cpu: Change thread assignments for heterogenous SMT

Trying to run an SE system with varying threads per core (SMT cores + Non-SMT
cores) caused failures due to the CPU id assignment logic.  The comment
about thread assignment (worrying about core 0 not having tid 0) seems
not to be valid given that our configuration scripts initialize them in
order.

This removes that constraint so a heterogenously threaded sytem can work.

src/cpu/base.cc

@@ -436,21 +436,17 @@ BaseCPU::getMasterPort(const string &if_name, PortID idx)
 void
 BaseCPU::registerThreadContexts()
 {
+    assert(system->multiThread || numThreads == 1);
+
     ThreadID size = threadContexts.size();
     for (ThreadID tid = 0; tid < size; ++tid) {
         ThreadContext *tc = threadContexts[tid];
 
-        /** This is so that contextId and cpuId match where there is a
-         * 1cpu:1context relationship.  Otherwise, the order of registration
-         * could affect the assignment and cpu 1 could have context id 3, for
-         * example.  We may even want to do something like this for SMT so that
-         * cpu 0 has the lowest thread contexts and cpu N has the highest, but
-         * I'll just do this for now
-         */
-        if (numThreads == 1)
-            tc->setContextId(system->registerThreadContext(tc, _cpuId));
-        else
+        if (system->multiThread) {
             tc->setContextId(system->registerThreadContext(tc));
+        } else {
+            tc->setContextId(system->registerThreadContext(tc, _cpuId));
+        }
 
         if (!FullSystem)
             tc->getProcessPtr()->assignThreadContext(tc->contextId());

src/sim/System.py

@@ -99,6 +99,9 @@ class System(MemObject):
             "Address to mask loading binaries with")
     load_offset = Param.UInt64(0, "Address to offset loading binaries with")
 
+    multi_thread = Param.Bool(False,
+            "Supports multi-threaded CPUs? Impacts Thread/Context IDs")
+
     # Dynamic voltage and frequency handler for the system, disabled by default
     # Provide list of domains that need to be controlled by the handler
     dvfs_handler = DVFSHandler()

src/sim/system.cc

@@ -80,6 +80,7 @@ int System::numSystemsRunning = 0;
 System::System(Params *p)
     : MemObject(p), _systemPort("system_port", this),
       _numContexts(0),
+      multiThread(p->multi_thread),
       pagePtr(0),
       init_param(p->init_param),
       physProxy(_systemPort, p->cache_line_size),

src/sim/system.hh

@@ -196,6 +196,7 @@ class System : public MemObject
 
     std::vector<ThreadContext *> threadContexts;
     int _numContexts;
+    const bool multiThread;
 
     ThreadContext *getThreadContext(ContextID tid)
     {