Research 140, 027's stage 2c measured: JSON read into a tree by recursive descent (aup's examples/json_tree.aup: value(d, i) -> Json, number giving the value and where it ends, each call's own let's lists, string's escapes joined by char(u), a depth-first walk passing items as values) against Python's reading walked so: JSONTestSuite's 95 y_ taken alike, no n_ taken, 2b's 6 i_ taken (a lone surrogate refused by char as encoding); 100,000 and 50,000 deep refused as memory (the frames fill the event's room, 0.2 s), 30,000 deep taken (past Python's json); 300 random documents alike; in make check (module_specs' digest now depth first, three paths); 142 lines against the same design's 105 in Python (json_rec.py; the bar missed: 24 lines of declarations, 23 of braces), 46 times cJSON's time and 1.6 times plain Python's (600 ns a byte, 248 Pure instructions; 71% in the steps, about 10% in calls, 66,000 calls a document, most of them small block fns like skip); characters against Python's ord in aup's check; probes: p5-chars compiles, p4-rec (a view calling itself) refused; instruction counts: the workloads' modules the same, batches within 0.001% of 2b (BFS -0.21%), the executor alone -0.001% to +0.30%, the bank's queries -0.30% to +0.11% (four runs each; bar 1%); json_rec_time.c (cJSON, a depth-first digest compared first); CHECKPOINT's direction row through 2c

1 files changed, 119 insertions, 0 deletions

diff --git a/research/experiments/044-specs/json_rec_time.c b/research/experiments/044-specs/json_rec_time.c
new file mode 100644
index 0000000..03c02c4
--- /dev/null
+++ b/research/experiments/044-specs/json_rec_time.c
@@ -0,0 +1,119 @@
+/* Research 140, design 027 stage 2c: aup's examples/json_tree.aup (run by
+ * Pure: a JSON text read into a tree by recursive descent, then a digest of
+ * the tree, depth first) against C:
+ * cJSON (the system's libcjson.so.1, its few declarations written here)
+ * reading the same document into its tree, and the same digest. Both digests
+ * must be equal; then each is timed (the median of seven runs of rounds).
+ *
+ *   cc -O2 -std=c17 -I../../../core json_rec_time.c ../../../build/libpure.a -l:libcjson.so.1 -lm -pthread -o json_rec_time
+ *   ./json_rec_time json_tree.pmod ROUNDS
+ *
+ * The document is json_time.c's: 2,000 objects of a key and an array of a
+ * number, true, null and a string with an escape. One JSON line. */
+#define _POSIX_C_SOURCE 200809L
+#include "module.h"
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+#include <time.h>
+
+typedef struct cJSON {
+    struct cJSON *next, *prev, *child;
+    int type;
+    char *valuestring;
+    int valueint;
+    double valuedouble;
+    char *string;
+} cJSON;
+cJSON *cJSON_Parse(const char *value);
+void cJSON_Delete(cJSON *item);
+enum { FALSE_ = 1, TRUE_ = 2, NULL_ = 4, NUMBER = 8, STRING = 16, ARRAY = 32, OBJECT = 64 };
+
+/* nulls, trues, falses, numbers, strings, arrays, objects, sum, text bytes, key bytes: depth first, as json_tree.aup. */
+typedef struct { double count[7], sum, text_bytes, key_bytes; } Digest;
+static void walk(const cJSON *x, Digest *d) {
+    switch (x->type & 0xFF) {
+    case NULL_: d->count[0]++; break;
+    case TRUE_: d->count[1]++; break;
+    case FALSE_: d->count[2]++; break;
+    case NUMBER: d->count[3]++, d->sum += x->valuedouble; break;
