AI/rocm-systems
2
0
Форкнуть 0
Код Задачи Запросы на слияние Действия Пакеты Проекты Релизы Вики Активность

P2P checkpoint.

Просмотр исходного кода 
Maintain enabled peer tables for each device.
Этот коммит содержится в:
Ben Sander
2016-04-06 15:49:32 -05:00
родитель b02e9163ab
Коммит 36926e6233
6 изменённых файлов: 96 добавлений и 52 удалений
Скачать Patch файл Скачать Diff файл Показать все файлы Свернуть все файлы
src/hip_error.cpp
+1 -1
Просмотреть файл
@@ -65,7 +65,7 @@ const char *hipGetErrorName(hipError_t hip_error)
//---
const char *hipGetErrorString(hipError_t hip_error)
{
std::call_once(hip_initialized, ihipInit);
HIP_INIT_API(hip_error);
// TODO - return a message explaining the error.
// TODO - This should be set up to return the same string reported in the the doxygen comments, somehow.
src/hip_hcc.cpp
+31 -21
Просмотреть файл
@@ -211,24 +211,31 @@ void ihipDeviceCriticalBase_t<DeviceMutex>::recomputePeerAgents()
template<>
void ihipDeviceCriticalBase_t<DeviceMutex>::addPeer(ihipDevice_t *peer)
bool ihipDeviceCriticalBase_t<DeviceMutex>::addPeer(ihipDevice_t *peer)
{
auto match = std::find(_peers.begin(), _peers.end(), peer);
if (match != std::end(_peers)) {
_peers.push_back(peer);
recomputePeerAgents();
return;
return true;
}
// If we get here - peer was already on list, silently ignore.
return false;
}
template<>
void ihipDeviceCriticalBase_t<DeviceMutex>::removePeer(ihipDevice_t *peer)
bool ihipDeviceCriticalBase_t<DeviceMutex>::removePeer(ihipDevice_t *peer)
{
_peers.remove(peer);
recomputePeerAgents();
auto match = std::find(_peers.begin(), _peers.end(), peer);
if (match != std::end(_peers)) {
_peers.remove(peer);
recomputePeerAgents();
return true;
} else {
return false;
}
}
//-------------------------------------------------------------------------------------------------
@@ -1073,22 +1080,25 @@ void ihipPostLaunchKernel(hipStream_t stream, hc::completion_future &kernelFutur
const char *ihipErrorString(hipError_t hip_error)
{
switch (hip_error) {
case hipSuccess : return "hipSuccess";
case hipErrorMemoryAllocation : return "hipErrorMemoryAllocation";
case hipErrorMemoryFree : return "hipErrorMemoryFree";
case hipErrorUnknownSymbol : return "hipErrorUnknownSymbol";
case hipErrorOutOfResources : return "hipErrorOutOfResources";
case hipErrorInvalidValue : return "hipErrorInvalidValue";
case hipErrorInvalidResourceHandle : return "hipErrorInvalidResourceHandle";
case hipErrorInvalidDevice : return "hipErrorInvalidDevice";
case hipErrorInvalidMemcpyDirection : return "hipErrorInvalidMemcpyDirection";
case hipErrorNoDevice : return "hipErrorNoDevice";
case hipErrorNotReady : return "hipErrorNotReady";
case hipErrorRuntimeMemory : return "hipErrorRuntimeMemory";
case hipErrorRuntimeOther : return "hipErrorRuntimeOther";
case hipErrorUnknown : return "hipErrorUnknown";
case hipErrorTbd : return "hipErrorTbd";
default : return "hipErrorUnknown";
case hipSuccess : return "hipSuccess";
case hipErrorMemoryAllocation : return "hipErrorMemoryAllocation";
case hipErrorMemoryFree : return "hipErrorMemoryFree";
case hipErrorUnknownSymbol : return "hipErrorUnknownSymbol";
case hipErrorOutOfResources : return "hipErrorOutOfResources";
case hipErrorInvalidValue : return "hipErrorInvalidValue";
case hipErrorInvalidResourceHandle : return "hipErrorInvalidResourceHandle";
case hipErrorInvalidDevice : return "hipErrorInvalidDevice";
case hipErrorInvalidMemcpyDirection : return "hipErrorInvalidMemcpyDirection";
case hipErrorNoDevice : return "hipErrorNoDevice";
case hipErrorNotReady : return "hipErrorNotReady";
case hipErrorPeerAccessNotEnabled : return "hipErrorPeerAccessNotEnabled";
case hipErrorPeerAccessAlreadyEnabled : return "hipErrorPeerAccessAlreadyEnabled";
case hipErrorRuntimeMemory : return "hipErrorRuntimeMemory";
case hipErrorRuntimeOther : return "hipErrorRuntimeOther";
case hipErrorUnknown : return "hipErrorUnknown";
case hipErrorTbd : return "hipErrorTbd";
default : return "hipErrorUnknown";
};
};
src/hip_peer.cpp
+43 -17
Просмотреть файл
@@ -26,7 +26,7 @@ THE SOFTWARE.
