qemu-e2k/contrib/plugins/cache.c
Mahmoud Mandour c79a2116af plugins/cache: supported multicore cache modelling
Multicore L1 cache modelling is introduced and is supported for both
full system emulation and linux-user.

For full-system emulation, L1 icache and dcache are maintained for each
available core, since this information is exposed to the plugin through
`qemu_plugin_n_vcpus()`.

For linux-user, a static number of cores is assumed (default 1 core, and
can be provided as a plugin argument `cores=N`). Every memory access
goes through one of these caches, this approach is taken as it's
somewhat akin to what happens on real setup, where a program that
dispatches more threads than the available cores, they'll thrash
each other

Signed-off-by: Mahmoud Mandour <ma.mandourr@gmail.com>
Message-Id: <20210803151301.123581-2-ma.mandourr@gmail.com>
Signed-off-by: Alex Bennée <alex.bennee@linaro.org>
2021-09-02 11:29:34 +01:00

729 lines
19 KiB
C

/*
* Copyright (C) 2021, Mahmoud Mandour <ma.mandourr@gmail.com>
*
* License: GNU GPL, version 2 or later.
* See the COPYING file in the top-level directory.
*/
#include <inttypes.h>
#include <stdio.h>
#include <glib.h>
#include <qemu-plugin.h>
QEMU_PLUGIN_EXPORT int qemu_plugin_version = QEMU_PLUGIN_VERSION;
static enum qemu_plugin_mem_rw rw = QEMU_PLUGIN_MEM_RW;
static GHashTable *miss_ht;
static GMutex hashtable_lock;
static GRand *rng;
static int limit;
static bool sys;
enum EvictionPolicy {
LRU,
FIFO,
RAND,
};
enum EvictionPolicy policy;
/*
* A CacheSet is a set of cache blocks. A memory block that maps to a set can be
* put in any of the blocks inside the set. The number of block per set is
* called the associativity (assoc).
*
* Each block contains the the stored tag and a valid bit. Since this is not
* a functional simulator, the data itself is not stored. We only identify
* whether a block is in the cache or not by searching for its tag.
*
* In order to search for memory data in the cache, the set identifier and tag
* are extracted from the address and the set is probed to see whether a tag
* match occur.
*
* An address is logically divided into three portions: The block offset,
* the set number, and the tag.
*
* The set number is used to identify the set in which the block may exist.
* The tag is compared against all the tags of a set to search for a match. If a
* match is found, then the access is a hit.
*
* The CacheSet also contains bookkeaping information about eviction details.
*/
typedef struct {
uint64_t tag;
bool valid;
} CacheBlock;
typedef struct {
CacheBlock *blocks;
uint64_t *lru_priorities;
uint64_t lru_gen_counter;
GQueue *fifo_queue;
} CacheSet;
typedef struct {
CacheSet *sets;
int num_sets;
int cachesize;
int assoc;
int blksize_shift;
uint64_t set_mask;
uint64_t tag_mask;
uint64_t accesses;
uint64_t misses;
} Cache;
typedef struct {
char *disas_str;
const char *symbol;
uint64_t addr;
uint64_t dmisses;
uint64_t imisses;
} InsnData;
void (*update_hit)(Cache *cache, int set, int blk);
void (*update_miss)(Cache *cache, int set, int blk);
void (*metadata_init)(Cache *cache);
void (*metadata_destroy)(Cache *cache);
static int cores;
static Cache **dcaches, **icaches;
static GMutex *dcache_locks;
static GMutex *icache_locks;
static uint64_t all_dmem_accesses;
static uint64_t all_imem_accesses;
static uint64_t all_imisses;
static uint64_t all_dmisses;
static int pow_of_two(int num)
{
g_assert((num & (num - 1)) == 0);
int ret = 0;
while (num /= 2) {
ret++;
}
return ret;
}
/*
* LRU evection policy: For each set, a generation counter is maintained
* alongside a priority array.
*
* On each set access, the generation counter is incremented.
*
* On a cache hit: The hit-block is assigned the current generation counter,
* indicating that it is the most recently used block.
*
* On a cache miss: The block with the least priority is searched and replaced
* with the newly-cached block, of which the priority is set to the current
* generation number.
