diff --git a/common/common.cpp b/common/common.cpp
index 309552b1a..963f5ccb8 100644
--- a/common/common.cpp
+++ b/common/common.cpp
@@ -231,6 +231,8 @@ int32_t setCpuAffinity(std::bitset<64> cpuMask) {
         if (cpuMask[i] == 1) {
             CPUSET_PRINT_DEBUG("Setting CPU %d\n", i);
             CPU_SET(i, &mask);
+        } else {
+            CPU_CLR(i, &mask);
         }
     }
 
@@ -328,7 +330,10 @@ int32_t get_num_physical_cores() {
         is_hybrid_core = false;
         if (is_hybrid) {
             if (pin_cpu(cpu) == 0) {
-                if (is_running_on_efficiency_core()) is_hybrid_core = true;
+                if (is_running_on_efficiency_core()) {
+                    is_hybrid_core = true;
+                    CPUSET_PRINT_DEBUG("Logical CPU is Hybrid: %d\n", cpu);
+                }
             }
         }
         numLogicalCores++;
@@ -365,6 +370,7 @@ int32_t get_num_physical_cores() {
                 CPU_SET(cpu, &mask);
                 CPUSET_PRINT_DEBUG("CPU %u is physical, siblings: %s\n", cpu, line.c_str());
             } else {
+                CPU_CLR(cpu, &mask);
                 cpuset_smt = true;
             }
         }
@@ -378,7 +384,7 @@ int32_t get_num_physical_cores() {
         if (pthread_setaffinity_np(pthread_self(), sizeof(mask), &mask) == -1) {
                 CPUSET_PRINT_DEBUG("pthread_setaffinity_np error\n");
         }
-        fprintf(stderr, "get_num_physical_cores Physical CPU count: %li\n", siblings.size());
+        CPUSET_PRINT_DEBUG("get_num_physical_cores Physical CPU count: %li\n", siblings.size());
 
         cpuset_best = cpuset;
         cpuset_worst = cpuset;
diff --git a/ggml.c b/ggml.c
index b5b11ca16..ef3200b09 100644
--- a/ggml.c
+++ b/ggml.c
@@ -18708,6 +18708,33 @@ enum ggml_status ggml_graph_compute(struct ggml_cgraph * cgraph, struct ggml_cpl
             };
 
             const int rc = ggml_thread_create(&workers[j].thrd, NULL, ggml_graph_compute_thread, &workers[j]);
+#if defined(__x86_64__) && defined(__linux__)
+            cpu_set_t procMask;
+            cpu_set_t threadMask;
+            //fprintf(stdout, "sched_getaffinity init\n");
+            if (sched_getaffinity(0, sizeof(cpu_set_t), &procMask) == -1) {
+                fprintf(stderr, "ggml_thread_create sched_getaffinity error\n");
+            } else {
+                int result = pthread_setaffinity_np(workers[j].thrd, sizeof(cpu_set_t), &procMask);
+                if (result !=0) fprintf(stderr, "ggml_thread_create pthread_setaffinity_np: %d", result);
+                //printf("Set returned by sched_getaffinity() contained:\n");
+                //for (size_t k = 0; k < CPU_SETSIZE; k++)
+                    //if (CPU_ISSET(k, &procMask))
+                        //printf("    CPU %zu\n", k);
+            }
+            /*
+            int s;
+            s = pthread_getaffinity_np(workers[j].thrd, sizeof(threadMask), &threadMask);
+            if (s != 0) {
+               fprintf(stderr, "ggml_thread_create pthread_getaffinity_np: %d\n", s);
+            } else {
+                printf("Set returned by pthread_getaffinity_np() contained:\n");
+                for (size_t l = 0; l < CPU_SETSIZE; l++)
+                    if (CPU_ISSET(l, &threadMask))
+                        printf("    CPU %zu\n", l);
+            } 
+            */           
+#endif
             GGML_ASSERT(rc == 0);
             UNUSED(rc);
         }