[axi-qspi/dv] Add behavioral QSPI flash model and top-level integration (Part 1/3)

[Raghavan-04]

Summary

Adds a behavioral QSPI flash memory model (uvm/tb/qspi_flash_model.sv) and integrates it into the top-level testbench (uvm/tb/axi_qspi_tb_top.sv) across the external bidirectional SPI pins. This establishes the physical slave endpoint necessary to verify memory-mapped reads, dummy cycles, mode-byte decode, and XIP fast-path transfers. Also resolves host-specific executable paths in the Makefile and documents flash preloading methods in the README.md.

Related issue or proposal

Related to #1 (Part 1/3: Behavioral Flash Model & Top-Level Hookup)

Design notes

• Command & Protocol Decoding: Implements SPI command decoding for Standard Read (0x03), Fast Read (0x0B), Dual Output Read (0x3B), Quad Output Read (0x6B), and Quad I/O Read (0xEB).
• Continuous Read Mode: Decodes mode byte 0xA0 during Quad I/O transactions, allowing subsequent transactions to skip the 8-bit command phase and transition directly to address nibbles.
• I/O Ownership & Timing: Tri-state bidirectional control on io[3:0]. Keeps lines High-Z during input phases (CMD, ADDR, MODE, DUMMY) and drives read data on negedge sclk with proper sampling on posedge sclk.
• Memory Preloading: Provides a byte-addressable internal memory array initialized by default to i[7:0] ^ 8'hA5, configurable via MEM_INIT_FILE parameter ($readmemh) or dynamic runtime task load_mem_file().
• Portability: Sanitized Makefile tool definitions (VERILATOR_BIN, UVM_HOME) to remove hardcoded Windows paths, ensuring compatibility with Linux containers and CI environments.

Verification

• RTL & Model Lint Check:

make lint

Result: Clean compilation with 0 errors and 0 warnings.

• Strict Model Lint Check:

verilator --language 1800-2017 -Wall -Wno-TIMESCALEMOD -Wno-EOFNEWLINE --lint-only uvm/tb/qspi_flash_model.sv

Result: Passed with 0 warnings.

Protocol & Timing Smoke Test:
Ran a cycle-accurate self-checking simulation verifying:

• Standard Read (0x03) at 0x000000 (received 0xA5)
• Fast Read (0x0B) with 8 dummy clocks at 0x000004 (received 0xA1)
• Quad I/O Read (0xEB) with 4 dummy clocks + Mode 0xA0 at 0x000010 (received 0xB5)
• Continuous Quad Read bypassing the command opcode at 0x000020 (received 0x85)
  Result: >>> ALL 4 FLASH PROTOCOL SMOKE TESTS PASSED! <<<

Checklist

• Scope was discussed before implementation when required.
• RTL is synthesizable and follows the repository style.
• Reset, errors, backpressure, and corner cases were considered.
• [] Tests fail without the fix where this is a bug fix.
• [] UVM/assertions/coverage were updated where applicable.
• Lint, compile, simulation, and relevant regressions pass.
• Documentation and register descriptions are current.
• Commits are focused, atomic, and free of unrelated changes.
• New code is original or correctly licensed and attributed.