aboutsummaryrefslogtreecommitdiff
path: root/src/compiler
Commit message (Collapse)AuthorAgeFilesLines
* x86-64 jit: inline new_k, recycle_leaf and own fast paths, in o3 regions and ↵Gravatar zinc 13 days1-2/+2
| | | | | | | | | | | | | | plain blocks a call to rt.new_k / rt.recycle_leaf with a constant class, or to rt.own, does its class-list pop or push (or owner store) inline and goes on after the call; an empty list, memdebug, a held range, a block outside memory, a translated word or a type/class mismatch takes the ordinary call. in a region this replaces the o3_invoke round trip. o3 frees of leaf objects now use recycle_leaf too (new_k stamps the type inside o3 since b86db60). ZYNOFASTALLOC=1 turns it off; ZYPERFMAP=1 writes a perf map of translated code. tests/programs/o3_freelist.zc: reuse, zeroing, an empty list, untyped neighbours, memdebug; jit vs ZYNOFASTALLOC vs ZYO3INTERP.
* prover: do not restore length facts after a call-loopGravatar zinc 13 days1-6/+6
| | | | | | | Restoring undirtied lafacts when the body might free was wrong: push can grow an array without dirtying its name, so a[99] after push warned against a stale length 0. Keep the old lafact_keep rule; reorder explain_prover so the constant/affine cases run before the touch() loops.
* prover: len/cross-array/fixed-new/affine proofs; keep facts past call loopsGravatar zinc 13 days1-19/+27
| | | | | | | | Rebased the prover WIP onto main. lafact_keep restores undirtied allocation facts even when the loop body might free (so a later touch() no longer drops every length fact). proven_p uses ptr_still so an outer call keeps the check. entry_fact still supplies the size for the "might free" explain reason. bounds_prover.zc and explain_prover.zc land with test.sh coverage.
* wip: prover gaps (inline len bound, cross-array, rt.new depth, fixed new, ↵Gravatar zinc 13 days4-42/+112
| | | | affine i*n+k); unbuilt and untested
* o3: use new_k (size-class free lists) like plain codeGravatar zinc 13 days1-4/+4
| | | | | | | Constant-size typed new inside o3 now calls rt.new_k instead of always taking new_t/new_raw. new_k and recycle_leaf skip the o3 argument-fatten path the way free/recycle already do. trees o3 new/free on the x86-64 jit dropped from about 4.1 ms to about 2.7 ms per iter (matches plain).
* actor: a declared actor without its own drop gets one that awaits owed ↵Gravatar zinc 13 days1-1/+3
| | | | replies and closes the endpoint; the handler reaches the struct through endpoint.state so the struct isn't shared; spawn hint skips mailbox structs
* o3: an o3 function called from normal code as a receiver method (run.method) ↵Gravatar zinc 13 days1-1/+3
| | | | enters and leaves the region and fattens its result, like a direct call; spec 4.10.3, o3_receiver_method_enter.zc
* o3: break and continue out of an o3 { } block call rt.o3_leave for each ↵Gravatar zinc 13 days1-0/+12
| | | | block they leave (the loop records the o3 depth it began at), so the depth returns to 0 and pending collections and drops run; spec 4.10.3, o3_break_continue_leave.zc
* destructors: hint: x: has drop and is live at spawn; the worker's copy drops ↵Gravatar zinc 13 days2-0/+14
| | | | it too, on x.spawn() of a mailbox struct for each live owned local with a drop; spec 6.15.4, lsp hint check
* destructors: nothing moves out of a value with a drop: pop on an array with ↵Gravatar zinc 13 days1-1/+8
| | | | a drop, or on an array field of a value with a drop, is error: a: can't move out of a value with drop (fields were already borrowed projections); spec 6.15.5, four conformance programs
* destructors: reassigning an m owner that still holds a value frees the old ↵Gravatar zinc 13 days3-4/+43
| | | | value first, running its drop (right side first, then the free, then the store), with hint: x: reassigning frees the old value; a new value that is part of the old one is an error; u, d and captured owners keep the leak error; the new value's size, leaf type and drop are recorded like a declaration's; spec 6.8.5, 6.15.2
