summaryrefslogtreecommitdiff
path: root/ir.c
AgeCommit message (Collapse)Author
2025-10-28speculative change to node_newv() to reduce fragmentationWormHeamer
2025-10-27fix a bunch of warningsWormHeamer
2025-10-27small size optimization on <= 1-node NodeListWormHeamer
2025-10-27fix node_set_in() setting always index 0WormHeamer
2025-10-27begin work on nodelistsWormHeamer
2025-08-30separate IR graph parts of Proc into a Graph structWormHeamer
2025-08-27add scratch arena to ProcWormHeamer
2025-08-10better error reporting around uninitialized valuesWormHeamer
2025-08-10fix peephole bug (communative(phi(a,b), phi(a,b)) =/= communative(a,b))WormHeamer
2025-08-10add type-specifiers to let, N_UNINIT for uninitialized valuesWormHeamer
2025-08-10separate peephole optimization outWormHeamer
2025-08-09add N_DEAD node type, and assert to help prevent use-after-freeWormHeamer
2025-08-08slightly tweak node_set_inWormHeamer
2025-08-08attempt at optimizing regions after graph generation....WormHeamer
2025-08-08rework some stuff, maybe bugfix, loosen associative typecheckingWormHeamer
2025-08-07hopefully make multiple returns form sensible graphsWormHeamer
2025-08-07i thiiiink if statement peepholes work now?WormHeamer
2025-08-07fix region nodes being culled prematurelyWormHeamer
2025-08-07preliminary peephole optimization of if statementsWormHeamer
2025-08-07add if statementsWormHeamer
2025-08-05more convenience macrosWormHeamer
2025-08-05cmp(a - b, b - a) -> cmp(a, b) [dubious]WormHeamer
2025-08-05add a bunch of rules to simplify comparisonsWormHeamer
2025-08-05better node_equiv for associative ops; not(cmp(a,b)) -> oppcmp(a,b)WormHeamer
2025-08-05convert divisions to right-shiftsWormHeamer
2025-08-05correction! equality is communative, but not _associative_WormHeamer
so constants can still be shuffled from left to right, not(X) from right to left, but the regrouping optimizations don't apply
2025-08-05tragic realization: = and <> aren't communative at allWormHeamer
i assumed they were because it kinda makes sense to write something like "a = b = c", but "a = (b = c)" and "(a = b) = c" are very very different expressions. whoops!
2025-08-05convert a + -lit -> a - lit; some add/sub inequivalence identitiesWormHeamer
2025-08-05known-not-equivalent identities: a [+/-] lit[x, x =/= 0]WormHeamer
2025-08-04actually do op(not(X), Y) -> op(Y, not(X)) to avoid loopingWormHeamer
2025-08-04shuffle communative op(not(X), Y) -> op(Y, not(X))WormHeamer
2025-08-04NODE_KEEP macroWormHeamer
2025-08-04fix bug of lhs getting culled if same node optimized out of rhsWormHeamer
2025-08-04make NodeInputs statically sized, peephole a + ~a and a ^ ~aWormHeamer
2025-08-04stuffsWormHeamer
2025-08-04fix communative peepholes trampling type boundariesWormHeamer
2025-08-04add optimizations based on only partially known valuesWormHeamer
2025-08-04use lisp acronyms to express peephole optimizationsWormHeamer
2025-08-04another communative operator peepholeWormHeamer
2025-08-04add projection nodes, fix peephole optimizationWormHeamer
2025-08-03add booleans and comparison operatorsWormHeamer
2025-08-03expand and clarify type system, fix double-free of nodesWormHeamer
2025-08-03add assignment statementsWormHeamer
2025-08-03factor out node stuff into ir.c & ir.hWormHeamer
generated by cgit v1.2.3 (git 2.46.0) at 2025-11-15 09:58:34 +0000