From ca214af2684b3c6e6982d1fe8be4e7ecc0849b00 Mon Sep 17 00:00:00 2001 From: eunhhu Date: Mon, 19 Oct 2026 02:28:34 +0100 Subject: 140's 3, inflate (RFC 1951), written in aup with no new opcode (aup's examples/inflate.aup, codes read a bit at a time from canonical counts as zlib's puff.c): 640 streams zlib made (6,488,480 bytes: random bytes, words, runs and repeats 32,768 back; levels 0, 1, 6, 9 and zlib's five strategies, so stored, fixed and dynamic blocks) inflated to the original's bytes on the three paths, the paths' outputs the same bytes; 240 damaged streams (cut, a bit flipped) read as inflate.py (the Python reference) reads them, 165 refused by both and 75 the same bytes; module_specs in make check: 144 streams on the three paths, 21 cut ones failing and leaving the last bytes; bars: written (no new opcode), three workarounds (lists passed in records: a fn takes no list; the output a list of one: a bytes value is no state; a block's tables built in a fn: a let's list lives in the handler's own block), 132 lines against Python's 81, 1,152 ns an output byte against the same algorithm in C 10.7 (108x) and zlib 3.5 (329x), pure Python 1,214; 191 instructions a byte (a block fn called for each bit was 246 and 1,553 ns: small block fns' inlining is 2c's leftover); the four timing tools' message calls fixed for 4b's API (no keys); CHECKPOINT's direction row --- research/experiments/044-specs/inflate_time.c | 231 ++++++++++++++++++++++++++ 1 file changed, 231 insertions(+) create mode 100644 research/experiments/044-specs/inflate_time.c (limited to 'research/experiments/044-specs/inflate_time.c') diff --git a/research/experiments/044-specs/inflate_time.c b/research/experiments/044-specs/inflate_time.c new file mode 100644 index 0000000..34cc2d5 --- /dev/null +++ b/research/experiments/044-specs/inflate_time.c @@ -0,0 +1,231 @@ +/* Research 140: inflate (RFC 1951) in aup (aup's examples/inflate.aup, run by Pure) against the same algorithm in C + * (codes read a bit at a time from canonical counts, as zlib's puff.c) and against zlib's inflate (tables), on the + * same raw deflate stream. Each output must be the original's bytes; then each is timed (the median of seven runs of + * rounds inflations). + * + * cc -O2 -std=c17 -I../../../core inflate_time.c ../../../build/libpure.a -lz -lm -pthread -o inflate_time + * ./inflate_time inflate.pmod STREAM ORIGINAL ROUNDS (PURE_JIT and PURE_BATCH choose Pure's path) + * + * One JSON line: the sizes, the nanoseconds per output byte of each, Pure's ratios to both and the instructions Pure + * counted for one stream. */ +#define _POSIX_C_SOURCE 200809L +#include "module.h" +#include +#include +#include +#include +#include + +/* ---- inflate in C, bit by bit (puff's algorithm, no tables) */ +typedef struct { + const unsigned char *in; + size_t size, at; /* at in bits */ + unsigned char *out; + size_t used, room; +} Stream; +typedef struct { short counts[16], symbols[288]; } Code; + +static int take(Stream *s, int n) { + int v = 0; + for (int k = 0; k < n; ++k, ++s->at) { + if (s->at >> 3 >= s->size) return -1; + v |= (s->in[s->at >> 3] >> (s->at & 7) & 1) << k; + } + return v; +} + +static void code(Code *c, const short *lengths, int n) { + memset(c->counts, 0, sizeof(c->counts)); + for (int s = 0; s < n; ++s) c->counts[lengths[s]]++; + short offsets[16] = {0}; + for (int l = 1; l < 15; ++l) offsets[l + 1] = (short)(offsets[l] + c->counts[l]); + for (int s = 0; s < n; ++s) + if (lengths[s]) c->symbols[offsets[lengths[s]]++] = (short)s; +} + +static int decode(Stream *s, const Code *c) { + int value = 0, first = 0, index = 0; + for (int n = 1; n < 16; ++n) { + int b = take(s, 1); + if (b < 0) return -1; + value |= b; + if (value - first < c->counts[n]) return c->symbols[index + value - first]; + index += c->counts[n]; + first = (first + c->counts[n]) << 1; + value <<= 1; + } + return -1; +} + +static const short length_base[29] = {3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 23, 27, 31, 35, 43, 51, 59, 67, 83, + 99, 115, 131, 163, 195, 227, 258}; +static const short length_extra[29] = {0, 0, 0, 0, 0, 0, 0, 0, 1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4, 5, 5, 5, 5, 0}; +static const short distance_base[30] = {1, 2, 3, 4, 5, 7, 9, 13, 17, 25, 33, 49, 65, 97, 129, 193, 257, 385, 513, 769, + 1025, 1537, 2049, 3073, 4097, 6145, 8193, 12289, 16385, 24577}; +static const short distance_extra[30] = {0, 0, 0, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 11, 11, + 12, 12, 13, 13}; + +static int put(Stream *s, unsigned char b) { + if (s->used == s->room) return -1; + s->out[s->used++] = b; + return 0; +} + +/* The bytes of a raw deflate stream; -1 if it is none (or the room is short). */ +static long puff(const unsigned char *in, size_t size, unsigned char *out, size_t room) { + Stream s = {in, size, 0, out, 0, room}; + static const short order[19] = {16, 17, 18, 0, 8, 7, 9, 6, 10, 5, 11, 4, 12, 3, 13, 2, 14, 1, 15}; + int last; + do { + last = take(&s, 1); + int kind = take(&s, 2); + if (last < 0 || kind < 0 || kind == 3) return -1; + if (kind == 0) { + size_t b = (s.at + 7) >> 3; + if (b + 4 > size) return -1; + size_t n = in[b] | in[b + 1] << 8; + if (n != ((in[b + 2] | in[b + 3] << 8) ^ 0xffffu) || b + 4 + n > size || s.used + n > room) return -1; + memcpy(out + s.used, in + b + 4, n); + s.used += n; + s.at = (b + 4 + n) * 8; + continue; + } + Code lc, dc; + short lengths[320]; + if (kind == 1) { + for (int i = 0; i < 288; ++i) lengths[i] = i < 144 ? 8 : i < 256 ? 9 : i < 280 ? 7 : 8; + code(&lc, lengths, 288); + for (int i = 0; i < 30; ++i) lengths[i] = 5; + code(&dc, lengths, 30); + } else { + int literals = take(&s, 5) + 257, count = take(&s, 5) + 1, n = take(&s, 4) + 4; + short small[19] = {0}; + for (int k = 0; k < n; ++k) small[order[k]] = (short)take(&s, 3); + Code c; + code(&c, small, 19); + int got = 0; + while (got < literals + count) { + int symbol = decode(&s, &c); + if (symbol < 0) return -1; + if (symbol < 16) { lengths[got++] = (short)symbol; continue; } + short repeat = 0; + int times; + if (symbol == 16) { + if (!got) return -1; + repeat = lengths[got - 1], times = 3 + take(&s, 2); + } else times = symbol == 17 ? 3 + take(&s, 3) : 11 + take(&s, 7); + if (got + times > literals + count) return -1; + while (times--) lengths[got++] = repeat; + } + code(&lc, lengths, literals); + code(&dc, lengths + literals, count); + } + for (;;) { + int symbol = decode(&s, &lc); + if (symbol < 0) return -1; + if (symbol == 256) break; + if (symbol < 256) { if (put(&s, (unsigned char)symbol)) return -1; continue; } + symbol -= 257; + if (symbol >= 29) return -1; + int length = length_base[symbol] + take(&s, length_extra[symbol]); + symbol = decode(&s, &dc); + if (symbol < 0 || symbol >= 30) return -1; + size_t distance = (size_t)(distance_base[symbol] + take(&s, distance_extra[symbol])); + if (distance > s.used) return -1; + while (length--) if (put(&s, s.out[s.used - distance])) return -1; + } + } while (!last); + return (long)s.used; +} + +static long zlib_inflate(const unsigned char *in, size_t size, unsigned char *out, size_t room) { + z_stream z = {0}; + if (inflateInit2(&z, -15) != Z_OK) return -1; + z.next_in = (unsigned char *)in, z.avail_in = (uInt)size, z.next_out = out, z.avail_out = (uInt)room; + int r = inflate(&z, Z_FINISH); + long used = (long)z.total_out; + inflateEnd(&z); + return r == Z_STREAM_END ? used : -1; +} + +static double now_ns(void) { + struct timespec t; + clock_gettime(CLOCK_MONOTONIC, &t); + return (double)t.tv_sec * 1e9 + (double)t.tv_nsec; +} + +static int by_value(const void *a, const void *b) { + double x = *(const double *)a, y = *(const double *)b; + return (x > y) - (x < y); +} + +static unsigned char *whole(const char *path, size_t *size) { + FILE *f = fopen(path, "rb"); + if (!f) return NULL; + fseek(f, 0, SEEK_END); + *size = (size_t)ftell(f); + rewind(f); + unsigned char *bytes = malloc(*size ? *size : 1); + if (fread(bytes, 1, *size, f) != *size) { free(bytes); bytes = NULL; } + fclose(f); + return bytes; +} + +int main(int argc, char **argv) { + if (argc != 5) { fprintf(stderr, "usage: %s MODULE STREAM ORIGINAL ROUNDS\n", argv[0]); return 2; } + size_t size, original_size; + unsigned char *stream = whole(argv[2], &size), *original = whole(argv[3], &original_size); + int rounds = atoi(argv[4]); + if (!stream || !original) { fprintf(stderr, "cannot read the stream or the original\n"); return 1; } + unsigned char *out = malloc(original_size + 1); + if (puff(stream, size, out, original_size) != (long)original_size || memcmp(out, original, original_size)) { + fprintf(stderr, "C's bytes differ\n"); + return 1; + } + if (zlib_inflate(stream, size, out, original_size + 1) != (long)original_size || memcmp(out, original, original_size)) { + fprintf(stderr, "zlib's bytes differ\n"); + return 1; + } + + PureModule module = {0}; + char error[160]; + if (pure_module_read(argv[1], &module, error, sizeof(error)) != PURE_OK) { fprintf(stderr, "%s\n", error); return 1; } + PureApp app; + if (pure_app_start(&app, &module, (PureData){0}) != PURE_OK) return 1; + int input = pure_module_input(&module, "inflate"), state = pure_module_state(&module, "inflated"); + uint64_t before = app.work; + if (pure_app_message(&app, (uint32_t)input, (PureSpan){stream, size}) != PURE_OK) { + fprintf(stderr, "event failed\n"); + return 1; + } + uint64_t work = app.work - before; + const PureList *list = app.lists[state]; + PureCell cell = pure_list_count(list) == 1 ? pure_list_get(list, 0, 0) : (PureCell){0}; + if (cell.size != original_size || memcmp(cell.text, original, original_size)) { fprintf(stderr, "Pure's bytes differ\n"); return 1; } + + double c_runs[7], zlib_runs[7], pure_runs[7]; + volatile unsigned char sink = 0; + for (int r = 0; r < 7; ++r) { + double start = now_ns(); + for (int i = 0; i < rounds; ++i) { puff(stream, size, out, original_size); sink ^= out[i % (original_size + 1)]; } + c_runs[r] = (now_ns() - start) / rounds; + start = now_ns(); + for (int i = 0; i < rounds; ++i) { zlib_inflate(stream, size, out, original_size + 1); sink ^= out[i % (original_size + 1)]; } + zlib_runs[r] = (now_ns() - start) / rounds; + start = now_ns(); + for (int i = 0; i < rounds; ++i) + if (pure_app_message(&app, (uint32_t)input, (PureSpan){stream, size}) != PURE_OK) return 1; + pure_runs[r] = (now_ns() - start) / rounds; + } + qsort(c_runs, 7, sizeof(double), by_value); + qsort(zlib_runs, 7, sizeof(double), by_value); + qsort(pure_runs, 7, sizeof(double), by_value); + double n = (double)original_size; + const char *jit = getenv("PURE_JIT"), *batch = getenv("PURE_BATCH"); + printf("{\"size\":%zu,\"stream\":%zu,\"jit\":\"%s\",\"batch\":\"%s\",\"c_ns_per_byte\":%.3f,\"zlib_ns_per_byte\":%.3f," + "\"pure_ns_per_byte\":%.3f,\"to_c\":%.1f,\"to_zlib\":%.1f,\"instructions\":%llu,\"instructions_per_byte\":%.1f}\n", + original_size, size, jit ? jit : "", batch ? batch : "", c_runs[3] / n, zlib_runs[3] / n, pure_runs[3] / n, + pure_runs[3] / c_runs[3], pure_runs[3] / zlib_runs[3], (unsigned long long)work, (double)work / n); + pure_app_stop(&app); + return 0; +} -- cgit v1.2.3