* compiler: a //:lines marker starts a new file, so unbounded; and unchecked; ↵Gravatar zinc 13 days1-1/+3
| | | | of the source before it end there: a program joined to the unbounded runtime in a self-hosted run-only image is bounds-checked again (since d55b5fd its index errors went unreported); runtime_image_test expects the trap's exit status 2; spec 4.15.1, 4.x unchecked, 5.15.3
* compiler: the drop globals' comment says mailbox send, so nothing the ↵Gravatar zinc 13 days1-1/+1
| | | | compiler reads mentions actors (test.sh check)
* destructors: heap(v), drop(v, f) / v.drop(f) / v.drop = f, struct drop ↵Gravatar zinc 13 days5-6/+125
| | | | methods, collector drops after the sweep (once, resurrection), freed blocks forget their drop, temporaries drop at statement end with a hint, actor send check; o3: no collection inside a region (pending until the outermost exit), no drops set or ended inside o3; spec 3.11.6, 4.10.3-4, 6.15
* defer: any single statement (if/else, loops, switch, unchecked, unbounded), ↵Gravatar zinc 13 days1-4/+8
| | | | not just an expression or a block
* fndef: b.2.13 unnamed prototype parameters in fewer ops (402197 -> 402112 ↵Gravatar zinc 13 days1-9/+8
| | | | words, same behaviour)
* detach doc comments on internal bounds helpers (genome size); todo: reuse ↵Gravatar zinc 13 days1-0/+19
| | | | capture_scan in the body scanner
* explain-checks: the record writer moves to src/lib/explain.zc, embedded and ↵Gravatar zinc 13 days3-54/+14
| | | | compiled only in that mode; the genome keeps the call sites and a hook
* bounds: body scan, loop-entry hoisting, pointer proofs from requested size, ↵Gravatar zinc 13 days7-13/+391
| | | | --explain-checks v1 (non-o3 code)
* zy: nonzero exit on compile errorsGravatar zinc 13 days1-3/+5
|
* rt: exact-size heap bounds; n, the words new asked for, lives in bits 32..62 ↵Gravatar zinc 13 days1-2/+2
| | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | | of p[-4] the bnd recipe and paylen read n instead of the power-of-two block: bnd: r6 = sub p 3; r6 = ld r6; r7 = xor p 0x2B7E1516; r6 = eq r6 r7; r7 = sub p 4; r7 = ld r7; r7 = shr r7 32; r7 = mul r7 q; r8 = mul i m; r8 = ltu r8 r7; r8 = xor r8 1; r0 = and r6 r8 paylen: return p[-4] >> 32 two loads instead of three; osize, bchk, release, recycle and the memdebug leak report all read n (rt.new(4) has length 4, rt.new(0) has 0). writers: new_raw and new_k set n; stack.h line 87 and tinyvm.ts (BigInt) set it on stack segments. release clears n when a block goes into quarantine (len 0, indexing traps; under memdebug bchk reports a use after free); every free-list path (fpush, mkblock, bdrop, recycle_leaf, balloc hold) rewrites the word as k << 16, and an array that grows gets a fresh block. a request over 2^31-1-HDR words stops in new_raw (oom), and the compiler only takes the new_k fast path for constant sizes that fit, so n is never truncated. bit users checked (all mask or preserve bits 32..62; none compares or copies the whole word): gc mark and sweep; release, free, recycle, recycle_leaf; share and own; isobj, quar, checkptr, obj_base; bclass and the bfree buddy merge; balloc, mkblock, fpush, funlink, heap_grow; frame_grow; stack_allocate in src/core/stack.h and src/js/tinyvm.ts; micro.c and core.c (check word only); the compiler's two p[-4] | F_INTERNAL writes (base.zc, builtins.zc). 32-bit ports: none (baremetal, wasm and micro use int64 Cell). tests: tests/programs/exactsize.zc (n-1/n for non-power-of-two sizes, new(0), packed bytes, recycle, grow, stack growth); 8-runtime: exact_size_bounds, pointer_index_n_minus_1, pointer_index_n_traps, pointer_index_n_traps_memdebug, new_zero_index_traps, new_too_big_traps, stale_index_after_free_traps, stale_index_after_free_memdebug. expectations changed: leak report shows n (tests/test.sh line 169, examples/learn/13-memdebug.out, docs/LEARN.md, 4-statements/memdebug_report_line.zc). programs adjusted to keep test.sh expectations: quarantine.zc (unbounded), bounds_heap.zc (rt.new(11)). spec 3.7.4 and 4.11 updated. fixpoint 402204 words (main 402197).
