Chez Scheme 的发展

custom linker

flat closures

stack-based representation of control

boxing of assigned variables

BiBOP typing with reserved (16-bit) fixnum range (BiBOP, big bag of pages, where memory is broken up into fixed-size segments and a separate segment table is used to identify the type of object contained in each segment)

stop-and-copy collector

code pointers cached in symbol code-pointer slot

peephole optimization

multiple back ends

multiple operating systems

automatic closing of files by the collector (坏主意！)

optimizing letrec expressions and loops

inlining of primitives

destination-driven code generation (obviating peephole optimizer): Destination-driven code generation is a simple top-down technique that allows the code generated for a program phrase to depend upon its context in an abstract syntax tree. The context is encapsulated in a data destination and a control destination. The data destination specifies where the value computed by an expression is to be stored, while the control destination specifies where program execution is to resume after computation of the value. (非常像 CP0 原型)

faster compiler

faster collector

constant-time continuation operations and stack-overflow recovery (Use segmented stacks. Tracing distinguish tail calls by comparing continuations)

various new compiler optimizations

faster compilation

RISC architecture ports

hybrid tagging model (tagging pointer mechanism from T dialect, BiBOP to identify whether a value contains a pointer or being mutable)

inline allocation (bump allocation)

closure length moved from closures to code headers

literal references moved from closures to code stream

flonum size cut in half

generational garbage collection

intraprocedural register allocation

local call optimizations (unnecessary procedure? checks and indirect jumps → direct jumps)

more compile-time checks

improved register allocation (Register allocation using lazy saves, eager restores, and greedy shuffling)

improved code generation

25X faster gensym (delay generation until the symbol is printed)

fast interpreter (为了在编译后的产物里支持 eval)

letrec optimization

procedure inlining (CP0)

faster continuation handling (one-shot continuations)

improved float printing

thread support

incompatible record checks

letrec/letrec* violation checks

compile-time format-string checks and compilation

unquote and unquote-splicing extension to zero+ subforms

various run-time library and code generation improvements

syntax-case performance improvements

improved bignum arithmetic speed