- Add volatile flag to Load/Store memory operations with instr_load_ex

    and instr_store_ex methods

  - Add FunctionAttrs struct with 15 function-level attributes (nounwind,
    noreturn, cold, hot, willreturn, nosync, nofree, norecurse, readnone,
    readonly, writeonly, inlinehint, alwaysinline, noinline, minsize,
    optsize) and LLVM lowering support

  - Implement GEP type inference with compute_gep_result_type() helper,
    GepIndexKind enum, and GepTypeError. Update instr_gep() to auto-infer
    result type; add instr_gep_ex() for explicit type specification

  - Add comprehensive intrinsics system with Intrinsic enum supporting:
    - Memory: memcpy, memset, memmove
    - Overflow: sadd/uadd/ssub/usub/smul/umul.with.overflow
    - Math: sqrt, sin, cos, exp, log, fabs, floor, ceil, pow, fma, etc.
    - Bit manipulation: ctpop, ctlz, cttz, bitreverse, bswap, fshl, fshr
    - Other: expect, assume, trap, debugtrap

  - Update ROADMAP.md to reflect completed items

Author: edg-l

CLAUDE.md

@@ -164,3 +164,4 @@ The MCJIT execution engine only supports certain parameter/return types reliably
 - Prefer explicit types over inference for public APIs
 - Use `Location` for source mapping on all constructs
 - Clone-friendly types (most types derive Clone)
+- Use `cargo fmt` after changes

ROADMAP.md

@@ -23,27 +23,27 @@ This document outlines the planned features and improvements for IRVM.
 ## Medium Priority (v0.3.0)
 
 ### Intrinsics System
-- [ ] Design intrinsics declaration mechanism
-- [ ] Implement `llvm.memcpy`, `llvm.memset`, `llvm.memmove`
-- [ ] Implement overflow intrinsics (`llvm.sadd.with.overflow`, etc.)
-- [ ] Implement math intrinsics (`llvm.sqrt`, `llvm.sin`, `llvm.cos`, etc.)
-- [ ] Implement bit manipulation intrinsics (`llvm.ctpop`, `llvm.ctlz`, `llvm.cttz`)
+- [x] Design intrinsics declaration mechanism
+- [x] Implement `llvm.memcpy`, `llvm.memset`, `llvm.memmove`
+- [x] Implement overflow intrinsics (`llvm.sadd.with.overflow`, etc.)
+- [x] Implement math intrinsics (`llvm.sqrt`, `llvm.sin`, `llvm.cos`, etc.)
+- [x] Implement bit manipulation intrinsics (`llvm.ctpop`, `llvm.ctlz`, `llvm.cttz`)
 
 ### Function & Call Attributes
-- [ ] Function attributes: `nounwind`, `noreturn`, `cold`, `hot`, `willreturn`, `nosync`, `nofree`
+- [x] Function attributes: `nounwind`, `noreturn`, `cold`, `hot`, `willreturn`, `nosync`, `nofree`
 - [ ] Parameter attributes: `nocapture`, `readonly`, `writeonly`, `dereferenceable(N)`
 - [ ] Return attributes: `noalias`, `noundef`, `dereferenceable(N)`
 - [ ] GC name support
 - [ ] Prefix/prologue data
 
 ### GEP Type Inference
-- [ ] Implement automatic result type computation for GEP
-- [ ] Add `compute_gep_result_type` helper function
-- [ ] Update `instr_gep` to infer result type automatically
-- [ ] Keep `instr_gep_ex` for explicit type specification
+- [x] Implement automatic result type computation for GEP
+- [x] Add `compute_gep_result_type` helper function
+- [x] Update `instr_gep` to infer result type automatically
+- [x] Keep `instr_gep_ex` for explicit type specification
 
 ### Memory Operation Enhancements
-- [ ] Add `volatile` flag to load/store
+- [x] Add `volatile` flag to load/store
 - [ ] Proper alignment validation
 - [ ] Non-temporal hints
 

irvm-lower/src/lib.rs

@@ -247,7 +247,7 @@ mod test {
             &storage,
         )?;
 
-        let ptr_idx1 = func.blocks[entry].instr_gep(
+        let ptr_idx1 = func.blocks[entry].instr_gep_ex(
             ptr_val,
             &[GepIndex::Const(1)],
             ptr_ty,