design

Josh and I spent considerable time discussing the ideal Constellation design:

Approach	Pros	Cons
C++23 POD class	Clean memory layout, SIMD-friendly	Requires C++23 for full benefits
Legacy class hierarchy	Compatible with existing code	Virtual overhead, complex inheritance
Template-based approach	Type-safe, optimizable	More complex for users

We settled on a hybrid approach:

• Core Constellation<N> template for new code (C++23 POD)
• Thin legacy adapters for backward compatibility
• Stateless slicing/metrics for better optimization

---

5 · CI Improvements

I implemented several CI improvements to address ongoing challenges:

• Per-subtree CMake targets to allow focused builds
• Test splitting to reduce memory pressure
• Selective type matrices to balance coverage and resource usage

These changes have significantly improved CI stability, especially for complex template code.

---

6 · Action items

Owner	Task
Krish	• Address feedback on Analog PR • Continue Digital core implementations • Begin work on Constellation implementation • Update CI configuration for Digital blocks
Josh	• Review Digital core components • Connect with CI team about forked PR Sonar issues • Provide guidance on Constellation design

---

Next sync: July 3rd to review Digital core progress and discuss Constellation implementation details.

---

7 · Technical Notes

CMake Module Structure

We refined the CMake structure to better support modular development:

# digital/CMakeLists.txt
add_library(gr4_digital INTERFACE)

# Core components (always built)
add_subdirectory(core)
target_link_libraries(gr4_digital INTERFACE gr4_digital_core)

# Optional components (can be built separately)
if(BUILD_DIGITAL_MAPPING)
    add_subdirectory(mapping)
    target_link_libraries(gr4_digital INTERFACE gr4_digital_mapping)
endif()

# ... other components similarly structured

This approach:

• Allows focused CI builds
• Reduces build times during development
• Maintains a clean interface for users