server-skynet-source-3rd-je.../src/sz.c
Qi Wang a11be50332 Implement opt.cache_oblivious.
Keep config.cache_oblivious for now to remain backward-compatible.
2021-02-11 11:32:01 -08:00

115 lines
3.1 KiB
C

#include "jemalloc/internal/jemalloc_preamble.h"
#include "jemalloc/internal/jemalloc_internal_includes.h"
#include "jemalloc/internal/sz.h"
JEMALLOC_ALIGNED(CACHELINE)
size_t sz_pind2sz_tab[SC_NPSIZES+1];
size_t sz_large_pad;
size_t
sz_psz_quantize_floor(size_t size) {
size_t ret;
pszind_t pind;
assert(size > 0);
assert((size & PAGE_MASK) == 0);
pind = sz_psz2ind(size - sz_large_pad + 1);
if (pind == 0) {
/*
* Avoid underflow. This short-circuit would also do the right
* thing for all sizes in the range for which there are
* PAGE-spaced size classes, but it's simplest to just handle
* the one case that would cause erroneous results.
*/
return size;
}
ret = sz_pind2sz(pind - 1) + sz_large_pad;
assert(ret <= size);
return ret;
}
size_t
sz_psz_quantize_ceil(size_t size) {
size_t ret;
assert(size > 0);
assert(size - sz_large_pad <= SC_LARGE_MAXCLASS);
assert((size & PAGE_MASK) == 0);
ret = sz_psz_quantize_floor(size);
if (ret < size) {
/*
* Skip a quantization that may have an adequately large extent,
* because under-sized extents may be mixed in. This only
* happens when an unusual size is requested, i.e. for aligned
* allocation, and is just one of several places where linear
* search would potentially find sufficiently aligned available
* memory somewhere lower.
*/
ret = sz_pind2sz(sz_psz2ind(ret - sz_large_pad + 1)) +
sz_large_pad;
}
return ret;
}
static void
sz_boot_pind2sz_tab(const sc_data_t *sc_data) {
int pind = 0;
for (unsigned i = 0; i < SC_NSIZES; i++) {
const sc_t *sc = &sc_data->sc[i];
if (sc->psz) {
sz_pind2sz_tab[pind] = (ZU(1) << sc->lg_base)
+ (ZU(sc->ndelta) << sc->lg_delta);
pind++;
}
}
for (int i = pind; i <= (int)SC_NPSIZES; i++) {
sz_pind2sz_tab[pind] = sc_data->large_maxclass + PAGE;
}
}
JEMALLOC_ALIGNED(CACHELINE)
size_t sz_index2size_tab[SC_NSIZES];
static void
sz_boot_index2size_tab(const sc_data_t *sc_data) {
for (unsigned i = 0; i < SC_NSIZES; i++) {
const sc_t *sc = &sc_data->sc[i];
sz_index2size_tab[i] = (ZU(1) << sc->lg_base)
+ (ZU(sc->ndelta) << (sc->lg_delta));
}
}
/*
* To keep this table small, we divide sizes by the tiny min size, which gives
* the smallest interval for which the result can change.
*/
JEMALLOC_ALIGNED(CACHELINE)
uint8_t sz_size2index_tab[(SC_LOOKUP_MAXCLASS >> SC_LG_TINY_MIN) + 1];
static void
sz_boot_size2index_tab(const sc_data_t *sc_data) {
size_t dst_max = (SC_LOOKUP_MAXCLASS >> SC_LG_TINY_MIN) + 1;
size_t dst_ind = 0;
for (unsigned sc_ind = 0; sc_ind < SC_NSIZES && dst_ind < dst_max;
sc_ind++) {
const sc_t *sc = &sc_data->sc[sc_ind];
size_t sz = (ZU(1) << sc->lg_base)
+ (ZU(sc->ndelta) << sc->lg_delta);
size_t max_ind = ((sz + (ZU(1) << SC_LG_TINY_MIN) - 1)
>> SC_LG_TINY_MIN);
for (; dst_ind <= max_ind && dst_ind < dst_max; dst_ind++) {
sz_size2index_tab[dst_ind] = sc_ind;
}
}
}
void
sz_boot(const sc_data_t *sc_data, bool cache_oblivious) {
sz_large_pad = cache_oblivious ? PAGE : 0;
sz_boot_pind2sz_tab(sc_data);
sz_boot_index2size_tab(sc_data);
sz_boot_size2index_tab(sc_data);
}