*/
static void lru_priorities_init(Cache *cache)
{
int i;
for (i = 0; i < cache->num_sets; i++) {
cache->sets[i].lru_priorities = g_new0(uint64_t, cache->assoc);
cache->sets[i].lru_gen_counter = 0;
}
}
static void lru_update_blk(Cache *cache, int set_idx, int blk_idx)
{
CacheSet *set = &cache->sets[set_idx];
set->lru_priorities[blk_idx] = cache->sets[set_idx].lru_gen_counter;
set->lru_gen_counter++;
}
static int lru_get_lru_block(Cache *cache, int set_idx)
{
int i, min_idx, min_priority;
min_priority = cache->sets[set_idx].lru_priorities[0];
min_idx = 0;
for (i = 1; i < cache->assoc; i++) {
if (cache->sets[set_idx].lru_priorities[i] < min_priority) {
min_priority = cache->sets[set_idx].lru_priorities[i];
min_idx = i;
}
}
return min_idx;
}
static void lru_priorities_destroy(Cache *cache)
{
int i;
for (i = 0; i < cache->num_sets; i++) {
g_free(cache->sets[i].lru_priorities);
}
}
/*
* FIFO eviction policy: a FIFO queue is maintained for each CacheSet that
* stores accesses to the cache.
*
* On a compulsory miss: The block index is enqueued to the fifo_queue to
* indicate that it's the latest cached block.
*
* On a conflict miss: The first-in block is removed from the cache and the new
* block is put in its place and enqueued to the FIFO queue.
*/
static void fifo_init(Cache *cache)
{
int i;
for (i = 0; i < cache->num_sets; i++) {
cache->sets[i].fifo_queue = g_queue_new();
}
}
static int fifo_get_first_block(Cache *cache, int set)
{
GQueue *q = cache->sets[set].fifo_queue;
return GPOINTER_TO_INT(g_queue_pop_tail(q));
}
static void fifo_update_on_miss(Cache *cache, int set, int blk_idx)
{
GQueue *q = cache->sets[set].fifo_queue;
g_queue_push_head(q, GINT_TO_POINTER(blk_idx));
}
static void fifo_destroy(Cache *cache)
{
int i;
for (i = 0; i < cache->num_sets; i++) {
g_queue_free(cache->sets[i].fifo_queue);
}
}
static inline uint64_t extract_tag(Cache *cache, uint64_t addr)
{
return addr & cache->tag_mask;
}
static inline uint64_t extract_set(Cache *cache, uint64_t addr)
{
return (addr & cache->set_mask) >> cache->blksize_shift;
}
static const char *cache_config_error(int blksize, int assoc, int cachesize)
{
if (cachesize % blksize != 0) {
return "cache size must be divisible by block size";
} else if (cachesize % (blksize * assoc) != 0) {
return "cache size must be divisible by set size (assoc * block size)";
} else {
return NULL;
}
}
static bool bad_cache_params(int blksize, int assoc, int cachesize)
{
return (cachesize % blksize) != 0 || (cachesize % (blksize * assoc) != 0);
}
static Cache *cache_init(int blksize, int assoc, int cachesize)
{
Cache *cache;
int i;
uint64_t blk_mask;
/*
* This function shall not be called directly, and hence expects suitable
* parameters.
*/
g_assert(!bad_cache_params(blksize, assoc, cachesize));
cache = g_new(Cache, 1);
cache->assoc = assoc;
cache->cachesize = cachesize;
cache->num_sets = cachesize / (blksize * assoc);
cache->sets = g_new(CacheSet, cache->num_sets);
cache->blksize_shift = pow_of_two(blksize);
cache->accesses = 0;
cache->misses = 0;
for (i = 0; i < cache->num_sets; i++) {
cache->sets[i].blocks = g_new0(CacheBlock, assoc);
}
blk_mask = blksize - 1;
cache->set_mask = ((cache->num_sets - 1) << cache->blksize_shift);
cache->tag_mask = ~(cache->set_mask | blk_mask);
if (metadata_init) {
metadata_init(cache);
}
return cache;
}
static Cache **caches_init(int blksize, int assoc, int cachesize)
{
Cache **caches;
int i;
if (bad_cache_params(blksize, assoc, cachesize)) {
return NULL;
}
caches = g_new(Cache *, cores);
for (i = 0; i < cores; i++) {
caches[i] = cache_init(blksize, assoc, cachesize);
}
return caches;
}
static int get_invalid_block(Cache *cache, uint64_t set)
{
int i;
for (i = 0; i < cache->assoc; i++) {
if (!cache->sets[set].blocks[i].valid) {
return i;
}
}
return -1;
}
static int get_replaced_block(Cache *cache, int set)
{
switch (policy) {
case RAND:
return g_rand_int_range(rng, 0, cache->assoc);
case LRU:
return lru_get_lru_block(cache, set);
case FIFO:
return fifo_get_first_block(cache, set);
default:
g_assert_not_reached();
}
}
static int in_cache(Cache *cache, uint64_t addr)
{
int i;
uint64_t tag, set;
tag = extract_tag(cache, addr);
set = extract_set(cache, addr);
for (i = 0; i < cache->assoc; i++) {
if (cache->sets[set].blocks[i].tag == tag &&
cache->sets[set].blocks[i].valid) {
return i;
}
}
return -1;
}
/**
* access_cache(): Simulate a cache access
* @cache: The cache under simulation
* @addr: The address of the requested memory location
*
* Returns true if the requsted data is hit in the cache and false when missed.
