Context management related changes in HIP.

-
-Contexts across threads are listed under device
-Device reset cleans up all contexts and re-initializes _primaryCtx

Change-Id: Ie1cfbb26d43a8dc6869be3e6ebaf7344ce374643


[ROCm/clr commit: c837b8d713]

projects/clr/hipamd/src/hip_context.cpp

@@ -58,9 +58,15 @@ hipError_t hipCtxCreate(hipCtx_t *ctx, unsigned int flags,  hipDevice_t device)
     HIP_INIT_API(ctx, flags, device); // FIXME - review if we want to init
     hipError_t e = hipSuccess;
     auto deviceHandle = ihipGetDevice(device);
-    *ctx = new ihipCtx_t(deviceHandle, g_deviceCnt, flags);
-    ihipSetTlsDefaultCtx(*ctx);
-    tls_ctxStack.push(*ctx);
+    {
+		// Obtain mutex access to the device critical data, release by destructor
+		LockedAccessor_DeviceCrit_t  deviceCrit(deviceHandle->criticalData());
+		auto ictx = new ihipCtx_t(deviceHandle, g_deviceCnt, flags);
+		*ctx = ictx;
+		ihipSetTlsDefaultCtx(*ctx);
+		tls_ctxStack.push(*ctx);
+		deviceCrit->addContext(ictx);
+    }
 
     return ihipLogStatus(e);
 }
@@ -125,6 +131,11 @@ hipError_t hipCtxDestroy(hipCtx_t ctx)
             //need to destroy the ctx associated with calling thread
             tls_ctxStack.pop();
         }
+        {
+			auto deviceHandle = ctx->getWriteableDevice();
+			deviceHandle->locked_removeContext(ctx);
+			ctx->locked_reset();
+    	}
         delete ctx; //As per CUDA docs , attempting to access ctx from those threads which has this ctx as current, will result in the error HIP_ERROR_CONTEXT_IS_DESTROYED.
     }
 

projects/clr/hipamd/src/hip_device.cpp

@@ -166,11 +166,16 @@ hipError_t hipDeviceReset(void)
     // This function currently does a user-level cleanup of known resources.
     // It could benefit from KFD support to perform a more "nuclear" clean that would include any associated kernel resources and page table entries.
 
-
+#if 0
     if (ctx) {
         // Release ctx resources (streams and memory):
         ctx->locked_reset();
     }
+#endif
+	if (ctx) {
+		ihipDevice_t *deviceHandle = ctx->getWriteableDevice();
+		deviceHandle->locked_reset();
+	}
 
     return ihipLogStatus(hipSuccess);
 }

projects/clr/hipamd/src/hip_hcc.cpp

@@ -522,6 +522,14 @@ void ihipCtxCriticalBase_t<CtxMutex>::addStream(ihipStream_t *stream)
     _streams.push_back(stream);
     tprintf(DB_SYNC, " addStream: %s\n", ToString(stream).c_str());
 }
+
+template<>
+void ihipDeviceCriticalBase_t<DeviceMutex>::addContext(ihipCtx_t *ctx)
+{
+    _ctxs.push_back(ctx);
+    tprintf(DB_SYNC, " addContext: %s\n", ToString(ctx).c_str());
+}
+
 //=============================================================================
 
 //=================================================================================================
@@ -530,7 +538,8 @@ void ihipCtxCriticalBase_t<CtxMutex>::addStream(ihipStream_t *stream)
 ihipDevice_t::ihipDevice_t(unsigned deviceId, unsigned deviceCnt, hc::accelerator &acc) :
     _deviceId(deviceId),
     _acc(acc),
-    _state(0)
+    _state(0),
+    _criticalData(this)
 {
     hsa_agent_t *agent = static_cast<hsa_agent_t*> (acc.get_hsa_agent());
     if (agent) {
@@ -557,7 +566,49 @@ ihipDevice_t::~ihipDevice_t()
     _primaryCtx = NULL;
 }
 
+void ihipDevice_t::locked_removeContext(ihipCtx_t *c)
+{
+    LockedAccessor_DeviceCrit_t  crit(_criticalData);
 
+    crit->ctxs().remove(c);
+    tprintf(DB_SYNC, " locked_removeContext: %s\n", ToString(c).c_str());
+}
+
+
+void ihipDevice_t::locked_reset()
+{
+    // Obtain mutex access to the device critical data, release by destructor
+    LockedAccessor_DeviceCrit_t  crit(_criticalData);
+
+
+    //---
+    //Wait for pending activity to complete?  TODO - check if this is required behavior:
+    tprintf(DB_SYNC, "locked_reset waiting for activity to complete.\n");
+
+    // Reset and remove streams:
+    // Delete all created streams including the default one.
+    for (auto ctxI=crit->const_ctxs().begin(); ctxI!=crit->const_ctxs().end(); ctxI++) {
+        ihipCtx_t *ctx = *ctxI;
+        (*ctxI)->locked_reset();
+        tprintf(DB_SYNC, " ctx cleanup %s\n", ToString(ctx).c_str());
+
+        delete ctx;
+    }
+    // Clear the list.
+    crit->ctxs().clear();
+
+
+    //reset _primaryCtx
+    _primaryCtx->locked_reset();
+    tprintf(DB_SYNC, " _primaryCtx cleanup %s\n", ToString(_primaryCtx).c_str());
+    // Reset and release all memory stored in the tracker:
+    // Reset will remove peer mapping so don't need to do this explicitly.
+    // FIXME - This is clearly a non-const action!  Is this a context reset or a device reset - maybe should reference count?
+
+    _state = 0;
+    am_memtracker_reset(_acc);
+
+};
 