* len: len on a char* is an error with typecheck off too (appendix b.2.14)Gravatar zinc 13 days1-1/+2
|
* prototypes: parameters may be unnamed, int f(int, int*); (appendix b.2.13)Gravatar zinc 13 days1-2/+11
|
* bool: normalising into a bool keeps an incompatible source type, so the ↵Gravatar zinc 13 days1-1/+2
| | | | return type check still fires (fixes typeerrors/ret.zc after b.2.6)
* lexer: a block comment that is never closed is 'unterminated comment' ↵Gravatar zinc 13 days1-0/+2
| | | | (appendix b.2.11)
* bool: a value stored into a bool is normalised to 0 or 1 (appendix b.2.6)Gravatar zinc 13 days1-0/+5
|
* enum: a member is defined as soon as it is declared, so later values can ↵Gravatar zinc 13 days2-2/+10
| | | | name it (appendix b.2.2)
* globals: an integer constant initializing a float or f32 global converts to ↵Gravatar zinc 13 days1-1/+2
| | | | float (float g = 2; is 2.0)
* Merge branch 'main' into work/specGravatar zinc 13 days9-37/+140
|\
| * merge bounds checkingGravatar zinc 13 days9-37/+140
| |\
| | * bounds checking: constant loop bound proof works, plain-op fixed-array ↵Gravatar zinc 13 days1-6/+8
| | | | | | | | | | | | check, docs (readme, spec 4.15 and friends, server, vscode), conformance and test.sh fixes, todo
| | * bounds checking: port agent/bounds-checking onto main 0879d0a; unbounded ↵Gravatar zinc 13 days9-36/+137
| | | | | | | | | | | | main in o3call, o3 tests expect no check, boundsrules, bounds_fixed, bounds_heap and zc0unbounded tests
* | | ternary: an int side and a float side convert the int on its own path only ↵Gravatar zinc 13 days1-3/+7
| | | | | | | | | | | | (appendix b.2.5)
* | | include: written inside a function body is 'include is top level only' ↵Gravatar zinc 13 days1-0/+1
| | | | | | | | | | | | (appendix b.2.9)
* | | globals: a float or f32 global takes a float literal initializer (appendix ↵Gravatar zinc 13 days1-0/+3
| | | | | | | | | | | | b.2.7)
* | | fndef: a prototype after a definition in the same module keeps the ↵Gravatar zinc 13 days1-4/+5
| | | | | | | | | | | | definition (appendix b.2.4)
* | | fndef: a second definition of a function in the same module is 'is defined ↵Gravatar zinc 13 days1-0/+4
| | | | | | | | | | | | twice' (appendix b.2.3)
* | | unary: u*x, m*x, d*x are not recasts when the tag names a global variable eitherGravatar zinc 13 days1-1/+6
| | |
* | | closures: allow 24 parameters, like functions (was 19)Gravatar zinc 13 days1-2/+2
|/ /
* | merge: seven self-hosted compiler fixes from conformanceGravatar zinc 13 days4-7/+8
|\ \ | |/ |/|
| * switch: case member(): with no names matches a tagged member by tag even ↵Gravatar zinc 13 days1-1/+1
| | | | | | | | when it has fields (spec 4.7.3)
| * memdebug: accept 'memdebug scrub;' (the check compared against rt.scrub); ↵Gravatar zinc 13 days1-1/+1
| | | | | | | | the scrub program reads the freed word through asm
| * return: a value returned from a void function is dropped and the function ↵Gravatar zinc 13 days1-1/+1
| | | | | | | | yields 0 (typecheck off, spec 4.8.3)
| * fndef: allow 24 parameters; the limit check ran after the 24th was countedGravatar zinc 13 days1-1/+1
| |
| * casts: a cast clears the byte-string flag unless the type is bstr, so ↵Gravatar zinc 13 days1-0/+1
| | | | | | | | (int*)s indexes words
| * unary: u*x, m*x, d*x are not recasts when the tag names a visible local ↵Gravatar zinc 13 days1-1/+1
| | | | | | | | (spec 1.6.3)
| * capture_scan: restore panic after the dry scan, so a bad literal in a ↵Gravatar zinc 13 days1-2/+2
| | | | | | | | function's first statement is reported
* | line tables and backtraces: port agent/line-tables onto main, op 56 on the ↵Gravatar zinc 13 days8-9/+163
|/ | | | x86-64 jit, docs, measurements, tests; default genome reservation 0x68000
* Merge the defer statement; todo: tick it, add its follow-upsGravatar zinc 13 days6-2/+117
|\
| * zc: add the defer statementGravatar zinc 13 days6-2/+117
| | | | | | | | | | | | | | | | | | | | | | | | defer expr; and defer { ... } register code that runs when the enclosing block ends: at the end of the block, and at return, break and continue that leave it. the tokens are kept at the defer and compiled again at each exit (block_close, body, kw_return, kw_break, kw_continue), newest first, before the block's variables are released, so the analyzer sees a deferred rt.free(p) as the transfer of p where it runs. names mean what they meant at the defer; values are read when it runs. return, break and continue inside a deferred statement are errors, reported once. works in unchecked code and closures (own base per function context).
generated by cgit v1.2.3 (git 2.39.1) at 2026-10-27 20:14:21 +0000