* The cache is updated on miss for the next access.
*/
static bool access_cache(Cache *cache, uint64_t addr)
{
int hit_blk, replaced_blk;
uint64_t tag, set;
tag = extract_tag(cache, addr);
set = extract_set(cache, addr);
hit_blk = in_cache(cache, addr);
if (hit_blk != -1) {
if (update_hit) {
update_hit(cache, set, hit_blk);
}
return true;
}
replaced_blk = get_invalid_block(cache, set);
if (replaced_blk == -1) {
replaced_blk = get_replaced_block(cache, set);
}
if (update_miss) {
update_miss(cache, set, replaced_blk);
}
cache->sets[set].blocks[replaced_blk].tag = tag;
cache->sets[set].blocks[replaced_blk].valid = true;
return false;
}
static void vcpu_mem_access(unsigned int vcpu_index, qemu_plugin_meminfo_t info,
uint64_t vaddr, void *userdata)
{
uint64_t effective_addr;
struct qemu_plugin_hwaddr *hwaddr;
int cache_idx;
InsnData *insn;
hwaddr = qemu_plugin_get_hwaddr(info, vaddr);
if (hwaddr && qemu_plugin_hwaddr_is_io(hwaddr)) {
return;
}
effective_addr = hwaddr ? qemu_plugin_hwaddr_phys_addr(hwaddr) : vaddr;
cache_idx = vcpu_index % cores;
g_mutex_lock(&dcache_locks[cache_idx]);
if (!access_cache(dcaches[cache_idx], effective_addr)) {
insn = (InsnData *) userdata;
__atomic_fetch_add(&insn->dmisses, 1, __ATOMIC_SEQ_CST);
dcaches[cache_idx]->misses++;
}
dcaches[cache_idx]->accesses++;
g_mutex_unlock(&dcache_locks[cache_idx]);
}
static void vcpu_insn_exec(unsigned int vcpu_index, void *userdata)
{
uint64_t insn_addr;
InsnData *insn;
int cache_idx;
insn_addr = ((InsnData *) userdata)->addr;
cache_idx = vcpu_index % cores;
g_mutex_lock(&icache_locks[cache_idx]);
if (!access_cache(icaches[cache_idx], insn_addr)) {
insn = (InsnData *) userdata;
__atomic_fetch_add(&insn->imisses, 1, __ATOMIC_SEQ_CST);
icaches[cache_idx]->misses++;
}
icaches[cache_idx]->accesses++;
g_mutex_unlock(&icache_locks[cache_idx]);
}
static void vcpu_tb_trans(qemu_plugin_id_t id, struct qemu_plugin_tb *tb)
{
size_t n_insns;
size_t i;
InsnData *data;
n_insns = qemu_plugin_tb_n_insns(tb);
for (i = 0; i < n_insns; i++) {
struct qemu_plugin_insn *insn = qemu_plugin_tb_get_insn(tb, i);
uint64_t effective_addr;
if (sys) {
effective_addr = (uint64_t) qemu_plugin_insn_haddr(insn);
} else {
effective_addr = (uint64_t) qemu_plugin_insn_vaddr(insn);
}
/*
* Instructions might get translated multiple times, we do not create
* new entries for those instructions. Instead, we fetch the same
* entry from the hash table and register it for the callback again.
