John Carmack on mutable variables

For performance critical code, you want to reuse L1D cache lines as much as possible. In many cases, allocation of a new immutable object boils down to malloc(). Newly allocated memory is unlikely to be found on L1D cache. OTOH, replacing data in recently accessed memory and reusing the memory is very likely to become L1D cache hit in runtime.

"Immutability" from a programming language statement perspective doesn't necessarily imply that the implementation duplicates memory or variables.

Similar to how "tail recursion can (usually) be lifted/lowered to a simple loop...", immutability from language statements can often be "collapsed" into mutating a single variable, and there may be one or two "dances" you need to do to either add helper functions, structure your code _slightly_ differently to get there, but it's similar to any kind of performance sensitive code.

Example foo(bar(baz(), bar_alt(baz_alt(), etc...))) [where function call nesting is "representing" an immutability graph] ...yeah, that'd have a lot of allocations and whatever.

But: foo().bar().bar_alt().baz().baz_alt().etc(...) you could imagine is always just stacking/mutating the same variable[s] "in place".

...don't get hung up on the syntax (it's wildly wrong), but imagine the concept. If all the functions "in the chain" are pure (no globals, no modifications), then they can be analyzed and reduced. Refer back to the "Why SSA?" article from a week or two ago: https://news.ycombinator.com/item?id=45674568 ...and you'll see how the logical lines of statements don't necessarily correspond to the physical movement of memory and registers.

You’re describing an edge case. Generally speaking, memory is only reused after old objects are deallocated. And here’s the relevant quote from the OP’s post:

> Having all the intermediate calculations still available is helpful in the debugger

For performance critical code, you wouldn't use malloc()-allocation at all, though whether using an arena allocator or putting stuff on the stack, your argument is still sane. Data locality is speed.