linux: compute CCD assignment for AMD cpus

leahneukirchen · Sep 1, 2024 · edc4417 · edc4417
1 parent 6ba9905
commit edc4417
Show file tree

Hide file tree

Showing 3 changed files with 58 additions and 0 deletions.
diff --git a/linux/LibSensors.c b/linux/LibSensors.c
@@ -156,6 +156,62 @@ static int tempDriverPriority(const sensors_chip_name* chip) {
    return -1;
 }
 
+void LibSensors_countCCDs(CPUData* cpus, unsigned int existingCPUs) {
+   /* For AMD k10temp/zenpower, temperatures are provided for CCDs only,
+      which is an aggregate of multiple cores.
+      There's no obvious mapping between hwmon sensors and sockets and CCDs.
+      Assume both are iterated in order.
+      Hypothesis: Each CCD has same size N = #Cores/#CCD
+      and is assigned N coreID in sequence.
+      Also assume all CPUs have same number of CCDs. */
+
+   int ccds = 0;
+
+   for (size_t i = 0; i < existingCPUs + 1; i++) {
+      cpus[i].ccdID = -1;
+   }
+
+   int n = 0;
+   for (const sensors_chip_name* chip = sym_sensors_get_detected_chips(NULL, &n); chip; chip = sym_sensors_get_detected_chips(NULL, &n)) {
+      int m = 0;
+      for (const sensors_feature* feature = sym_sensors_get_features(chip, &m); feature; feature = sym_sensors_get_features(chip, &m)) {
+         if (feature->type != SENSORS_FEATURE_TEMP)
+            continue;
+
+         if (!feature->name || !String_startsWith(feature->name, "temp"))
+            continue;
+
+         char *label = sym_sensors_get_label(chip, feature);
+         if (label) {
+            if (String_startsWith(label, "Tccd")) {
+               ccds++;
+            }
+            free(label);
+         }
+      }
+   }
+
+   if (ccds > 0) {
+      int ccdsPerCore = existingCPUs / ccds;
+
+      int ccd = 0;
+      int nc = ccdsPerCore;
+      for (int p = 0; p < (int)existingCPUs; p++) {
+         for (int c = 0; c < (int)existingCPUs; c++) {
+            for (size_t i = 1; i <= existingCPUs; i++) {
+               if (cpus[i].physicalID == p && cpus[i].coreID == c) {
+                  cpus[i].ccdID = ccd;
+                  if (--nc == 0) {
+                     nc = ccdsPerCore;
+                     ccd++;
+                  }
+               }
+            }
+         }
+      }
+   }
+}
+
 void LibSensors_getCPUTemperatures(CPUData* cpus, unsigned int existingCPUs, unsigned int activeCPUs) {
    assert(existingCPUs > 0 && existingCPUs < 16384);
 

diff --git a/linux/LibSensors.h b/linux/LibSensors.h
@@ -14,6 +14,7 @@ int LibSensors_init(void);
 void LibSensors_cleanup(void);
 int LibSensors_reload(void);
 
+void LibSensors_countCCDs(CPUData* cpus, unsigned int existingCPUs);
 void LibSensors_getCPUTemperatures(CPUData* cpus, unsigned int existingCPUs, unsigned int activeCPUs);
 
 #endif /* HEADER_LibSensors */
diff --git a/linux/LinuxMachine.h b/linux/LinuxMachine.h
@@ -51,6 +51,7 @@ typedef struct CPUData_ {
 
    int physicalID;		/* different for each CPU socket */
    int coreID;			/* same for hyperthreading */
+   int ccdID;			/* same for each AMD chiplet */
    #endif
 
    bool online;