*/
g_mutex_lock(&hashtable_lock);
data = g_hash_table_lookup(miss_ht, GUINT_TO_POINTER(effective_addr));
if (data == NULL) {
data = g_new0(InsnData, 1);
data->disas_str = qemu_plugin_insn_disas(insn);
data->symbol = qemu_plugin_insn_symbol(insn);
data->addr = effective_addr;
g_hash_table_insert(miss_ht, GUINT_TO_POINTER(effective_addr),
(gpointer) data);
}
g_mutex_unlock(&hashtable_lock);
qemu_plugin_register_vcpu_mem_cb(insn, vcpu_mem_access,
QEMU_PLUGIN_CB_NO_REGS,
rw, data);
qemu_plugin_register_vcpu_insn_exec_cb(insn, vcpu_insn_exec,
QEMU_PLUGIN_CB_NO_REGS, data);
}
}
static void insn_free(gpointer data)
{
InsnData *insn = (InsnData *) data;
g_free(insn->disas_str);
g_free(insn);
}
static void cache_free(Cache *cache)
{
for (int i = 0; i < cache->num_sets; i++) {
g_free(cache->sets[i].blocks);
}
if (metadata_destroy) {
metadata_destroy(cache);
}
g_free(cache->sets);
g_free(cache);
}
static void caches_free(Cache **caches)
{
int i;
for (i = 0; i < cores; i++) {
cache_free(caches[i]);
}
}
static int dcmp(gconstpointer a, gconstpointer b)
{
InsnData *insn_a = (InsnData *) a;
InsnData *insn_b = (InsnData *) b;
return insn_a->dmisses < insn_b->dmisses ? 1 : -1;
}
static void append_stats_line(GString *line, uint64_t daccess, uint64_t dmisses,
uint64_t iaccess, uint64_t imisses)
{
double dmiss_rate, imiss_rate;
dmiss_rate = ((double) dmisses) / (daccess) * 100.0;
imiss_rate = ((double) imisses) / (iaccess) * 100.0;
g_string_append_printf(line, "%-14lu %-12lu %9.4lf%% %-14lu %-12lu"
" %9.4lf%%\n",
daccess,
dmisses,
daccess ? dmiss_rate : 0.0,
iaccess,
imisses,
iaccess ? imiss_rate : 0.0);
}
static void sum_stats(void)
{
int i;
g_assert(cores > 1);
for (i = 0; i < cores; i++) {
all_imisses += icaches[i]->misses;
all_dmisses += dcaches[i]->misses;
all_imem_accesses += icaches[i]->accesses;
all_dmem_accesses += dcaches[i]->accesses;
}
}
static int icmp(gconstpointer a, gconstpointer b)
{
InsnData *insn_a = (InsnData *) a;
InsnData *insn_b = (InsnData *) b;
return insn_a->imisses < insn_b->imisses ? 1 : -1;
}
static void log_stats(void)
{
int i;
Cache *icache, *dcache;
g_autoptr(GString) rep = g_string_new("core #, data accesses, data misses,"
" dmiss rate, insn accesses,"
" insn misses, imiss rate\n");
for (i = 0; i < cores; i++) {
g_string_append_printf(rep, "%-8d", i);
dcache = dcaches[i];
icache = icaches[i];
append_stats_line(rep, dcache->accesses, dcache->misses,
icache->accesses, icache->misses);
}
if (cores > 1) {
sum_stats();
g_string_append_printf(rep, "%-8s", "sum");
append_stats_line(rep, all_dmem_accesses, all_dmisses,
all_imem_accesses, all_imisses);
}
g_string_append(rep, "\n");
qemu_plugin_outs(rep->str);
}
static void log_top_insns(void)
{
int i;
GList *curr, *miss_insns;
InsnData *insn;
miss_insns = g_hash_table_get_values(miss_ht);
miss_insns = g_list_sort(miss_insns, dcmp);
g_autoptr(GString) rep = g_string_new("");
g_string_append_printf(rep, "%s", "address, data misses, instruction\n");
for (curr = miss_insns, i = 0; curr && i < limit; i++, curr = curr->next) {
insn = (InsnData *) curr->data;
g_string_append_printf(rep, "0x%" PRIx64, insn->addr);
if (insn->symbol) {
g_string_append_printf(rep, " (%s)", insn->symbol);
}
g_string_append_printf(rep, ", %ld, %s\n", insn->dmisses,
insn->disas_str);
}
miss_insns = g_list_sort(miss_insns, icmp);
