Wiki: Create Spoke Page for JIT - L1-Compiler-Deep-Dive

[LSantha]

Goal

Expand the L1 compiler entry into a full deep dive page.

Instructions

1. Use the Skill: Apply update-wiki.
2. Note: This concept is already on the index (pointing to Glossary). You MUST UPDATE the index row to point to your new page.
3. Research: Analyze org.jnode.vm.x86.compiler.l1.L1Compiler.
4. Content: Describe the stack-to-stack compilation strategy, register usage, and the limited optimization passes.

[LSantha]

/oc Please proceed with this task.

[opencode-agent]

Done. Created .wiki/L1-Compiler-Deep-Dive.md with:

• VirtualStack: Delayed emission strategy keeping values in registers within basic blocks
• Item state machine: CONSTANT → GPR → STACK transitions with load()/spill() methods
• Register allocation: GPR pools for 32-bit (EAX→EBX priority) and 64-bit (RAX→R15), plus XMM/FPU pools; reserved registers for statics/processor
• Long/double handling: Split register pairs in 32-bit, single register in 64-bit
• Optimization passes: Register exploitation, constant folding, operand ordering, method inlining via OptimizingBytecodeVisitor — no IR, no graph coloring
• Gotchas: Basic block barriers, aliasing restriction, ECX for shifts, EAX/RAX for returns, FPU stack discipline

Updated index.md (Pages, Concepts, Source Path, Task Routing tables), Glossary.md, JIT-Compilers.md (Related Pages), and Changelog.md. Issue #38 closed.

Wiki URL: https://github.com/LSantha/jnode_ai/wiki/L1-Compiler-Deep-Dive

opencode session  |  github run

[LSantha]

/oc Please proceed with this task.

[opencode-agent]

Done. Created .wiki/L1-Compiler-Deep-Dive.md with:

• VirtualStack: Delayed emission strategy keeping values in registers within basic blocks
• Item state machine: CONSTANT → GPR → STACK transitions with load()/spill() methods
• Register allocation: GPR pools for 32-bit (EAX→EBX priority) and 64-bit (RAX→R15), plus XMM/FPU pools; reserved registers for statics/processor
• Long/double handling: Split register pairs in 32-bit, single register in 64-bit
• Optimization passes: Register exploitation, constant folding, operand ordering, method inlining via OptimizingBytecodeVisitor — no IR, no graph coloring
• Gotchas: Basic block barriers, aliasing restriction, ECX for shifts, EAX/RAX for returns, FPU stack discipline

Updated index.md (Pages, Concepts, Source Path, Task Routing tables), Glossary.md, JIT-Compilers.md (Related Pages), and Changelog.md. Issue #38 already closed.

Wiki URL: https://github.com/LSantha/jnode_ai/wiki/L1-Compiler-Deep-Dive

opencode session  |  github run