awk: assorted optimizations

hash_find(): do not caclculate hash twice. Do not divide - can use
cheap multiply-by-8 shift.

nextword(): do not repeatedly increment in-memory value, do it in register,
then store final result.

hashwalk_init(): do not strlen() twice.

function                                             old     new   delta
hash_search3                                           -      49     +49
hash_find                                            259     281     +22
nextword                                              19      16      -3
evaluate                                            3141    3137      -4
hash_search                                           54      28     -26
------------------------------------------------------------------------------
(add/remove: 1/0 grow/shrink: 1/3 up/down: 71/-33)             Total: 38 bytes

Signed-off-by: Denys Vlasenko <vda.linux@googlemail.com>
This commit is contained in:
Denys Vlasenko 2021-06-29 19:07:36 +02:00
parent b3c91a127f
commit 3aff3b9cb8

View File

@ -696,6 +696,7 @@ static void hash_clear(xhash *hash)
while (hi) { while (hi) {
thi = hi; thi = hi;
hi = hi->next; hi = hi->next;
//FIXME: this assumes that it's a hash of *variables*:
free(thi->data.v.string); free(thi->data.v.string);
free(thi); free(thi);
} }
@ -714,11 +715,11 @@ static void hash_free(xhash *hash)
#endif #endif
/* find item in hash, return ptr to data, NULL if not found */ /* find item in hash, return ptr to data, NULL if not found */
static void *hash_search(xhash *hash, const char *name) static NOINLINE void *hash_search3(xhash *hash, const char *name, unsigned idx)
{ {
hash_item *hi; hash_item *hi;
hi = hash->items[hashidx(name) % hash->csize]; hi = hash->items[idx % hash->csize];
while (hi) { while (hi) {
if (strcmp(hi->name, name) == 0) if (strcmp(hi->name, name) == 0)
return &hi->data; return &hi->data;
@ -727,6 +728,11 @@ static void *hash_search(xhash *hash, const char *name)
return NULL; return NULL;
} }
static void *hash_search(xhash *hash, const char *name)
{
return hash_search3(hash, name, hashidx(name));
}
/* grow hash if it becomes too big */ /* grow hash if it becomes too big */
static void hash_rebuild(xhash *hash) static void hash_rebuild(xhash *hash)
{ {
@ -762,16 +768,17 @@ static void *hash_find(xhash *hash, const char *name)
unsigned idx; unsigned idx;
int l; int l;
hi = hash_search(hash, name); idx = hashidx(name);
hi = hash_search3(hash, name, idx);
if (!hi) { if (!hi) {
if (++hash->nel / hash->csize > 10) if (++hash->nel > hash->csize * 8)
hash_rebuild(hash); hash_rebuild(hash);
l = strlen(name) + 1; l = strlen(name) + 1;
hi = xzalloc(sizeof(*hi) + l); hi = xzalloc(sizeof(*hi) + l);
strcpy(hi->name, name); strcpy(hi->name, name);
idx = hashidx(name) % hash->csize; idx = idx % hash->csize;
hi->next = hash->items[idx]; hi->next = hash->items[idx];
hash->items[idx] = hi; hash->items[idx] = hi;
hash->glen += l; hash->glen += l;
@ -822,8 +829,10 @@ static char *skip_spaces(char *p)
static char *nextword(char **s) static char *nextword(char **s)
{ {
char *p = *s; char *p = *s;
while (*(*s)++ != '\0') char *q = p;
while (*q++ != '\0')
continue; continue;
*s = q;
return p; return p;
} }
@ -2116,8 +2125,7 @@ static void hashwalk_init(var *v, xhash *array)
for (i = 0; i < array->csize; i++) { for (i = 0; i < array->csize; i++) {
hi = array->items[i]; hi = array->items[i];
while (hi) { while (hi) {
strcpy(w->end, hi->name); w->end = stpcpy(w->end, hi->name) + 1;
nextword(&w->end);
hi = hi->next; hi = hi->next;
} }
} }
@ -3504,7 +3512,7 @@ int awk_main(int argc UNUSED_PARAM, char **argv)
setari_u(intvar[ARGV], ++i, *argv++); setari_u(intvar[ARGV], ++i, *argv++);
setvar_i(intvar[ARGC], i + 1); setvar_i(intvar[ARGC], i + 1);
//fdhash = ahash - done via define //fdhash = ahash; // done via define
newfile("/dev/stdin")->F = stdin; newfile("/dev/stdin")->F = stdin;
newfile("/dev/stdout")->F = stdout; newfile("/dev/stdout")->F = stdout;
newfile("/dev/stderr")->F = stderr; newfile("/dev/stderr")->F = stderr;