- comparisons of vectors give masks (boolxN): select(m, a, b), m.any/all/count, & | ^ !;
f32x8(list, i, fill) and v.store_part(list, i) for the end of a list (masked load/store)
- v.shuffle(constant lanes), v.lo(), v.hi(), concat(a, b), table.lookup(indices):
pshufb / vpshufb on x86 (SSSE3 and up), tbl on ARM64 (checked in the assembly, LLVM 18 and 21)
- acc.dot_add(a, b): i32 lanes gain 4 byte products each, exactly; vpdpbusd with AVX-VNNI or AVX-512 VNNI,
usdot/sdot/udot on ARM64, the sum written out elsewhere (pmaddwd); run here with all three paths agreeing
- @cpu("avx2,fma") fn: LLVM target-features for one function (+evex512 for AVX-512, which LLVM 18+ needs);
cpu_has("avx2"): CPUID + XGETBV on x86, hwcaps on Linux ARM64, sysctl on macOS; a wide vector can't be passed
between functions with different @cpu (its registers differ), and the compiler says so
- the compiler asks clang which features --cpu means (its predefined macros, cached), so dot_add picks the instruction
- Nix's clang wrapper dropped -march=native in nix-shell (reproduced): finch tells it NIX_ENFORCE_NO_NATIVE=0, shell.nix too,
and warns if --cpu native still made no difference
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Psc42moPLwP43Cmb9oZqSR
- a * b + c is two roundings on every processor and --cpu: the IR has no fast-math flags, and clang
compiles it with -ffp-contract=on (checked in the assembly: never fused, also with FMA hardware)
- fma(a, b, c) / v.fma(b, c): one rounding, always (llvm.fma)
- muladd(a, b, c) / v.muladd(b, c): the fastest way here (llvm.fmuladd), what v.fma was before
- f32_bits / f32_from_bits / f64_bits / f64_from_bits: bitcasts, also of vectors
- f16_to_f32 / f32_to_f16 (LLVM half), bf16_to_f32 / f32_to_bf16 (integer ops, ties to even, NaN kept)
- docs (en, pl): what is guaranteed about floats, the new built-ins
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Psc42moPLwP43Cmb9oZqSR
- fNxM / iNxM / uNxM up to 512 bits, as LLVM vectors (<8 x float>): LLVM picks SSE, AVX2,
AVX-512 or NEON for them
- f32x8(x) in every lane, f32x8(a, ..., h), f32x8(list, i) loads, f32x8(i32x8 v) converts
- + - * / and the bit operators lane by lane, with a plain number on either side counting for every lane;
v[i], v[i] = x, v.len, == / != (every lane), print and str
- methods: sum min max dot abs sqrt floor ceil round fma store; min/max/abs/clamp lane by lane
- wide vectors (and structs holding them) are loaded and stored with align 16, what malloc and the stack give
- --cpu <name> (and cpu = "..." in finch.toml): native, x86-64-v3, apple-m1, ...
- debug info shows vectors; the language server completes their methods; Kate and VS Code highlight them
- f32 x = -1.5 works: a minus in front of a float written in the code keeps it such a number
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01Psc42moPLwP43Cmb9oZqSR
Also frees main's args and the message of a failing main (valgrind was not clean on two tests).
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Methods are functions whose first parameter is the receiver's address; a method
that changes self can't be called on a temporary or a read-only value, and
calling one on a variable keeps a parameter from being borrowed. An argument
that lives inside the receiver is copied (c.merge(c)). str(x) and print of
arrays and structs go through one formatter, as in the C++ compiler.
boot.sh: 12 passed, 0 failed, 10 skipped.
- fn Struct.method(...) with self by reference, also on C structs
- map[K]V: insertion-ordered hash maps, literals, m[k] += 1, has/get/remove/keys/values
- -> T! / -> !, return error(...), or <value>, or { err }, try; fallible int/float/read_file/write_file/delete_file
- for i, x in list / for k, v in map
- print and str() share one text formatter
- LSP: method and map completion, method signatures; grammar and snippets for VS Code and Kate
- tests: 74 (methods, errors, maps + 11 failure cases); docs at all three levels, en and pl
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- examples/opengl.fch: OpenGL 3.3 core, shaders, VAO/VBO, a spinning triangle (checked offscreen)
- examples/vulkan.fch: instance, GPUs with properties/memory/queues, logical device and queue
- ptr(n) and int(p) conversions for C APIs that pass offsets as pointers
- .ptr works on string literals; a comma after the last call argument is allowed
- a missing header included by an imported header is now named in the error
- the old .fn extension hint is gone (Finch is new, there are no old files)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
.fn is already used by other tools (e.g. Fennel). Modules are now looked up as name.fch;
a leftover name.fn gets a clear message telling you to rename it.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
- struct with default values; constructors Point(1, 2) / Point(x: 1); field access, auto-deref through pointers
- []T arrays with bounds checks, literals, for-each, push/pop/insert/remove/clear/resize/find/contains/join/slice/reverse/sort
- str carries its length: .len, s[i], +, comparison, sub/find/split/trim/upper/lower/replace/repeat/bytes, str(x), int(s), float(s)
- Memory: values own their memory and are freed at the end of their block; assignment copies, temporaries move;
parameters are lent and only copied when the function changes them; defer; new()/free() for manual heap memory
- input(), read_file/write_file/file_exists, exit(), fn main([]str args)
- Finch modules: import name, module.fn(), module.Struct
- C structs from headers (fields, fixed arrays, nested); passed and returned by value with the System V x86-64 ABI;
addr(fn) gives C a callback
- link "file.c" compiles and links your own C code
- -g emits DWARF debug info (gdb/lldb)
- Small C runtime (runtime/finch_rt.c) embedded in the compiler, cached in ~/.cache/finch
- Tests: 58, all clean under valgrind (MEMCHECK=1 tests/run.sh)
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
A small, fast programming language on LLVM with C-like syntax.
- Lexer, recursive-descent parser, single-pass type checker + LLVM IR codegen, O2 pipeline
- Types: int/float/bool/char/str, sized i8..u64/f32/f64, pointers ptr[T] with addr()/.value/null
- Functions (fn ... -> T), if/else, while, for..in, break/continue, Go-style operator precedence
- Implicit conversions only when lossless; literals adapt to their target type
- Runtime checks for division by zero, MIN/-1 and null .value
- C interop: import "header.h" via libclang, link "lib" with pkg-config, link-error advice
- Human-readable errors with source excerpts
- Test suite (34 tests), examples, docs for beginners/technicians/engineers in English and Polish
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