g_string_append_printf(rep, "%s", "\naddress, fetch misses, instruction\n");
for (curr = miss_insns, i = 0; curr && i < limit; i++, curr = curr->next) {
insn = (InsnData *) curr->data;
g_string_append_printf(rep, "0x%" PRIx64, insn->addr);
if (insn->symbol) {
g_string_append_printf(rep, " (%s)", insn->symbol);
}
g_string_append_printf(rep, ", %ld, %s\n", insn->imisses,
insn->disas_str);
}
qemu_plugin_outs(rep->str);
g_list_free(miss_insns);
}
static void plugin_exit(qemu_plugin_id_t id, void *p)
{
log_stats();
log_top_insns();
caches_free(dcaches);
caches_free(icaches);
g_hash_table_destroy(miss_ht);
}
static void policy_init(void)
{
switch (policy) {
case LRU:
update_hit = lru_update_blk;
update_miss = lru_update_blk;
metadata_init = lru_priorities_init;
metadata_destroy = lru_priorities_destroy;
break;
case FIFO:
update_miss = fifo_update_on_miss;
metadata_init = fifo_init;
metadata_destroy = fifo_destroy;
break;
case RAND:
rng = g_rand_new();
break;
default:
g_assert_not_reached();
}
}
QEMU_PLUGIN_EXPORT
int qemu_plugin_install(qemu_plugin_id_t id, const qemu_info_t *info,
int argc, char **argv)
{
int i;
int iassoc, iblksize, icachesize;
int dassoc, dblksize, dcachesize;
limit = 32;
sys = info->system_emulation;
dassoc = 8;
dblksize = 64;
dcachesize = dblksize * dassoc * 32;
iassoc = 8;
iblksize = 64;
icachesize = iblksize * iassoc * 32;
policy = LRU;
cores = sys ? qemu_plugin_n_vcpus() : 1;
for (i = 0; i < argc; i++) {
char *opt = argv[i];
if (g_str_has_prefix(opt, "iblksize=")) {
iblksize = g_ascii_strtoll(opt + 9, NULL, 10);
} else if (g_str_has_prefix(opt, "iassoc=")) {
iassoc = g_ascii_strtoll(opt + 7, NULL, 10);
} else if (g_str_has_prefix(opt, "icachesize=")) {
icachesize = g_ascii_strtoll(opt + 11, NULL, 10);
} else if (g_str_has_prefix(opt, "dblksize=")) {
dblksize = g_ascii_strtoll(opt + 9, NULL, 10);
} else if (g_str_has_prefix(opt, "dassoc=")) {
dassoc = g_ascii_strtoll(opt + 7, NULL, 10);
} else if (g_str_has_prefix(opt, "dcachesize=")) {
dcachesize = g_ascii_strtoll(opt + 11, NULL, 10);
} else if (g_str_has_prefix(opt, "limit=")) {
limit = g_ascii_strtoll(opt + 6, NULL, 10);
} else if (g_str_has_prefix(opt, "cores=")) {
cores = g_ascii_strtoll(opt + 6, NULL, 10);
} else if (g_str_has_prefix(opt, "evict=")) {
gchar *p = opt + 6;
if (g_strcmp0(p, "rand") == 0) {
policy = RAND;
} else if (g_strcmp0(p, "lru") == 0) {
policy = LRU;
} else if (g_strcmp0(p, "fifo") == 0) {
policy = FIFO;
} else {
fprintf(stderr, "invalid eviction policy: %s\n", opt);
return -1;
}
} else {
fprintf(stderr, "option parsing failed: %s\n", opt);
return -1;
}
}
policy_init();
dcaches = caches_init(dblksize, dassoc, dcachesize);
if (!dcaches) {
const char *err = cache_config_error(dblksize, dassoc, dcachesize);
fprintf(stderr, "dcache cannot be constructed from given parameters\n");
fprintf(stderr, "%s\n", err);
return -1;
}
icaches = caches_init(iblksize, iassoc, icachesize);
if (!icaches) {
const char *err = cache_config_error(iblksize, iassoc, icachesize);
fprintf(stderr, "icache cannot be constructed from given parameters\n");
fprintf(stderr, "%s\n", err);
return -1;
}
dcache_locks = g_new0(GMutex, cores);
icache_locks = g_new0(GMutex, cores);
qemu_plugin_register_vcpu_tb_trans_cb(id, vcpu_tb_trans);
qemu_plugin_register_atexit_cb(id, plugin_exit, NULL);
miss_ht = g_hash_table_new_full(NULL, g_direct_equal, NULL, insn_free);
return 0;
}