* @warning HCC returns 0 in *canAccessPeer ; Need to update this function when RT supports P2P
*/
//---
hipError_t hipDeviceCanAccessPeer ( int* canAccessPeer, int deviceId, int peerDeviceId)
hipError_t hipDeviceCanAccessPeer (int* canAccessPeer, int deviceId, int peerDeviceId)
{
HIP_INIT_API(canAccessPeer, deviceId, peerDeviceId);
@@ -38,7 +38,7 @@ hipError_t hipDeviceCanAccessPeer ( int* canAccessPeer, int deviceId, int peerD
if ((device != NULL) && (peerDevice != NULL)) {
#if USE_PEER_TO_PEER==2
*canAccessPeer = peerDevice->_acc.is_peer(device->_acc);
*canAccessPeer = peerDevice->_acc.get_is_peer(device->_acc);
#else
*canAccessPeer = 0;
#endif
@@ -56,16 +56,32 @@ hipError_t hipDeviceCanAccessPeer ( int* canAccessPeer, int deviceId, int peerD
}
/**
* warning Need to update this function when RT supports P2P
*/
//---
hipError_t hipDeviceDisablePeerAccess ( int peerDevice )
hipError_t hipDeviceDisablePeerAccess (int peerDeviceId)
{
HIP_INIT_API(peerDevice);
HIP_INIT_API(peerDeviceId);
// TODO-p2p
return ihipLogStatus(hipSuccess);
hipError_t err = hipSuccess;
#if USE_PEER_TO_PEER
auto thisDevice = ihipGetTlsDefaultDevice();
auto peerDevice = ihipGetDevice(peerDeviceId);
if ((thisDevice != NULL) && (peerDevice != NULL)) {
LockedAccessor_DeviceCrit_t crit(thisDevice->criticalData());
bool changed = crit->removePeer(peerDevice);
if (changed) {
#if USE_PEER_TO_PEER==2
am_memtracker_update_peers(device->_acc, crit->peerCnt(), crit->peerAgents());
#endif
} else {
err = hipErrorPeerAccessNotEnabled; // never enabled P2P access.
}
} else {
err = hipErrorInvalidDevice;
}
#endif
return ihipLogStatus(err);
};
@@ -74,18 +90,27 @@ hipError_t hipDeviceDisablePeerAccess ( int peerDevice )
*/
//---
// Enable registering memory on peerDevice for direct access from the current device.
hipError_t hipDeviceEnablePeerAccess (int peerDeviceId, unsigned int flags)
hipError_t hipDeviceEnablePeerAccess (int peerDeviceId, unsigned int flags)
{
std::call_once(hip_initialized, ihipInit);
HIP_INIT_API(peerDeviceId, flags);
hipError_t err = hipSuccess;
#if USE_PEER_TO_PEER
if (flags != 0) {
err = hipErrorInvalidValue;
} else {
auto thisDevice = ihipGetTlsDefaultDevice();
auto peerDevice = ihipGetDevice(peerDeviceId);
if (peerDevice != NULL) {
if ((thisDevice != NULL) && (peerDevice != NULL)) {
LockedAccessor_DeviceCrit_t crit(thisDevice->criticalData());
bool isNewPeer = crit->addPeer(peerDevice);
if (isNewPeer) {
#if USE_PEER_TO_PEER==2
am_memtracker_update_peers(device->_acc, crit->peerCnt(), crit->peerAgents());
#endif
} else {
err = hipErrorPeerAccessAlreadyEnabled;
}
} else {
err = hipErrorInvalidDevice;
}
@@ -97,9 +122,10 @@ hipError_t hipDeviceEnablePeerAccess (int peerDeviceId, unsigned int flags)
//---
hipError_t hipMemcpyPeer ( void* dst, int dstDevice, const void* src, int srcDevice, size_t sizeBytes )
hipError_t hipMemcpyPeer (void* dst, int dstDevice, const void* src, int srcDevice, size_t sizeBytes)
{
std::call_once(hip_initialized, ihipInit);
HIP_INIT_API(dst, dstDevice, src, srcDevice, sizeBytes);
// HCC has a unified memory architecture so device specifiers are not required.
return hipMemcpy(dst, src, sizeBytes, hipMemcpyDefault);
};
@@ -109,9 +135,9 @@ hipError_t hipMemcpyPeer ( void* dst, int dstDevice, const void* src, int srcD
* @bug This function uses a synchronous copy
*/
//---
hipError_t hipMemcpyPeerAsync ( void* dst, int dstDevice, const void* src, int srcDevice, size_t sizeBytes, hipStream_t stream )
hipError_t hipMemcpyPeerAsync (void* dst, int dstDevice, const void* src, int srcDevice, size_t sizeBytes, hipStream_t stream)
{
std::call_once(hip_initialized, ihipInit);
HIP_INIT_API(dst, dstDevice, src, srcDevice, sizeBytes, stream);
// HCC has a unified memory architecture so device specifiers are not required.
return hipMemcpyAsync(dst, src, sizeBytes, hipMemcpyDefault, stream);
};