mirror of
https://github.com/tangerineturtle1/NX2navi.git
synced 2026-11-17 04:59:57 -04:00
Run the game on the die with the larger L3 cache on processors whose dies differ (AMD's two-die X3D processors).
Windows places a game on the V-Cache die only when it recognizes the exe as a game, and it does not know this one, so the threads were spread over both dies by frequency preference. cache_die = 1 (the default) narrows the process to the logical processors that share the largest L3; it never widens the set the process was given, and does nothing where every L3 has the same size. The recorder's reserved core is then chosen inside that set. Co-Authored-By: Claude Fable 5.2 <[EMAIL]>
This commit is contained in:
@@ -155,6 +155,40 @@ int plat_cpu_cores(uint64_t *all, uint64_t *first) {
|
||||
return cores;
|
||||
}
|
||||
|
||||
// The L3 cache of each logical processor: the cache/index<k> of /sys/devices/system/cpu/cpu<n> whose level is 3, its
|
||||
// size in KiB ("98304K").
|
||||
uint64_t plat_cpu_cache_die(uint32_t *l3_kb) {
|
||||
cpu_set_t set;
|
||||
*l3_kb = 0;
|
||||
if (sched_getaffinity(0, sizeof set, &set)) return 0;
|
||||
uint64_t die = 0, in_set = 0;
|
||||
uint32_t big = 0, small = ~0u;
|
||||
for (int i = 0; i < 64; i++) {
|
||||
uint32_t kb = 0;
|
||||
for (int k = 0; k < 8 && !kb; k++) {
|
||||
char path[96], text[32] = "";
|
||||
uint64_t level = 0;
|
||||
snprintf(path, sizeof path, "/sys/devices/system/cpu/cpu%d/cache/index%d/level", i, k);
|
||||
if (!read_u64_file(path, &level)) break;
|
||||
if (level != 3) continue;
|
||||
snprintf(path, sizeof path, "/sys/devices/system/cpu/cpu%d/cache/index%d/size", i, k);
|
||||
FILE *f = fopen(path, "r");
|
||||
if (f) {
|
||||
if (fgets(text, sizeof text, f)) kb = (uint32_t)strtoul(text, NULL, 10) * (strchr(text, 'M') ? 1024 : 1);
|
||||
fclose(f);
|
||||
}
|
||||
}
|
||||
if (!kb) continue;
|
||||
if (kb > big) big = kb, die = 0;
|
||||
if (kb == big) die |= 1ull << i;
|
||||
if (kb < small) small = kb;
|
||||
if (CPU_ISSET(i, &set)) in_set |= 1ull << i;
|
||||
}
|
||||
if (!die || big == small) return 0;
|
||||
*l3_kb = big;
|
||||
return die & in_set;
|
||||
}
|
||||
|
||||
// Linux has no process-wide affinity: each thread of /proc/self/task gets the mask, and threads started later inherit
|
||||
// it from the thread that starts them.
|
||||
int plat_process_affinity(uint64_t mask) {
|
||||
|
||||
@@ -41,6 +41,10 @@ uint64_t plat_thread_cpu_ns(uint32_t tid); // CPU time of the thread with id ti
|
||||
// The logical processors the process may use (bit n = processor n, the first 64), into *all, and the lowest of each
|
||||
// physical core among them into *first; returns the number of physical cores, 0 when unknown.
|
||||
int plat_cpu_cores(uint64_t *all, uint64_t *first);
|
||||
// The logical processors of the process's set (the first 64) that share the largest L3 cache, when the L3 caches
|
||||
// are not all the same size: the die with AMD's 3D V-Cache on a two-die X3D processor. 0 when they are (one die, or
|
||||
// equal dies) or the topology is unknown. *l3_kb: that cache's size.
|
||||
uint64_t plat_cpu_cache_die(uint32_t *l3_kb);
|
||||
int plat_process_affinity(uint64_t mask); // every thread of the process to the processors in mask; 1 on success
|
||||
|
||||
// Interrupting a thread (kernel.c's asynchronous preemption). A thread makes its own handle; another thread stops it
|
||||
|
||||
@@ -95,6 +95,32 @@ int plat_cpu_cores(uint64_t *all, uint64_t *first) {
|
||||
return cores;
|
||||
}
|
||||
|
||||
uint64_t plat_cpu_cache_die(uint32_t *l3_kb) {
|
||||
DWORD_PTR proc, sys;
|
||||
*l3_kb = 0;
|
||||
if (!GetProcessAffinityMask(GetCurrentProcess(), &proc, &sys)) return 0;
|
||||
DWORD len = 0;
|
||||
GetLogicalProcessorInformationEx(RelationCache, NULL, &len);
|
||||
SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX *buf = len ? malloc(len) : NULL;
|
||||
uint64_t die = 0;
|
||||
DWORD big = 0, small = ~0u;
|
||||
if (buf && GetLogicalProcessorInformationEx(RelationCache, buf, &len))
|
||||
for (DWORD off = 0; off < len;) {
|
||||
SYSTEM_LOGICAL_PROCESSOR_INFORMATION_EX *e = (void *)((char *)buf + off);
|
||||
const CACHE_RELATIONSHIP *c = &e->Cache;
|
||||
if (c->Level == 3 && c->GroupMask.Group == 0) { // every L3, also those outside the process's set
|
||||
if (c->CacheSize > big) big = c->CacheSize, die = 0;
|
||||
if (c->CacheSize == big) die |= c->GroupMask.Mask;
|
||||
if (c->CacheSize < small) small = c->CacheSize;
|
||||
}
|
||||
off += e->Size;
|
||||
}
|
||||
free(buf);
|
||||
if (!die || big == small) return 0;
|
||||
*l3_kb = big >> 10;
|
||||
return die & proc;
|
||||
}
|
||||
|
||||
int plat_process_affinity(uint64_t mask) { return SetProcessAffinityMask(GetCurrentProcess(), (DWORD_PTR)mask) != 0; }
|
||||
|
||||
uint64_t plat_thread_cpu_ns(uint32_t tid) {
|
||||
|
||||
Reference in New Issue
Block a user