 #define ErrorCheck(x) error_check(x, __LINE__, __FILE__)
 
@@ -861,8 +912,14 @@ ihipCtx_t::ihipCtx_t(ihipDevice_t *device, unsigned deviceCnt, unsigned flags) :
     _device(device),
     _criticalData(this, deviceCnt)
 {
-    locked_reset();
+    //locked_reset();
+    LockedAccessor_CtxCrit_t  crit(_criticalData);
+	_defaultStream = new ihipStream_t(this, getDevice()->_acc.get_default_view(), hipStreamDefault);
+	crit->addStream(_defaultStream);
 
+
+		// Reset peer list to just me:
+    crit->resetPeerWatchers(this);
     tprintf(DB_SYNC, "created ctx with defaultStream=%p (%s)\n", _defaultStream, ToString(_defaultStream).c_str());
 };
 
@@ -905,7 +962,7 @@ void ihipCtx_t::locked_reset()
     _defaultStream = new ihipStream_t(this, getDevice()->_acc.get_default_view(), hipStreamDefault);
     crit->addStream(_defaultStream);
 
-
+#if 0
     // Reset peer list to just me:
     crit->resetPeerWatchers(this);
 
@@ -915,7 +972,7 @@ void ihipCtx_t::locked_reset()
     ihipDevice_t *device = getWriteableDevice();
     device->_state = 0;
     am_memtracker_reset(device->_acc);
-
+#endif
 };
 
 

projects/clr/hipamd/src/hip_hcc.h

@@ -143,6 +143,8 @@ extern const char *API_COLOR_END;
 
 #define CTX_THREAD_SAFE 1
 
+#define DEVICE_THREAD_SAFE 1
+
 
 // Compile debug trace mode - this prints debug messages to stderr when env var HIP_DB is set.
 // May be set to 0 to remove debug if checks - possible code size and performance difference?
@@ -371,6 +373,13 @@ typedef FakeMutex StreamMutex;
 typedef std::mutex CtxMutex;
 #else
 typedef FakeMutex CtxMutex;
+#warning "Ctx thread-safe disabled"
+#endif
+
+#if DEVICE_THREAD_SAFE
+typedef std::mutex DeviceMutex;
+#else
+typedef FakeMutex DeviceMutex;
 #warning "Device thread-safe disabled"
 #endif
 
@@ -595,7 +604,40 @@ struct ihipEvent_t {
 
 
 
+//=============================================================================
+//class ihipDeviceCriticalBase_t
+template <typename MUTEX_TYPE>
+class ihipDeviceCriticalBase_t : LockedBase<MUTEX_TYPE>
+{
+public:
+    ihipDeviceCriticalBase_t(ihipDevice_t *parentDevice) :
+        _parent(parentDevice)
+    {
+    };
 
+    ~ihipDeviceCriticalBase_t()  {
+
+    }
+
+    // Contexts:
+    void addContext(ihipCtx_t *ctx);
+    void removeContext(ihipCtx_t *ctx);
+    std::list<ihipCtx_t*> &ctxs() { return _ctxs; };
+    const std::list<ihipCtx_t*> &const_ctxs() const { return _ctxs; };
+	int getcount() {return _ctxCount;};
+    friend class LockedAccessor<ihipDeviceCriticalBase_t>;
+private:
+    ihipDevice_t                 *_parent;
+
+    //--- Context Tracker:
+    std::list< ihipCtx_t* > _ctxs;   // contexts associated with this device across all threads.
+
+	int	_ctxCount;
+};
+
+typedef ihipDeviceCriticalBase_t<DeviceMutex> ihipDeviceCritical_t;
+
+typedef LockedAccessor<ihipDeviceCritical_t> LockedAccessor_DeviceCrit_t;
 
 //----
 // Properties of the HIP device.
@@ -608,7 +650,9 @@ public:
 
     // Accessors:
     ihipCtx_t *getPrimaryCtx() const { return _primaryCtx; };
-
+    void locked_removeContext(ihipCtx_t *c);
+	void locked_reset();
+    ihipDeviceCritical_t  &criticalData() { return _criticalData; };
 public:
     unsigned                _deviceId; // device ID
 
@@ -628,6 +672,8 @@ public:
 
 private:
     hipError_t initProperties(hipDeviceProp_t* prop);
+private:
+	ihipDeviceCritical_t       _criticalData;
 };
 //=============================================================================