+    case STRING: d->count[4]++, d->text_bytes += (double)strlen(x->valuestring); break;
+    case ARRAY: d->count[5]++; for (const cJSON *y = x->child; y; y = y->next) walk(y, d); break;
+    default: d->count[6]++; for (const cJSON *y = x->child; y; y = y->next) d->key_bytes += (double)strlen(y->string), walk(y, d);
+    }
+}
+static Digest digest(const char *text, cJSON **queue) {
+    (void)queue;
+    Digest d = {{0}, 0, 0, 0};
+    cJSON *root = cJSON_Parse(text);
+    walk(root, &d);
+    cJSON_Delete(root);
+    return d;
+}
+
+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);
+}
+
+int main(int argc, char **argv) {
+    if (argc != 3) { fprintf(stderr, "usage: %s MODULE ROUNDS\n", argv[0]); return 2; }
+    int rounds = atoi(argv[2]);
+    static char doc[1 << 18];
+    static cJSON *queue[1 << 16];
+    size_t n = 0;
+    doc[n++] = '[';
+    for (int k = 0; k < 2000; ++k)
+        n += (size_t)sprintf(doc + n, "%s{\"key%d\": [1.5e3, true, null, \"text \\u00e9 %d\"]}", k ? "," : "", k, k);
+    doc[n++] = ']';
+    doc[n] = 0;
+    PureModule module;
+    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 check = pure_module_input(&module, "check");
+    if (pure_app_message(&app, (uint32_t)check, NULL, -1, (PureSpan){(const unsigned char *)doc, n}) != PURE_OK) {
+        fprintf(stderr, "event failed\n");
+        return 1;
+    }
+    uint64_t work = app.work;
+    Digest c = digest(doc, queue);
+    /* The states after valid: nulls, trues, falses, numbers, strings, arrays, objects, sum, text_bytes, key_bytes. */
+    const char *names[] = {"nulls", "trues", "falses", "numbers", "strings", "arrays", "objects", "sum", "text_bytes", "key_bytes"};
+    double want[] = {c.count[0], c.count[1], c.count[2], c.count[3], c.count[4], c.count[5], c.count[6], c.sum, c.text_bytes, c.key_bytes};
+    for (unsigned k = 0; k < 10; ++k) {
+        int state = -1;
+        for (uint32_t s = 0; s < module.states; ++s) if (!strcmp(module.state_names[s], names[k])) state = (int)s;
+        if (state < 0 || app.values[state].num != want[k]) {
+            fprintf(stderr, "%s: %g against %g\n", names[k], state < 0 ? -1 : app.values[state].num, want[k]);
+            return 1;
+        }
+    }
+    double c_runs[7], pure_runs[7];
+    volatile double sink = 0;
+    for (int r = 0; r < 7; ++r) {
+        double start = now_ns();
+        for (int i = 0; i < rounds; ++i) sink += digest(doc, queue).sum;
+        c_runs[r] = (now_ns() - start) / rounds;
+        start = now_ns();
+        for (int i = 0; i < rounds; ++i)
+            if (pure_app_message(&app, (uint32_t)check, NULL, -1, (PureSpan){(const unsigned char *)doc, n}) != PURE_OK) return 1;
+        pure_runs[r] = (now_ns() - start) / rounds;
+    }
+    qsort(c_runs, 7, sizeof(double), by_value);
+    qsort(pure_runs, 7, sizeof(double), by_value);
+    const char *jit = getenv("PURE_JIT"), *batch = getenv("PURE_BATCH");
+    printf("{\"size\":%zu,\"jit\":\"%s\",\"batch\":\"%s\",\"c_ns_per_byte\":%.3f,\"pure_ns_per_byte\":%.3f,\"ratio\":%.1f,"
+           "\"instructions\":%llu,\"instructions_per_byte\":%.1f}\n",
+           n, jit ? jit : "", batch ? batch : "", c_runs[3] / (double)n, pure_runs[3] / (double)n, pure_runs[3] / c_runs[3],
+           (unsigned long long)work, (double)work / (double)n);
+    pure_app_stop(&app);
+    return 0;
+}