Add DCE and parallel store optimization for token ordering. (#149)

Author: maleadt

src/compiler/codegen.jl

@@ -4,6 +4,7 @@ include("codegen/utils.jl")
 include("codegen/passes/token_keys.jl")        # TokenKey, TokenRole, ACQUIRE_TOKEN_KEY
 include("codegen/passes/alias_analysis.jl")    # alias_analysis_pass!
 include("codegen/passes/token_order.jl")       # token_order_pass!
+include("codegen/passes/dce.jl")              # dce_pass!
 include("codegen/kernel.jl")
 include("codegen/control_flow.jl")
 include("codegen/statements.jl")

src/compiler/codegen/kernel.jl

@@ -148,6 +148,7 @@ function emit_kernel!(writer::BytecodeWriter, func_buf::Vector{UInt8},
     # Run alias analysis and token ordering pass on the structured IR.
     alias_result = alias_analysis_pass!(sci)
     token_order_pass!(sci, alias_result)
+    dce_pass!(sci)
 
     # Cache the token bytecode type for codegen
     ctx.token_type = Token(tt)