From 6333fdecf3a4687ddab3bccec6d78072bd67faf9 Mon Sep 17 00:00:00 2001
From: David Yat Sin <David.YatSin@amd.com>
Date: Fri, 1 Dec 2023 20:42:33 +0000
Subject: [PATCH] Use pthread_setaffinity_np

On some systems, pthread_addr_setaffinity_np does not exist, so we need
to use pthread_setaffinity_np on thread after pthread_create

Provided by Julian Samaroo on github

https: //github.com/RadeonOpenCompute/ROCR-Runtime/pull/143
Change-Id: I4649f94333f2d7b0a5993b370a4bfc48d92acecb
---
 .../hsa-runtime/core/util/lnx/os_linux.cpp    | 40 +++++++++----------
 1 file changed, 20 insertions(+), 20 deletions(-)

diff --git a/runtime/hsa-runtime/core/util/lnx/os_linux.cpp b/runtime/hsa-runtime/core/util/lnx/os_linux.cpp
index bdf76657eb..54a9c99ffc 100644
--- a/runtime/hsa-runtime/core/util/lnx/os_linux.cpp
+++ b/runtime/hsa-runtime/core/util/lnx/os_linux.cpp
@@ -110,26 +110,7 @@ class os_thread {
         return;
       }
     }
-
-    if (core::Runtime::runtime_singleton_->flag().override_cpu_affinity()) {
-      int cores = get_nprocs_conf();
-      cpu_set_t* cpuset = CPU_ALLOC(cores);
-      if (cpuset == nullptr) {
-        fprintf(stderr, "CPU_ALLOC failed: %s\n", strerror(errno));
-        return;
-      }
-      CPU_ZERO_S(CPU_ALLOC_SIZE(cores), cpuset);
-      for (int i = 0; i < cores; i++) {
-        CPU_SET_S(i, CPU_ALLOC_SIZE(cores), cpuset);
-      }
-      err = pthread_attr_setaffinity_np(&attrib, CPU_ALLOC_SIZE(cores), cpuset);
-      CPU_FREE(cpuset);
-      if (err != 0) {
-        fprintf(stderr, "pthread_attr_setaffinity_np failed: %s\n", strerror(err));
-        return;
-      }
-    }
-
+\
     err = pthread_create(&thread, &attrib, ThreadTrampoline, args.get());
 
     // Probably a stack size error since system limits can be different from PTHREAD_STACK_MIN
@@ -148,6 +129,25 @@ class os_thread {
       }
     }
 
+    if (core::Runtime::runtime_singleton_->flag().override_cpu_affinity()) {
+      int cores = get_nprocs_conf();
+      cpu_set_t* cpuset = CPU_ALLOC(cores);
+      if (cpuset == nullptr) {
+        fprintf(stderr, "CPU_ALLOC failed: %s\n", strerror(errno));
+        return;
+      }
+      CPU_ZERO_S(CPU_ALLOC_SIZE(cores), cpuset);
+      for (int i = 0; i < cores; i++) {
+        CPU_SET_S(i, CPU_ALLOC_SIZE(cores), cpuset);
+      }
+      err = pthread_setaffinity_np(thread, CPU_ALLOC_SIZE(cores), cpuset);
+      CPU_FREE(cpuset);
+      if (err != 0) {
+        fprintf(stderr, "pthread_attr_setaffinity_np failed: %s\n", strerror(err));
+        return;
+      }
+    }
+
     if (err == 0)
       args.release();
     else