Skip to content

[PAL/vm-common] Use per-CPU L1d,L1i,L2 caches and one shared L3 #32

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
dimakuv merged 1 commit into from
Jul 12, 2024
Merged

[PAL/vm-common] Use per-CPU L1d,L1i,L2 caches and one shared L3 #32

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
56 changes: 42 additions & 14 deletions pal/src/host/vm-common/pal_common_misc.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -36,15 +36,8 @@ int pal_common_get_topo_info(struct pal_topo_info* topo_info) {
*
* Note the `static` keyword -- all arrays are allocated in BSS.
*/
static struct pal_cache_info caches[4] = {
{ .type = CACHE_TYPE_DATA, .level = 1, .size = 32 * 1024,
.coherency_line_size = 64, .number_of_sets = 64, .physical_line_partition = 1 },
{ .type = CACHE_TYPE_INSTRUCTION, .level = 1, .size = 32 * 1024,
.coherency_line_size = 64, .number_of_sets = 64, .physical_line_partition = 1 },
{ .type = CACHE_TYPE_UNIFIED, .level = 2, .size = 256 * 1024,
.coherency_line_size = 64, .number_of_sets = 1024, .physical_line_partition = 1 },
{ .type = CACHE_TYPE_UNIFIED, .level = 3, .size = 12288 * 1024,
.coherency_line_size = 64, .number_of_sets = 12288, .physical_line_partition = 1 },
static struct pal_cache_info caches[MAX_NUM_CPUS * MAX_CACHES] = {
0 /* to be filled below */
};
static struct pal_cpu_thread_info threads[MAX_NUM_CPUS] = {
0 /* to be filled below */
Expand All @@ -60,13 +53,48 @@ int pal_common_get_topo_info(struct pal_topo_info* topo_info) {
};
static size_t distances[1] = { 10 };

size_t caches_cnt = 0;
/* add one shared L3 */
caches[caches_cnt].type = CACHE_TYPE_UNIFIED;
caches[caches_cnt].level = 3;
caches[caches_cnt].size = 12288 * 1024;
caches[caches_cnt].coherency_line_size = 64;
caches[caches_cnt].number_of_sets = 12288;
caches[caches_cnt].physical_line_partition = 1;
caches_cnt++;

for (size_t i = 0; i < g_num_cpus; i++) {
threads[i].is_online = true;
threads[i].core_id = i;
threads[i].ids_of_caches[0] = 0;
threads[i].ids_of_caches[1] = 1;
threads[i].ids_of_caches[2] = 2;
threads[i].ids_of_caches[3] = 3;

caches[caches_cnt].type = CACHE_TYPE_DATA;
caches[caches_cnt].level = 1;
caches[caches_cnt].size = 32 * 1024;
caches[caches_cnt].coherency_line_size = 64;
caches[caches_cnt].number_of_sets = 64;
caches[caches_cnt].physical_line_partition = 1;
threads[i].ids_of_caches[0] = caches_cnt;
caches_cnt++;

caches[caches_cnt].type = CACHE_TYPE_INSTRUCTION;
caches[caches_cnt].level = 1;
caches[caches_cnt].size = 32 * 1024;
caches[caches_cnt].coherency_line_size = 64;
caches[caches_cnt].number_of_sets = 64;
caches[caches_cnt].physical_line_partition = 1;
threads[i].ids_of_caches[1] = caches_cnt;
caches_cnt++;

caches[caches_cnt].type = CACHE_TYPE_UNIFIED;
caches[caches_cnt].level = 2;
caches[caches_cnt].size = 256 * 1024;
caches[caches_cnt].coherency_line_size = 64;
caches[caches_cnt].number_of_sets = 1024;
caches[caches_cnt].physical_line_partition = 1;
threads[i].ids_of_caches[2] = caches_cnt;
caches_cnt++;

threads[i].ids_of_caches[3] = /* shared L3 */0;

cores[i].socket_id = 0;
cores[i].node_id = 0;
Expand All @@ -79,7 +107,7 @@ int pal_common_get_topo_info(struct pal_topo_info* topo_info) {
topo_info->numa_nodes = numa_nodes;
topo_info->numa_distance_matrix = distances;

topo_info->caches_cnt = 4;
topo_info->caches_cnt = /* per-CPU L1d, L1i, L2 */g_num_cpus * 3 + /* shared L3 */1;
topo_info->threads_cnt = g_num_cpus;
topo_info->cores_cnt = g_num_cpus;
topo_info->sockets_cnt = 1;
Expand Down