AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

P2P checkpoint.

Browse Source 
Maintain enabled peer tables for each device.
This commit is contained in:
Ben Sander
2016-04-06 15:49:32 -05:00
parent b02e9163ab
commit 36926e6233
6 changed files with 96 additions and 52 deletions
Download Patch File Download Diff File Expand all files Collapse all files
include/hcc_detail/hip_hcc.h
+9 -3
View File
@@ -519,9 +519,11 @@ public:
// "Allocate" a stream ID:
ihipStream_t::SeqNum_t incStreamId() { return _stream_id++; };
void recomputePeerAgents();
void addPeer(ihipDevice_t *peer);
void removePeer(ihipDevice_t *peer);
bool addPeer(ihipDevice_t *peer);
bool removePeer(ihipDevice_t *peer);
uint32_t peerCnt() const { return _peerCnt; };
uint32_t peerAgents() const { return _peerAgents; };
private:
@@ -532,6 +534,8 @@ private:
std::list<ihipDevice_t*> _peers; // list of enabled peer devices.
uint32_t _peerCnt; // number of enabled peers
hsa_agent_t *_peerAgents; // efficient packed array of enabled agents (to use for allocations.)
private:
void recomputePeerAgents();
};
// Note Mutex selected based on DeviceMutex
@@ -560,6 +564,8 @@ public: // Functions:
void locked_waitAllStreams();
void locked_syncDefaultStream(bool waitOnSelf);
ihipDeviceCritical_t &criticalData() { return _criticalData; }; // TODO, move private. Fix P2P.
public: // Data, set at initialization:
unsigned _device_index; // index into g_devices.
include/hcc_detail/hip_runtime_api.h
+9 -9
View File
@@ -913,21 +913,21 @@ hipError_t hipDeviceCanAccessPeer (int* canAccessPeer, int deviceId, int peerDev
/**
* @brief Disable registering memory on peerDevice for direct access from the current device.
* @brief Disable direct access from current device's virtual address space to memory allocations physically located on a peer device.
*
* If there are any allocations on peerDevice which were registered in the current device using hipPeerRegister() then these allocations will be automatically unregistered.
* Returns hipErrorPeerAccessNotEnabled if direct access to memory on peerDevice has not yet been enabled from the current device.
*
* @param [in] peerDevice
* @param [in] peerDeviceId
*
* TODO:cudaErrorPeerAccessNotEnabled and cudaErrorInvalidDevice error not supported in HIP, return hipErrorUnknown
* Returns #hipSuccess, #hipErrorUnknown
*/
hipError_t hipDeviceDisablePeerAccess (int peerDeviceId);
/**
* @brief Enable registering memory on peerDevice for direct access from the current device.
* @brief Enable direct access from current device's virtual address space to memory allocations physically located on a peer device.
*
* @param [in] peerDevice
* @param [in] peerDeviceId
* @param [in] flags
*
* TODO:cudaErrorInvalidDevice error not supported in HIP, return hipErrorUnknown
@@ -939,14 +939,14 @@ hipError_t hipDeviceEnablePeerAccess (int peerDeviceId, unsigned int flags);
* @brief Copies memory from one device to memory on another device.
*
* @param [out] dst - Destination device pointer.
* @param [in] dstDevice - Destination device
* @param [in] dstDeviceId - Destination device
* @param [in] src - Source device pointer
* @param [in] srcDevice - Source device
* @param [in] srcDeviceId - Source device
* @param [in] sizeBytes - Size of memory copy in bytes
*
* Returns #hipSuccess, #hipErrorInvalidValue, #hipErrorInvalidDevice
*/
hipError_t hipMemcpyPeer ( void* dst, int dstDevice, const void* src, int srcDevice, size_t sizeBytes );
hipError_t hipMemcpyPeer (void* dst, int dstDeviceId, const void* src, int srcDeviceId, size_t sizeBytes);
/**
* @brief Copies memory from one device to memory on another device.
@@ -961,7 +961,7 @@ hipError_t hipMemcpyPeer ( void* dst, int dstDevice, const void* src, int srcD
* Returns #hipSuccess, #hipErrorInvalidValue, #hipErrorInvalidDevice
*/
#if __cplusplus
hipError_t hipMemcpyPeerAsync ( void* dst, int dstDevice, const void* src, int srcDevice, size_t sizeBytes, hipStream_t stream=0 );
hipError_t hipMemcpyPeerAsync ( void* dst, int dstDeviceId, const void* src, int srcDevice, size_t sizeBytes, hipStream_t stream=0 );
#else
hipError_t hipMemcpyPeerAsync(void* dst, int dstDevice, const void* src, int srcDevice, size_t sizeBytes, hipStream_t stream);
#endif